Author SHA1 Message Date
jschoubben 015334020a Merge main (delivery 1, the parse-only sdk) into L3's re-land, and pin mesh-sdk 2da0de7, the release's final head (issue 774)
mesh/delivery ready: it is merged on the Approve
mesh/check pass 015334020ad2 on a3c098a5c618 | unit 5168 run, 4873 kept, 63 skipped | integration 6041 run, 4903 kept, 71 skipped | over bound: unit, …
mesh/delivery-group group feat/774-l3-reland ready: every member ready, and composed together they pass
The conflicts were go.mod, go.sum and testdata/beside/CAPTURED alone, taken from the release. 2da0de7 is 6e4a301's code;
its conformance events are unchanged.
2026-10-30 01:13:38 +01:00
mesh-tools a3c098a5c6 Merge pull request 'Pin the parse-only mesh-sdk in the node-engine, so every node can compose design 47's manifests (issue 774)' (#176) from fix/774-new-words-first into main 2026-10-30 00:10:21 +00:00
jschoubben 2fd165f5af Pin the parse-only mesh-sdk 78c9ec8 in place of 6e4a301, so main's engine reads the release's new words and means nothing new by them (issue 774)
mesh/delivery merged: source.main moved to novox/mesh-host@a3c098a5c618
mesh/delivery-group group fix/774-new-words-first merged: every member is merged
mesh/check pass 2fd165f5af8e on 8c5af3e4a3a3 | unit 9174 run, 546 kept, 51 skipped | integration 9952 run, 659 kept, 60 skipped | over bound: unit, in…
78c9ec8 is sdk main 307b623, whose tree is main's pin dbcaa29, with one commit that parses the new words (a credential's
form, an artifact's bus, a log's max-age-days) and names the new seats; an offer with no form still means pair. No code
of ours changes. Unit and integration suites pass against a local bus and main's catalogue and controller beside: 1260
tests, 7 skipped for a container or a throwaway machine.
2026-10-30 00:39:20 +01:00
jschoubben db1151c773 Pin mesh-sdk 6e4a301 on main's engine, so the controller and the engine read the release's new manifest words before the release uses them (issue 774)
No code of ours changes. Whole unit and integration suites pass at main's code against a local bus: 1260 tests, 7
skipped for a container or a throwaway machine.
2026-10-30 00:25:10 +01:00
jschoubben 02f78c9949 Pin mesh-sdk 6e4a301, the release's final sdk (issue 774)
The whole integration suite passes at this head against a local bus (JetStream, and a bus with the witness grants) and
the release's mesh-catalog and mesh-controller beside it: 1260 tests, 7 skipped for a container or a throwaway machine.
2026-10-30 00:17:00 +01:00
jschoubben 192633edb8 Pin mesh-sdk 007b6fd, the release's final sdk (issue 774) 2026-10-30 00:07:07 +01:00
jschoubben e5e8ff5783 Say each account's root verdict and the engine's contract in the node's health report, so the self-check's DA and D-root judge without the controller (issue 1952, ADR 0266)
Every account resource marked root never goes into NodeHealth.Root; the statement's contract into Contract. Relogin,
root and found are judged each on its own, so an account may be both. Pins mesh-sdk f3677eb. Test: the report's row in
TestAHealthStatementIsSaidOnTheLinkOpenNow carries the agent account's verdict and the contract.
2026-10-29 23:49:36 +01:00
jschoubben c3ff18de4e Name the machines a failing network part points at in the node's health report, so the watchman holds it where the controller does (issue 774)
delivery.NetworkPart.Targets (mesh-sdk 6eae076) is filled where the report is made, by the controller's own rule:
- the hub's node for "hub";
- each resolver's node by its address on the private network;
- none when the part points at nothing, at this machine, or at an address no other mesh machine has.

The names come from the placements the node's own composer holds (Decided.Address, Decided.Hub). A spoke's tunnel names
only the hub, so the tunnel cannot name a resolver on another spoke. A node that does not compose its own names no
target.

Test: rows in TestAHealthStatementIsSaidOnTheLinkOpenNow (hub, two resolvers, a VPN's resolver, its own, a healthy
part).
2026-10-29 23:05:02 +01:00
jschoubben 3ed9abf658 Format the health report's aliases as gofmt does (issue 774) 2026-10-29 22:59:38 +01:00
jschoubben 552f97c919 Take the node's health report layout from mesh-sdk 94d0de5, so the engine and the watchman read one form (issue 774)
delivery.NodeHealth and its parts replace this package's copies. The message's network and units parts are aliases of
the same types, so its JSON is unchanged. The engine keeps the derivation from its statement and the write on a change.
Test: the key's bytes, decoded and encoded again by the sdk, are the same bytes.
2026-10-29 22:59:22 +01:00
jschoubben 65fbc1e6f8 Write each node's health report into its own key of mesh-host_health on a change, so the watchman raises the machine's alerts without the controller (issue 774, the operator's alert table)
Until now the six machine alerts came from a health message passing to the controller. The report is now kept in the
engine's declared state `health`, key `<node>`, created by the grants part as `acts` is. Only this node's engine
writes it, after comparing it on every look and only when what it says changed; it is read first after a restart.

It carries:
- the network parts (machine-network; network-rewritten through the resolver file part's writer and owner);
- the tunnel's peers, each seen within five minutes or not (network-unreachable);
- the failed units no module owns (machine-units);
- the accounts marked relogin needed (relogin-needed);
- the directories used as found (directory-used-as-found).

The streaks and the time of the look are left out, so a look alone writes nothing. A peer is reported seen or not,
never its handshake time, so a renewed handshake writes nothing.

The layout is internal/link's (NodeHealth) until it moves into mesh-sdk. Pins mesh-sdk 58a8324.

Tests:
- rows in TestOnTheHubAnyFreshPeerIsAHealthyTunnel: the peers, and a change only when one is seen or lost;
- a case in TestAHealthStatementIsSaidOnTheLinkOpenNow: what the report takes, and one write per change across a
  restart;
- the manifest's states in TestTheManifestConsumesTheBackupEvent.
2026-10-29 22:56:14 +01:00
jschoubben cbe8528f91 Resolve ${secret:<name>} through this node's provision keys, and check the reference by the vault's own name (issue 1954 L6-E1, E2; ADR 0570 clauses 1, 6; ADR 0094)
- A placeholder is `${secret:<name>}`, the name the module calls the credential (catalogue.SecretLocal): the local
  name where it keeps several values of one provision, else the provision. The `<provision>.<local>` form goes.
- The engine finds the one key under `<node>.<module>.` of mesh-provisioner's `provisions` going by that name. That is
  what node.<node> may read, by a watch to its caught-up mark; plans are not read.
- The reference's ID must be a credential name of that provision (provisions.VaultName). The vault's answer must carry
  exactly that name and generation. Anything else refuses the write, naming both.

Pins mesh-sdk 66ff33d. Tests: rows in TestAWriteAndAnApplyOfTheSameFileSerialise (a local name, a name no provision
goes by, a reference to no credential of the provision, a mismatched answer).
2026-10-29 22:41:44 +01:00
jschoubben 5497d362dd Refuse a provision's value unless the vault's credential name and generation are the ones its reference names (issue 1954 L6-E1; ADR 0570 clauses 1, 6)
The value written is the one the provision names: a1-47 makes Reference.ID the vault's credential name. A mismatch in
either refuses the write, naming both. Test: a row in TestAWriteAndAnApplyOfTheSameFileSerialise.
2026-10-29 22:37:44 +01:00
jschoubben 6b32e07164 Fill a provision's ${secret:} at the write from the vault's value sealed to this machine, so the provisioner's client never holds it (issue 1954 L6-E1, E2; ADR 0570 clause 6)
The provisioner's client fills a handed module's file but for each `${secret:<provision>}` (or a whole `secrets` file
that is the placeholder alone).
- The node-engine reads the provision's key `<node>.<consumer>.<provision>[.<local>]` in mesh-provisioner's
  `provisions`. The key must be `provided` and carry the vault's reference.
- It asks the vault's `own` for that pair credential sealed to this machine, at least at the reference's generation,
  and opens it with the node's sealing key.
- It writes the file as declared (its owner), 0600 when the declaration gives no mode.
- Anything short of that refuses the write, naming the provision and why.

A module the node composes itself already has its `secrets` files written so: its composition carries the vault's
value from `own`, written 0600 with the declared owner.

Pins mesh-sdk 4a96793 (go/provisions). Tests: rows in TestAWriteAndAnApplyOfTheSameFileSerialise.
2026-10-29 22:36:53 +01:00
jschoubben f7814853f9 Pin mesh-sdk 2e5cc4f, the re-land group's sdk head (issue 774) 2026-10-29 22:18:45 +01:00
jschoubben 58aacaad2d Pin mesh-sdk 7000bd5, the re-land group's sdk head: a tools bundle declares its own bus account (issue 1952 S5, issue 774) 2026-10-29 21:57:09 +01:00
jschoubben 81672b183d Keep the node's own as it runs from an engine's first start, and refuse the node's own apply removing a module no want removed (issue 774, review round 1)
The reviewer showed how 76e214a still lets the 07:54 outage happen at the re-landed engine's first start on novox.
applied/ is empty there, so loadApplied loads nothing. The kept declaration retires, because every placed module has a
have. Nothing then carries the node's own composition. The guard misses nats, which has no want there, so its
nats.conf is guessed as nats's and removed.

- At the start, when the store has no node's own composition, the node's own runs as the kept declaration has it, and
  the store keeps it now. The kept declaration retires only once the store holds `node`.
- What this engine applied runs until loadHad reads the composition a differing have names. The bucket no longer keeps
  the one novox's have names, so the stored one stays.
- In the node's own apply, a resource of another module goes only when that module's want says removed
  (Scope.RemovedByWant). A module the node's own composition carried has no want, so Placed never named it.
  Cost: the node's own apply cannot drop one of its own dotted resources without such a want. It is refused whole and
  said, never removed silently.

Tests, with no new test functions:
- A first-start case in the agent's Everything test (empty store, then a start with the declaration retired). It fails
  on 76e214a.
- Two rows in the guard's table: no want refuses; a want saying removed removes.
- A case for the store in the authority test.
2026-10-29 21:25:51 +01:00
jschoubben 76e214a8b9 Keep every composition the engine applied on the machine, so a module whose composition the bus no longer keeps runs as applied and is never composed as nothing (issue 774)
On novox on 2026-10-29 at 07:54 UTC the kept declaration was retired and the node's own composition was in no bucket.
The composer's first pass reads what runs before the agent reads any have (the agent waits for the composer's caught,
sent after that pass), so the node's own was composed with its folded modules running nothing, and the bus server and
four held modules were removed.

- Quick fix: read the node's own have in loadHad too. It still waits for the composer's first pass, and still finds
  nothing when the bucket no longer keeps the composition the have names.
- Thorough fix, chosen: the engine keeps each composition it applied (the node's own too) in `applied/` beside its
  state, 0600. The agent reads it at Run's start, before any watch, as what runs ahead of the bucket and the kept
  declaration. A have naming another running composition wins. An apply keeps; a removal forgets.

Test: a case in TestEverythingIsWhatTheAgentRunsAndWhatIsKeptForTheRest (no new test function). It fails on L3's
engine 4f9227a (0 compositions at the first own composition) and passes here.
2026-10-29 21:16:38 +01:00
jschoubben acf02d52ff Pin mesh-sdk 22c9ddd, so a carried or held module serves its tools from what runs and a data flag alone moves no data (issues 1126, 924) 2026-10-29 21:05:12 +01:00
jschoubben a7acee3eed Refuse an apply that would remove a resource of a module a want still places here, so a composition that lost what runs can never remove the bus server or a held module again (issue 774) 2026-10-29 21:04:17 +01:00
jschoubben d2645a27e9 Compose a file that runs as the node last wrote it, so a stale kept declaration's export no longer names a swept object (issue 924)
novox's mesh-vault has failed since 01:16 on 2026-10-29: its have says failed, so loadHad (fbea777) takes no
composition from it, and what runs falls back to the kept declaration, whose export (mesh-vault.kept) names
f099eaa1, an object the controller's sweep removed. The node wrote a newer export since. The composer carries that
file as it runs (mesh-sdk frombus keptOf), so what runs now names the digest and size of what is on disk, but only
where the node's record says it wrote exactly that content at that path (Applied.Wrote, Target). The apply then
finds the file in place and needs no object.

Quick fix weighed: have the controller put the object back. Thorough (chosen, in the node): what runs is what the node
wrote, never a guess from an older declaration. The rows join the existing whole-apply test; the test count stays.
2026-10-29 21:04:17 +01:00
jschoubben 7d39510450 Keep the content of every file given by digest on the node, so an object the controller's sweep removed never refuses a composition the node already had (issue 924)
On novox at 01:16 on 2026-10-29, after the node began composing its own modules, mesh-vault's apply failed while the
vault ran: its own composition carries the export as it runs (mesh-sdk frombus KeptRuns) by a digest whose object the
controller's sweep (mesh-controller cmd/mesh-controller/declaration_sweep.go, sweepNode and sweepDeclarationFiles)
had removed, since the sweep keeps only what the controller's compositions in composed-<node> and the wants name.
The node may only read its bucket (ADR 0486 §2), so the node keeps what it had: a file here that verifies, or an
object fetched, goes to <state dir>/objects under the object's name, and a kept one is used before the bus is asked.
An object unused for 14 days is let go. A fetch that still fails says the keep never held it, before its cause.

Quick fix weighed: re-put the object from the controller's store. Thorough (chosen): the node never depends on the
bucket for content it once had, for every module, since the sweep and a node composing its own can always disagree.
The whole-apply test gains three rows that fail on the old code; the test count stays.
2026-10-29 21:04:17 +01:00
jschoubben f38ac87ac9 What runs from the kept declaration says Data as the module's manifest does, and only a have that runs counts at the start (issue 924)
The kept-declaration fallback built a composition without Data, so after a restart a data module (mongodb, mssql
on novox) was judged to move its data against its first composition and the operator was asked a false yes. Now
every such composition takes Data from the manifest (agent.Env.HoldsData, the composer's DataItems), and a module
whose manifest cannot be read is said to be unjudgeable rather than taken to hold none. The composition a have
names comes before the kept declaration in the judgement too.

loadHad takes a have only in the phases baseline takes as running (running, held, a composer's mark): one backing
up, applying, checking, failed or waiting names a composition not applied (review of 031e106). Two rows added to
the same table test; the count stays.
2026-10-29 21:04:17 +01:00
jschoubben d8de0a8853 Judge a module the agent has not taken yet by the composition its have names, and knock the filling at the start and whenever a module is taken as running (issue 924)
A held module on a node composing its own never gets a composition, so the agent never took it and the filling
judged its files by the kept declaration, older than power's three new drop-ins: refused from every engine start
on ace, g14 and shanks until some module settled (2026-10-29 00:09 UTC). The second meeting of issue 924.

Quick fix: knock the filling once the haves are read at the start. Thorough fix (chosen, the kind recurs for every
file added since the switch): what runs of such a module is the composition its have names, from the bucket or
its revisions, ahead of the kept declaration; the filling is knocked on every change of what runs, not only on a
settle. One unit test replaced by a table of four cases; the count stays.
2026-10-29 21:04:17 +01:00
jschoubben 45d0348044 Revert "Merge pull request 'Revert L3: its node-engine removed every held module on novox, the bus server among them (issue 774)' (#174) from revert/774-l3 into main"
This reverts commit 1ec73915b8, reversing
changes made to 0a4c35f731.
2026-10-29 21:04:13 +01:00
mesh-tools 8c5af3e4a3 Merge pull request 'Pin mesh-sdk dbcaa29, so a held module's hold no longer spreads to the modules that read from it (issue 1960)' (#175) from fix/1960-no-hold-spread into main 2026-10-29 15:35:33 +00:00
jschoubben ece607e215 Pin mesh-sdk dbcaa29, so a held module's hold no longer spreads to the modules that read from it (issue 1960)
mesh/delivery merged: source.main moved to novox/mesh-host@8c5af3e4a3a3
mesh/delivery-group group fix/1960-no-hold-spread merged: every member is merged
mesh/check pass ece607e21522 on 1ec73915b807 | unit 3318 run, 1951 kept, 38 skipped | integration 3994 run, 2125 kept, 38 skipped | over bound: unit, …
2026-10-29 16:00:21 +01:00
jschoubben 86a05babed Expect proxy handed beside held de-spiegel: a hold no longer spreads (issue 1960)
mesh-sdk's composer no longer holds a module behind a held one it reads from; the selfcompose test now expects
proxy handed, composed without de-spiegel's routes, and de-spiegel told alone. Needs mesh-sdk pinned at the fix.
2026-10-29 15:58:19 +01:00
mesh-tools 1ec73915b8 Merge pull request 'Revert L3: its node-engine removed every held module on novox, the bus server among them (issue 774)' (#174) from revert/774-l3 into main 2026-10-29 08:59:58 +00:00
47 changed files with 4753 additions and 216 deletions
+107
View File
@@ -0,0 +1,107 @@
package main
import (
"errors"
"fmt"
"os"
"path/filepath"
"regexp"
"strings"
)
// appliedStore keeps every composition the agent applied, one file per module (the node's own as `node.json`), in the
// directory `applied` beside the node's state (agent.Env.Applied, issue 774). It is the machine's own record of what runs:
// read at the agent's start before anything is watched, so the node never composes its own as if a module it runs ran
// nothing because the bucket no longer keeps that composition or the kept declaration is retired. Files are 0600 in a
// 0700 directory: a composition carries values sealed to this machine.
type appliedStore struct{ dir string }
func appliedStoreFor(statePath string) appliedStore {
return appliedStore{dir: filepath.Join(filepath.Dir(statePath), "applied")}
}
// appliedName is a module name the store takes as a file name: a catalogue module's or the node's own, never a path.
var appliedName = regexp.MustCompile(`^[a-z0-9][a-z0-9._-]*$`)
func (s appliedStore) path(module string) (string, error) {
if !appliedName.MatchString(module) || strings.Contains(module, "..") {
return "", fmt.Errorf("%q is not a module name the applied store takes", module)
}
return filepath.Join(s.dir, module+".json"), nil
}
// holds says whether the store keeps the module's composition.
func (s appliedStore) holds(module string) bool {
path, err := s.path(module)
if err != nil {
return false
}
_, err = os.Stat(path)
return err == nil
}
func (s appliedStore) Load() (map[string][]byte, error) {
entries, err := os.ReadDir(s.dir)
if errors.Is(err, os.ErrNotExist) {
return nil, nil
}
if err != nil {
return nil, err
}
out := map[string][]byte{}
var errs []error
for _, e := range entries {
module, ok := strings.CutSuffix(e.Name(), ".json")
if !ok || e.IsDir() || !appliedName.MatchString(module) {
continue
}
raw, err := os.ReadFile(filepath.Join(s.dir, e.Name()))
if err != nil {
errs = append(errs, err)
continue
}
out[module] = raw
}
return out, errors.Join(errs...)
}
func (s appliedStore) Keep(module string, raw []byte) error {
path, err := s.path(module)
if err != nil {
return err
}
if was, err := os.ReadFile(path); err == nil && string(was) == string(raw) {
return nil
}
if err := os.MkdirAll(s.dir, 0o700); err != nil {
return err
}
tmp, err := os.CreateTemp(s.dir, "."+module+".*")
if err != nil {
return err
}
defer os.Remove(tmp.Name())
if _, err := tmp.Write(raw); err != nil {
tmp.Close()
return err
}
if err := tmp.Sync(); err != nil {
tmp.Close()
return err
}
if err := tmp.Close(); err != nil {
return err
}
return os.Rename(tmp.Name(), path)
}
func (s appliedStore) Drop(module string) error {
path, err := s.path(module)
if err != nil {
return err
}
if err := os.Remove(path); err != nil && !errors.Is(err, os.ErrNotExist) {
return err
}
return nil
}
+143 -2
View File
@@ -1,16 +1,27 @@
package main
import (
"context"
"crypto/ed25519"
"crypto/x509"
"encoding/base64"
"encoding/json"
"encoding/pem"
"os"
"path/filepath"
"testing"
"time"
"git.novox.be/novox/mesh-sdk/go/buscompose"
"git.novox.be/novox/mesh-sdk/go/delivery"
"git.novox.be/novox/mesh-sdk/go/overlay"
"github.com/novox/mesh-host/internal/identity"
)
// Issue 1126 (ADR 0555 §3): the mesh's own certificate a composition carries is kept for the machine's record, and the
// record's keys read it; a composition carrying none changes nothing.
// record's keys read it; a composition carrying none changes nothing. And (L3 s6, s7; ADR 0550 §2) the record says what
// the node trusts a mesh signature from, and the node keeps the certificate the vault signed for it.
func TestTheAppliedAuthorityIsKeptForTheRecord(t *testing.T) {
state := filepath.Join(t.TempDir(), "state", "state.json")
res := func(id, content string) json.RawMessage {
@@ -26,7 +37,137 @@ func TestTheAppliedAuthorityIsKeptForTheRecord(t *testing.T) {
if err := keepAuthority(state, delivery.Composed{Resources: []json.RawMessage{res("ca-trust.certificate-authority", "CA-1")}}); err != nil {
t.Fatal(err)
}
if got := recordKeys(state)(); got.Authority != "CA-1" {
if got := recordKeys(state, nil)(); got.Authority != "CA-1" {
t.Errorf("the record's keys read the authority %q", got.Authority)
}
// L3 s6: before the vault's signers are read the node trusts the controller's key alone; once read, a message the
// vault signed is accepted and the record names both; once the lease is handed away the controller's key is refused
// and the record names the vault's alone. Each change tells the record's writer.
controllerPub, controller, _ := ed25519.GenerateKey(nil)
vaultPub, vault, _ := ed25519.GenerateKey(nil)
trust := newMeshTrust(controllerPub)
told := 0
trust.onChange(func() { told++ })
keys := recordKeys(state, trust.Trusts)
ctrlFP, vaultFP := delivery.SignerFingerprint(controllerPub), delivery.SignerFingerprint(vaultPub)
if got := keys().Trusts; !delivery.TrustsKey(got, delivery.SignerController, ctrlFP) || delivery.TrustsKey(got, delivery.SignerVault, vaultFP) {
t.Errorf("before the signers are read the record trusts %v", got)
}
enrol, _ := delivery.SignContext(delivery.KindEnrolment)
msg := []byte("a reply")
if _, err := trust.Verify(delivery.KindEnrolment, msg, delivery.SignMessage(vault, enrol, msg)); err == nil {
t.Error("a vault-signed message is accepted before the vault's key is read")
}
vaultCA := testAuthority(t, vaultPub, vault)
keyValue, _ := json.Marshal(delivery.NewSignerKey(vaultPub, time.Now()))
fp, _ := delivery.AuthorityFingerprint(vaultCA)
authValue, _ := json.Marshal(delivery.SignerAuthority{Certificate: vaultCA, Fingerprint: fp, Made: time.Now()})
trust.heardSigner(delivery.SignersKeyKey, keyValue, false, func(string) {})
trust.heardSigner(delivery.SignersAuthorityKey, authValue, false, func(string) {})
if s, err := trust.Verify(delivery.KindEnrolment, msg, delivery.SignMessage(vault, enrol, msg)); err != nil || s.Name != delivery.SignerVault {
t.Errorf("a vault-signed message on a node that read signers: %+v, %v", s, err)
}
if got := keys().Trusts; !delivery.TrustsKey(got, delivery.SignerController, ctrlFP) || !delivery.TrustsKey(got, delivery.SignerVault, vaultFP) ||
trust.VaultAuthority() != vaultCA || told != 2 {
t.Errorf("after the signers are read the record trusts %v (told %d)", got, told)
}
handed, _ := json.Marshal(buscompose.LeaseHolder{HandedTo: "mesh-vault", HandedBy: "person.operator"})
trust.heardLease(handed, true, nil)
if _, err := trust.Verify(delivery.KindEnrolment, msg, delivery.SignMessage(controller, enrol, msg)); err == nil || trust.Controller() != nil {
t.Error("after the lease is handed away the controller's key is still trusted")
}
if got := keys().Trusts; delivery.TrustsKey(got, delivery.SignerController, ctrlFP) || !delivery.TrustsKey(got, delivery.SignerVault, vaultFP) || told != 3 {
t.Errorf("after the lease is handed away the record trusts %v (told %d)", got, told)
}
trust.heardLease(handed, true, nil)
if told != 3 {
t.Error("a lease heard again unchanged wrote the record again")
}
// A signers' key naming another fingerprint is not trusted.
bad, _ := json.Marshal(delivery.SignerKey{Key: base64.StdEncoding.EncodeToString(controllerPub), Fingerprint: vaultFP})
trust.heardSigner(delivery.SignersKeyKey, bad, false, func(string) {})
if !delivery.TrustsKey(keys().Trusts, delivery.SignerVault, vaultFP) {
t.Error("a signers' key that does not check replaced the vault's")
}
// L3 s7: the certificate the vault signs for this node's internal name and serving key is kept where the record reads
// it, and the record is written again; one signed by another authority is refused and nothing is kept.
serving, err := identity.GenerateServingKey()
if err != nil {
t.Fatal(err)
}
if err := identity.WriteServingKey(identity.ServingKeyPath(state), serving); err != nil {
t.Fatal(err)
}
servingPub, _ := base64.StdEncoding.DecodeString(serving.Public)
signedBy := func(ca string, key ed25519.PrivateKey) string {
parent, _ := delivery.ParseCertificate(ca)
der, err := x509.CreateCertificate(nil, delivery.NodeCertificateTemplate(overlay.InternalName("n1"), servingPub, time.Now().Add(-time.Hour)),
parent, ed25519.PublicKey(servingPub), key)
if err != nil {
t.Fatal(err)
}
return string(pem.EncodeToMemory(&pem.Block{Type: "CERTIFICATE", Bytes: der}))
}
var asked []any
answer := signedBy(vaultCA, vault)
ask := func(_ context.Context, verb string, args any) (json.RawMessage, error) {
asked = append(asked, verb, args)
return json.Marshal(delivery.SignAnswer{Kind: delivery.KindNodeCertificate, Node: "n1", Certificate: answer})
}
written := 0
if err := certifyThisNode(context.Background(), state, "n1", trust, ask, func() { written++ }); err != nil {
t.Fatal(err)
}
if got := recordKeys(state, nil)(); got.Certificate != answer || written != 1 || len(asked) != 2 || asked[0] != delivery.SignVerb {
t.Errorf("the vault's certificate: kept %v, written %d, asked %v", got.Certificate == answer, written, asked)
}
answer = signedBy(testAuthority(t, controllerPub, controller), controller)
if err := certifyThisNode(context.Background(), state, "n1", trust, ask, func() { written++ }); err == nil || written != 1 {
t.Errorf("a certificate of another authority was kept: %v", err)
}
// Issue 774: what the engine applied is kept beside the state too, one file per module, 0600 in a 0700 directory,
// written by a rename so a crash leaves the one before; a name that is no module's is refused, never a path.
t.Run("the compositions the engine applied are kept beside the state", func(t *testing.T) {
applied := appliedStoreFor(filepath.Join(t.TempDir(), "state", "state.json"))
if m, err := applied.Load(); err != nil || len(m) != 0 || applied.holds("node") {
t.Fatalf("an empty store loads %v, %v", m, err)
}
for _, tc := range []struct {
module string
takes bool
}{{"node", true}, {"nats", true}, {"mesh-delivery-view", true}, {"../state", false}, {"a/b", false}, {"..", false}, {"", false}} {
if err := applied.Keep(tc.module, []byte(`{"v":1}`)); (err == nil) != tc.takes {
t.Errorf("keeping %q: %v, want taken %v", tc.module, err, tc.takes)
}
}
if err := applied.Keep("node", []byte(`{"v":2}`)); err != nil {
t.Fatal(err)
}
entries, _ := os.ReadDir(applied.dir)
dirInfo, _ := os.Stat(applied.dir)
fileInfo, _ := os.Stat(filepath.Join(applied.dir, "node.json"))
if len(entries) != 3 || dirInfo.Mode().Perm() != 0o700 || fileInfo.Mode().Perm() != 0o600 {
t.Fatalf("the store holds %d files (want 3, no temporary one left), directory %v, file %v", len(entries), dirInfo.Mode(), fileInfo.Mode())
}
if err := applied.Drop("nats"); err != nil || applied.Drop("nats") != nil || applied.holds("nats") {
t.Fatalf("a dropped module is still held, or dropping twice fails: %v", err)
}
if m, err := applied.Load(); err != nil || string(m["node"]) != `{"v":2}` || len(m) != 2 {
t.Fatalf("the store loads %v, %v", m, err)
}
})
}
// testAuthority is a self-signed authority certificate for key, as the vault makes its own.
func testAuthority(t *testing.T, pub ed25519.PublicKey, key ed25519.PrivateKey) string {
t.Helper()
tpl := delivery.AuthorityTemplate(pub, time.Now().Add(-time.Hour))
der, err := x509.CreateCertificate(nil, tpl, tpl, pub, key)
if err != nil {
t.Fatal(err)
}
return string(pem.EncodeToMemory(&pem.Block{Type: "CERTIFICATE", Bytes: der}))
}
+277
View File
@@ -0,0 +1,277 @@
package main
import (
"bufio"
"context"
"errors"
"fmt"
"os"
"path/filepath"
"strings"
"sync/atomic"
"syscall"
"time"
stdio "git.novox.be/novox/mesh-sdk/go"
meshasker "git.novox.be/novox/mesh-sdk/go/asker"
"git.novox.be/novox/mesh-sdk/go/asks"
"github.com/novox/mesh-host/internal/accounts"
"github.com/novox/mesh-host/internal/apply"
"github.com/novox/mesh-host/internal/link"
"github.com/novox/mesh-host/internal/store"
)
// The node seat node-engine's judgements on this machine (novox/hq ADR 0563): what a hand-over and a library's take
// would set and what the disk holds now, and whether this machine is root-free.
// diskOf is what is on disk at a directory, as an act compares it when the operator's yes arrives: its owner and
// group, its mode, and which file it is.
func diskOf(info os.FileInfo) string {
owner := "unknown"
if uid, gid, ok := apply.OwnerOf(info); ok {
owner = fmt.Sprintf("%d:%d", uid, gid)
}
which := ""
if st, ok := info.Sys().(*syscall.Stat_t); ok {
which = fmt.Sprintf(" %v:%v", st.Dev, st.Ino)
}
return fmt.Sprintf("%s %o%s", owner, info.Mode().Perm(), which)
}
// handOverJudgeFrom judges a hand-over through the node seat as the controller's ask is judged (judgeFound): the
// declared owner and mode it will set, and the disk as it is.
func handOverJudgeFrom(statePath string) func(string) (link.Judged, error) {
return func(path string) (link.Judged, error) {
found, info, err := judgeFound(statePath, path)
if err != nil {
return link.Judged{}, err
}
return link.Judged{Subject: path, Path: path, Sets: orUnsaid(found.Declared), Disk: diskOf(info),
Words: "A directory of " + moduleOf(found.ID) + " is used as found: the mesh left its owner and mode."}, nil
}
}
func moduleOf(id string) string {
if at := strings.LastIndex(id, "."); at > 0 {
return id[:at]
}
return id
}
// accessJudgeFrom judges the take act on a library through the node seat (ADR 0328 §7, ADR 0563 §4): declared on this
// machine, its home, and its root holding data the mesh has not taken; what the take sets, and the disk as it is.
func accessJudgeFrom(statePath string) func(string, string) (link.Judged, error) {
return func(act, access string) (link.Judged, error) {
d, err := declaredNow(statePath, nil)
if errors.Is(err, store.ErrNothingDeclared) {
return link.Judged{}, errors.New("this node keeps no declaration, so it is declared no library")
}
if err != nil {
return link.Judged{}, fmt.Errorf("what runs on this node does not read: %w", err)
}
for _, l := range d.Libraries {
if apply.LibraryAccess(l.Library) != access {
continue
}
if !l.Home {
return link.Judged{}, fmt.Errorf("this machine is not the home of the library %s, and a library is "+
"taken on its home", l.Library)
}
ctx, cancel := context.WithTimeout(context.Background(), lookBound)
defer cancel()
if v := apply.JudgeLibrary(ctx, l); v.Kind != apply.VerdictWaitKeep {
return link.Judged{}, fmt.Errorf("the library %s does not wait for the operator's word to be kept "+
"(it is judged %s), so there is nothing to take", l.Library, v.Kind)
}
info, err := os.Lstat(l.Root)
if err != nil {
return link.Judged{}, fmt.Errorf("the library %s's root cannot be read: %v", l.Library, err)
}
return link.Judged{Subject: access, Path: filepath.Clean(l.Root),
Sets: "the group " + l.Group + ", group write and setgid on everything under its root",
Disk: diskOf(info), Words: "The library " + l.Library + " holds data the mesh has not taken."}, nil
}
return link.Judged{}, fmt.Errorf("the access %s names no library declared on this machine", access)
}
}
// askTakes asks the take of every library the look found waiting for the operator's word, once per library (ADR 0563
// §4): asked by the engine itself, never a second time while one is open or after one was asked; access-act asks again
// by hand.
func askTakes(acts *link.HandActs, waiting []string, say func(string)) {
if acts == nil {
return
}
for _, access := range waiting {
if acts.AskedFor(link.ActAccess, access) {
continue
}
if _, err := acts.AskAccess(link.EngineSeat+"-"+acts.Node, link.ActTake, access,
"its root holds data outside its group", true); err != nil && say != nil {
say("the take of " + access + " could not be asked: " + err.Error())
}
}
}
// lastAccounts is the newest look at every account (the health loop's), which root-free judges from.
var lastAccounts atomic.Pointer[accounts.Statement]
// rootFreshFor is how old the verdict root-free judges from may be (ADR 0563 §6, the controller's verdictFreshFor).
const rootFreshFor = 15 * time.Minute
// judgeRootFree is this machine's root-free (ADR 0259 §8, ADR 0563 §6): free only on the engine's own verdict, at most
// rootFreshFor old, that every account declared never to become root without a person cannot, at least one such
// account declared, and the login shell's execute not served here. Anything else is no pass, and says why.
func judgeRootFree(node string, st *accounts.Statement, now time.Time, served bool, servedWhy string) link.RootVerdict {
v := link.RootVerdict{Machine: node, Judged: now.UTC()}
var not []string
confined := ""
switch {
case st == nil:
not = append(not, "this machine's engine has judged no account yet")
case now.Sub(st.At) > rootFreshFor:
not = append(not, fmt.Sprintf("the engine's newest judgement of its accounts was made at %s, more than %d "+
"minutes ago, and a verdict that old is not a verdict on now", st.At.UTC().Format(time.RFC3339),
int(rootFreshFor.Minutes())))
default:
var named []string
for _, a := range st.Accounts {
if !a.Root {
continue
}
named = append(named, a.Name)
if a.State != accounts.Healthy {
not = append(not, "the account "+a.Name+" "+orNoneKnownWord(a.Reason))
}
}
if len(named) == 0 {
not = append(not, "no account here is declared never to become root, so agents run as the operator's "+
"account, which may become root")
} else if len(not) == 0 {
confined = "the account " + strings.Join(named, ", ") + " cannot become root without a person (judged " +
st.At.UTC().Format(time.RFC3339) + ")"
}
}
if served {
not = append(not, "the login shell runs any command an agent gives it as the machine's runtime account, which "+
"can become root ("+servedWhy+")")
}
if len(not) > 0 {
v.Why = strings.Join(not, "; ")
return v
}
v.Free, v.Why = true, confined+"; the login shell's execute is not served here"
return v
}
func orNoneKnownWord(s string) string {
if s == "" {
return "is not judged"
}
return s
}
// settingFileSuffix is the resource the login shell's holder renders its setting execute into (ADR 0268: zsh's
// tools-conf, `execute=serve` or `execute=withhold`), read beside the bus's discovery.
const settingFileSuffix = ".tools-conf"
// executeServedHere is whether the login shell's execute is served on this machine (ADR 0268), failing closed: served
// while the bus's discovery hears it answered here or could not be asked, or while the holder's setting file applied
// here says anything but withhold or cannot be read. why says which.
func executeServedHere(statePath string, heard func() (bool, error)) (bool, string) {
var why []string
if heard == nil {
why = append(why, "the bus could not be asked whether execute is answered here")
} else if h, err := heard(); err != nil {
why = append(why, "the bus could not be asked whether execute is answered here: "+err.Error())
} else if h {
why = append(why, "the bus hears execute answered here")
}
known, err := store.Load(statePath)
if err != nil {
why = append(why, "what this machine applied could not be read: "+err.Error())
}
for _, r := range known.Resources {
if r.Type != "file" || !strings.HasSuffix(r.ID, settingFileSuffix) {
continue
}
if value, err := settingIn(r.Target); err != nil {
why = append(why, moduleOf(r.ID)+"'s setting execute cannot be read: "+err.Error())
} else if value != "withhold" {
why = append(why, moduleOf(r.ID)+"'s setting execute here is "+value)
}
}
return len(why) > 0, strings.Join(why, "; ")
}
// settingIn is the value of execute in a setting file, "" when it says none.
func settingIn(path string) (string, error) {
f, err := os.Open(path)
if err != nil {
return "", err
}
defer f.Close()
value := ""
lines := bufio.NewScanner(f)
for lines.Scan() {
t := strings.TrimSpace(lines.Text())
if key, v, ok := strings.Cut(t, "="); ok && !strings.HasPrefix(t, "#") && strings.TrimSpace(key) == "execute" {
value = strings.TrimSpace(v)
}
}
if value == "" {
value = "not said"
}
return value, lines.Err()
}
// keepWaiting is every library a look found waiting for the operator's word to keep it (ADR 0328 §8).
func keepWaiting(vs []link.ResourceHealth) []string {
var out []string
for _, v := range vs {
for _, w := range v.AccessWaits {
if w.Reason == link.WaitKeep {
out = append(out, w.Access)
}
}
}
return out
}
// engineActs is the node seat this engine holds (ADR 0563): judged and performed by the same code as the controller's
// signed asks, its asks to the operator over the session's link, its acts kept beside the state.
func engineActs(opts options, node string, queue *link.Session, search *accounts.SetuidCache) *link.HandActs {
state := link.BusState{Session: queue, Node: node}
h := &link.HandActs{Node: node, State: state,
JudgeHandOver: handOverJudgeFrom(opts.state), HandOver: handOverFrom(opts.state),
JudgeAccess: accessJudgeFrom(opts.state), AccessAct: accessActFrom(opts.state, opts.timeout),
JudgeDirectories: judgeDirectoriesFrom(opts.state), SetuidSearch: queue.SetuidSearch, Searching: search.Running,
RootFree: func() link.RootVerdict {
served, why := executeServedHere(opts.state, func() (bool, error) {
return queue.HeardServed("node-login-shell", "execute")
})
return judgeRootFree(node, lastAccounts.Load(), time.Now(), served, why)
},
Say: queue.SayEngine}
asker := asks.EngineAsker(node)
h.Asker = &meshasker.Client{Module: asker, State: state, Publish: queue.PublishAsk,
Record: func(id string) (*stdio.StateEntry, error) { return queue.RouterRecord(asker + "." + id) }}
return h
}
// reconcileEvery is how often the engine catches up its open asks from the router's records (a warrant heard while no
// link was open, an ask past its expiry): besides each link's opening.
const reconcileEvery = 5 * time.Minute
func reconcileActs(ctx context.Context, h *link.HandActs) {
t := time.NewTicker(reconcileEvery)
defer t.Stop()
for {
select {
case <-ctx.Done():
return
case <-t.C:
h.Reconcile()
}
}
}
+73
View File
@@ -2,6 +2,7 @@ package main
import (
"context"
"encoding/json"
"fmt"
"os"
"path/filepath"
@@ -10,6 +11,7 @@ import (
"testing"
"git.novox.be/novox/mesh-sdk/go/declaration"
"git.novox.be/novox/mesh-sdk/go/delivery"
"github.com/nats-io/nats.go"
"github.com/novox/mesh-host/internal/link"
)
@@ -82,6 +84,77 @@ func TestADeclarationWithAFileByDigestAppliesFromTheObject(t *testing.T) {
if err != nil || string(got) != string(body) {
t.Errorf("1695/AC1: the file holds %d bytes (%v), want %d", len(got), err, len(body))
}
// Issue 924 (novox, 2026-10-29 01:16): the controller's sweep removed an object a composition the node ran still named,
// and the module's apply failed while it ran. Content the node once had is its own keep's, and the bus is not asked.
gone := aFilesLink{err: fmt.Errorf("the object %s was deleted from the bucket declaration-files-novox", name)}
never := []byte(strings.Repeat("never had here\n", 6000))
for _, row := range []struct {
name string
raw []byte
onDisk []byte // nil removes the file
refused string // "" applies
}{
{"the file here drifted, its object deleted", raw, []byte("the module wrote it since"), ""},
{"the file here is gone, its object deleted", raw, nil, ""},
{"never had here, its object deleted", referenceDeclaration(target, never), nil, "never held it"},
} {
if row.onDisk == nil {
_ = os.Remove(target)
} else if err := os.WriteFile(target, row.onDisk, 0o600); err != nil {
t.Fatal(err)
}
report = applyAndKeep(link.WithFiles(context.Background(), gone), opts, row.raw, nil, nil, nil)
if row.refused == "" {
got, _ := os.ReadFile(target)
if report.Refused != "" || string(got) != string(body) {
t.Errorf("924: %s: refused %q, the file holds %d bytes; want it applied from the node's own keep",
row.name, report.Refused, len(got))
}
} else if !strings.Contains(report.Refused, row.refused) || !strings.Contains(report.Refused, "was deleted") {
t.Errorf("924: %s: refused %q; want a refusal saying the bucket and the keep both lack it", row.name, report.Refused)
}
}
// Issue 924, the live case: what runs came from the kept declaration, whose export names a digest swept long ago,
// while the node wrote a newer export since. Composed as it runs, the file names what the node wrote, and applies with
// no object; a file someone else changed since is left as named.
newer := []byte(strings.Repeat("the export the node wrote since\n", 3000))
newerName, _ := declaration.ObjectName(declaration.DigestOf(newer))
ctx = link.WithFiles(context.Background(), aFilesLink{objects: map[string][]byte{newerName: newer}})
if report = applyAndKeep(ctx, opts, referenceDeclaration(target, newer), nil, nil, nil); report.Refused != "" {
t.Fatalf("924: the newer export: %s", report.Refused)
}
stale := json.RawMessage(`{"id":"m.big","type":"file","path":"` + target + `","digest":"` + declaration.DigestOf(body) +
`","size":` + strconv.Itoa(len(body)) + `}`)
for _, row := range []struct {
name string
edit []byte // what someone else wrote there since; nil for nobody
digest string
}{
{"the file is what the node wrote", nil, declaration.DigestOf(newer)},
{"someone changed the file since", []byte("a hand edit"), declaration.DigestOf(body)},
} {
if row.edit != nil {
if err := os.WriteFile(target, row.edit, 0o600); err != nil {
t.Fatal(err)
}
}
runs := runsAsWritten([]delivery.Composed{{Node: "n", Module: "m", Resources: []json.RawMessage{stale}}}, opts.state)
var got struct {
Digest string `json:"digest"`
}
_ = json.Unmarshal(runs[0].Resources[0], &got)
if got.Digest != row.digest {
t.Errorf("924: %s: what runs names %s, want %s", row.name, got.Digest, row.digest)
}
if row.edit == nil {
raw := []byte(`{"declaration":1,"resources":[` + string(runs[0].Resources[0]) + `]}`)
if report = applyAndKeep(link.WithFiles(context.Background(), gone), opts, raw, nil, nil, nil); report.Refused != "" {
t.Errorf("924: %s: composed as it runs, refused %q; want applied with no object", row.name, report.Refused)
}
}
}
}
// 1695/AC8: the hand-run `apply <file>` and `reconcile` have no bus: a reference whose file here verifies is applied
+76 -10
View File
@@ -3,22 +3,25 @@ package main
import (
"context"
"crypto/ed25519"
"encoding/json"
"errors"
"fmt"
"path/filepath"
"sync/atomic"
"git.novox.be/novox/mesh-sdk/go/declaration"
"git.novox.be/novox/mesh-sdk/go/provisions"
"github.com/novox/mesh-host/internal/apply"
"github.com/novox/mesh-host/internal/filling"
"github.com/novox/mesh-host/internal/link"
"github.com/novox/mesh-host/internal/store"
"github.com/novox/mesh-host/internal/system"
)
// serveFilling writes filled files on the provisioner client's request for as long as this engine serves (novox/hq
// ADR 0349 §4). A socket that cannot be served is said once, and this engine goes on applying.
func serveFilling(ctx context.Context, opts options, signer ed25519.PublicKey, say func(string)) {
e := fillingEngine(opts, signer, say)
func serveFilling(ctx context.Context, opts options, signer ed25519.PublicKey, queue *link.Session, node string, say func(string)) {
e := fillingEngine(opts, signer, provisionSecretsOn(queue, node, opts.state), say)
fillingNow.Store(e)
defer fillingNow.CompareAndSwap(e, nil)
if err := filling.Serve(ctx, filling.Socket, e, say); err != nil {
@@ -30,9 +33,10 @@ func serveFilling(ctx context.Context, opts options, signer ed25519.PublicKey, s
var fillingNow atomic.Pointer[filling.Engine]
// knockFilling tells the provisioner's client to fill again what the filling refused or failed and the declaration
// applied now may let through (filling.Engine.Knock), once a module's apply or removal settled (agent.Env.Settled).
// What it told, or why it could not, is said.
func knockFilling(module string, say func(string)) {
// applied now may let through (filling.Engine.Knock), after what runs changed: a module's apply or removal settled
// (agent.Env.Settled, "<module> settled"), the haves were read at the start, or a module was taken as running
// (agent.Env.RunsChanged, issue 924). What it told, or why it could not, is said.
func knockFilling(after string, say func(string)) {
e := fillingNow.Load()
if e == nil {
return
@@ -40,16 +44,19 @@ func knockFilling(module string, say func(string)) {
told, err := e.Knock(filling.Socket)
switch {
case err != nil:
say(fmt.Sprintf("after %s settled: %v", module, err))
say(fmt.Sprintf("after %s: %v", after, err))
case told != "":
say(fmt.Sprintf("after %s settled, the provisioner's client is told to fill again: %s", module, told))
say(fmt.Sprintf("after %s, the provisioner's client is told to fill again: %s", after, told))
}
}
// fillingEngine is the write of a filled file as this engine does it: the declaration it keeps, verified, read under
// the apply's lock — the same two locks every apply takes, the process's and the one beside the state — and the file
// written through the apply (apply.WriteFilled), its outcome recorded in the node's state as an apply's is.
func fillingEngine(opts options, signer ed25519.PublicKey, say func(string)) *filling.Engine {
//
// A provision's `${secret:}` the client left in the content is filled first (secrets, novox/hq ADR 0570 clause 6), and a
// file so filled is written 0600 when the declaration gives no mode.
func fillingEngine(opts options, signer ed25519.PublicKey, secrets provisionSecrets, say func(string)) *filling.Engine {
return &filling.Engine{
Declared: func() (*declaration.Declaration, error) {
return declaredNow(opts.state, signer)
@@ -73,8 +80,13 @@ func fillingEngine(opts options, signer ed25519.PublicKey, say func(string)) *fi
if err != nil {
return "", "", err
}
report, updated, writeErr := apply.WriteFilled(ctx, built, d, known, path, content, apply.ExecRunner,
announceOr(say), apply.KeepIn(filepath.Dir(opts.state)), apply.WindowsIn(filepath.Dir(opts.state)))
content, secret, err := secrets.fill(ctx, filledModule(d, path), content)
if err != nil {
return "", "", fmt.Errorf("%s was not written: %w", path, err)
}
report, updated, writeErr := apply.WriteFilledWith(ctx, built, d, known, path, content, apply.ExecRunner,
announceOr(say), apply.KeepIn(filepath.Dir(opts.state)), apply.WindowsIn(filepath.Dir(opts.state)),
apply.Filled{Secret: secret})
// Saved whichever way it went, as an apply's: what was written is on the machine either way.
if err := store.Save(opts.state, updated); err != nil {
return "", "", errors.Join(writeErr, errors.New("written, and the node's state could not be saved: "+
@@ -93,3 +105,57 @@ func fillingEngine(opts options, signer ed25519.PublicKey, say func(string)) *fi
},
}
}
// filledModule is the module whose file the declaration declares at path, or "" when none or several do.
func filledModule(d *declaration.Declaration, path string) string {
module := ""
for _, r := range d.Resources {
if f, ok := r.(*declaration.File); ok && f.Path == path {
if module != "" {
return ""
}
module, _ = d.ModuleOfID(f.ID)
}
}
return module
}
// provisionSecretsOn reads provisions and asks the vault over the link, and opens with this machine's sealing key.
func provisionSecretsOn(queue *link.Session, node, statePath string) provisionSecrets {
p := provisionSecrets{node: node, open: sealOpener(statePath)}
if queue == nil {
return p
}
// What the bus holds under prefix, read by a watch to its caught-up mark: no poll, no list of every key.
p.list = func(ctx context.Context, prefix string) (map[string][]byte, error) {
wctx, cancel := context.WithCancel(ctx)
defer cancel()
updates, err := queue.KVWatch(wctx, provisions.ProvisionsBucket, prefix+">")
if err != nil {
return nil, err
}
out := map[string][]byte{}
for {
select {
case <-ctx.Done():
return nil, ctx.Err()
case u, ok := <-updates:
if !ok {
return nil, fmt.Errorf("the watch of %s ended before it caught up", provisions.ProvisionsBucket)
}
switch {
case u.Caught:
return out, nil
case u.Deleted:
delete(out, u.Key)
default:
out[u.Key] = u.Value
}
}
}
}
p.own = func(ctx context.Context, args map[string]any) (json.RawMessage, error) {
return queue.AskVault(ctx, "own", args)
}
return p
}
+102 -1
View File
@@ -7,9 +7,12 @@ import (
"os"
"os/user"
"path/filepath"
"strings"
"testing"
"time"
"git.novox.be/novox/mesh-sdk/go/provisions"
"github.com/novox/mesh-host/internal/filling"
"github.com/novox/mesh-host/internal/store"
)
@@ -45,7 +48,34 @@ func TestAWriteAndAnApplyOfTheSameFileSerialise(t *testing.T) {
store.Declared{Declaration: raw, Signature: ed25519.Sign(private, raw)}); err != nil {
t.Fatal(err)
}
e := fillingEngine(opts, public, func(string) {})
// The provision m reads, as mesh-provisioner keeps it, and the vault's value of it, sealed to this machine (faked:
// the opener reads the box's "v").
provision := map[string]string{}
vaultName := "provision/n2/db"
secrets := provisionSecrets{node: "n1",
list: func(_ context.Context, prefix string) (map[string][]byte, error) {
out := map[string][]byte{}
for k, v := range provision {
if strings.HasPrefix(k, prefix) {
out[k] = []byte(v)
}
}
return out, nil
},
own: func(_ context.Context, args map[string]any) (json.RawMessage, error) {
raw, _ := json.Marshal(args)
var a struct{ Credentials []map[string]any }
_ = json.Unmarshal(raw, &a)
c := a.Credentials[0]
c["name"], c["generation"], c["sealed"] = vaultName, 2, map[string]string{"v": "s3cret"}
return json.Marshal(map[string]any{"node": "n1", "credentials": []any{c}})
},
open: func(sealed string) ([]byte, error) {
var b struct{ V string }
err := json.Unmarshal([]byte(sealed), &b)
return []byte(b.V), err
}}
e := fillingEngine(opts, public, secrets, func(string) {})
uid := uint32(os.Getuid())
if err := e.Accepts(uid); err != nil {
t.Fatal(err)
@@ -102,4 +132,75 @@ func TestAWriteAndAnApplyOfTheSameFileSerialise(t *testing.T) {
if a := <-done; a.Action != "updated" {
t.Fatalf("the second write: %+v", a)
}
// ADR 0570 clause 6: the client leaves a provision's `${secret:}` to the node-engine, which fills it at the write
// from the vault's value of a provided provision, and writes the file 0600; a provision not provided refuses it.
keyed := func(state string, ref bool, id, local string) string {
p := provisions.Provision{Provision: "db", Local: local, Consumer: provisions.End{Module: "m", Node: "n1"},
Provider: provisions.End{Module: "postgres", Node: "n2"}, State: state, At: time.Now()}
if state != provisions.Provided {
p.Why = "the provider has not answered"
}
if ref {
p.Reference = &provisions.Reference{ID: id, Generation: 2}
}
raw, err := provisions.Encode(p)
if err != nil {
t.Fatal(err)
}
return string(raw)
}
for _, tc := range []struct {
name, state string
ref bool
content string
want string // the file after; "" keeps what was there
mode os.FileMode
vault string // the vault's credential name, when not the one the provision names
refID string // the reference's id, when not the provision's shared value
local string // the provision's local name (ADR 0094): the placeholder names it
}{
{name: "a provision not provided is refused", state: provisions.Asked, content: "password=${secret:db}\n"},
{name: "provided with no reference is refused", state: provisions.Provided, content: "password=${secret:db}\n"},
{name: "a reference to no credential of the provision is refused", state: provisions.Provided, ref: true,
refID: "provision/n9/db", content: "password=${secret:db}\n"},
{name: "a name no provision goes by is refused", state: provisions.Provided, ref: true, content: "password=${secret:other}\n"},
{name: "the vault's value is another than the provision names", state: provisions.Provided, ref: true,
vault: "pair/n2/db/n1/m", content: "password=${secret:db}\n"},
{name: "provided: filled at the write, 0600", state: provisions.Provided, ref: true,
content: "password=${secret:db}\n", want: "password=s3cret\n", mode: 0o600},
{name: "a local name is the provision's by its key", state: provisions.Provided, ref: true, local: "admin",
content: "admin=${secret:admin}\n", want: "admin=s3cret\n", mode: 0o600},
{name: "no placeholder is written as it came, as the declaration says", content: "plain\n", want: "plain\n", mode: 0o644},
} {
before, _ := os.ReadFile(path)
if tc.state != "" {
id := "provision/n2/db"
if tc.refID != "" {
id = tc.refID
}
clear(provision)
key := "n1.m.db"
if tc.local != "" {
key += "." + tc.local
}
provision[key] = keyed(tc.state, tc.ref, id, tc.local)
}
vaultName = "provision/n2/db"
if tc.vault != "" {
vaultName = tc.vault
}
a := e.Write(context.Background(), uid, filling.Request{Path: path, Content: []byte(tc.content)})
got, _ := os.ReadFile(path)
if tc.want == "" {
if a.Failed == "" && a.Refused == "" || string(got) != string(before) {
t.Errorf("%s: answered %+v, the file holds %q", tc.name, a, got)
}
continue
}
info, _ := os.Stat(path)
if a.Failed != "" || a.Refused != "" || string(got) != tc.want || info.Mode().Perm() != tc.mode {
t.Errorf("%s: answered %+v, the file holds %q, mode %v", tc.name, a, got, info.Mode())
}
}
}
+33 -14
View File
@@ -68,14 +68,23 @@ func foundDirectories(known store.State) []store.Applied {
// known (novox/hq ADR 0272): typed on the control-node, from the operator's own account. The directory's
// path is the resource's target beside the reason, never in it: the controller says a reason on the
// operator's channel, which withholds a path (ADR 0234 §6), and keeps the target as evidence.
//
// From the lease write, heard as the engine hears it (meshTrust), the line has no answerer: the way out is the node
// seat's hand-over with the directory's path, which asks the operator on their channel (novox/hq ADR 0563 §5).
func handOverLine() string {
node, _ := foundNode.Load().(string)
if node == "" {
node = "<node>"
}
if t := foundTrust.Load(); t != nil && t.HandedAway() {
return fmt.Sprintf("`%s/node-engine.hand-over` with the directory's path, then your yes on your channel", node)
}
return fmt.Sprintf("`nox node hand-over %s <directory>` on the control-node", node)
}
// foundTrust is what the engine trusts, whose lease decides the way out a directory used as found names.
var foundTrust atomic.Pointer[meshTrust]
// withFoundDirectories adds to a statement every directory used as found, as its module's resource of kind
// directory, unhealthy: the reason names its owner and mode as they are now, what was declared, and the line
// that hands it over. The path is the target, which the controller keeps as evidence.
@@ -129,16 +138,30 @@ func foundNow() []store.Applied {
// never follows a path it did not state: a hand-over is root giving a directory to an account at the next
// apply, and a path that resolves elsewhere would give that elsewhere.
func handOverFound(statePath, path, by string) (string, error) {
found, _, err := judgeFound(statePath, path)
if err != nil {
return "", err
}
if err := store.AddHandedOver(statePath, store.HandOver{ID: found.ID, Path: path, At: time.Now().UTC(), By: by}); err != nil {
return "", err
}
return fmt.Sprintf("%s (%s) is handed to the mesh by %s: the next apply gives it %s, and it is the mesh's from "+
"then on — removed when empty and no longer declared", path, found.ID, by, orUnsaid(found.Declared)), nil
}
// judgeFound is the judgement of a hand-over's path, recording nothing: the directory this machine uses as found at
// that path, as the engine states it, and what is on disk there.
func judgeFound(statePath, path string) (store.Applied, os.FileInfo, error) {
if !filepath.IsAbs(path) {
return "", fmt.Errorf("hand-over takes the directory's absolute path, and %q is not one; nothing was handed over", path)
return store.Applied{}, nil, fmt.Errorf("hand-over takes the directory's absolute path, and %q is not one; nothing was handed over", path)
}
if filepath.Clean(path) != path {
return "", fmt.Errorf("%q is not the directory's path as the engine states it (no `..`, no doubled or trailing "+
return store.Applied{}, nil, fmt.Errorf("%q is not the directory's path as the engine states it (no `..`, no doubled or trailing "+
"separator); nothing was handed over", path)
}
known, err := store.Load(statePath)
if err != nil {
return "", err
return store.Applied{}, nil, err
}
var found *store.Applied
for _, r := range foundDirectories(known) {
@@ -149,31 +172,27 @@ func handOverFound(statePath, path, by string) (string, error) {
}
}
if found == nil {
return "", errors.New(path + " is not a directory this machine uses as found (the module's condition names " +
return store.Applied{}, nil, errors.New(path + " is not a directory this machine uses as found (the module's condition names " +
"the ones that are); nothing was handed over")
}
info, err := os.Lstat(path)
switch {
case err != nil:
return "", fmt.Errorf("%s cannot be read on this machine: %v; nothing was handed over", path, err)
return store.Applied{}, nil, fmt.Errorf("%s cannot be read on this machine: %v; nothing was handed over", path, err)
case info.Mode()&os.ModeSymlink != 0:
return "", fmt.Errorf("%s is a link, and the mesh hands over no link (novox/hq ADR 0012); nothing was handed over", path)
return store.Applied{}, nil, fmt.Errorf("%s is a link, and the mesh hands over no link (novox/hq ADR 0012); nothing was handed over", path)
case !info.IsDir():
return "", fmt.Errorf("%s is not a directory on this machine; nothing was handed over", path)
return store.Applied{}, nil, fmt.Errorf("%s is not a directory on this machine; nothing was handed over", path)
}
resolved, err := filepath.EvalSymlinks(path)
if err != nil {
return "", fmt.Errorf("%s cannot be resolved on this machine: %v; nothing was handed over", path, err)
return store.Applied{}, nil, fmt.Errorf("%s cannot be resolved on this machine: %v; nothing was handed over", path, err)
}
if resolved != path {
return "", fmt.Errorf("%s is reached through a link (it is %s), and the mesh hands over nothing through one "+
return store.Applied{}, nil, fmt.Errorf("%s is reached through a link (it is %s), and the mesh hands over nothing through one "+
"(novox/hq issue 339); nothing was handed over", path, resolved)
}
if err := store.AddHandedOver(statePath, store.HandOver{ID: found.ID, Path: path, At: time.Now().UTC(), By: by}); err != nil {
return "", err
}
return fmt.Sprintf("%s (%s) is handed to the mesh by %s: the next apply gives it %s, and it is the mesh's from "+
"then on — removed when empty and no longer declared", path, found.ID, by, orUnsaid(found.Declared)), nil
return *found, info, nil
}
// handOverFrom is the engine's hand-over for the controller's ask: the state beside statePath, the path as asked.
+15
View File
@@ -40,6 +40,21 @@ func TestAFoundDirectoryIsStatedUnhealthy(t *testing.T) {
strings.Contains(r.Reason, "mesh-host") || strings.Contains(r.Reason, dir) {
t.Fatalf("the way out is not the operator's line, or carries the path: %q", r.Reason)
}
// ADR 0563 §5: from the lease write the remedy is the node seat's hand-over, answered on the operator's channel, and
// no longer the controller's terminal line; the lease is the one the engine hears.
handed := newMeshTrust(nil)
handed.handedAway = true
foundTrust.Store(handed)
if r := withFoundDirectories(&link.Health{}, foundDirectories(known)).Resources[0]; !strings.Contains(r.Reason,
"`laptop/node-engine.hand-over`") || strings.Contains(r.Reason, "nox node hand-over") || strings.Contains(r.Reason, dir) {
t.Fatalf("after the lease write the way out is %q", r.Reason)
}
foundTrust.Store(newMeshTrust(nil))
if r := withFoundDirectories(&link.Health{}, foundDirectories(known)).Resources[0]; !strings.Contains(r.Reason,
"`nox node hand-over laptop <directory>`") {
t.Fatalf("while the controller holds its lease the way out is %q", r.Reason)
}
foundTrust.Store((*meshTrust)(nil))
foundNode.Store("")
if r := withFoundDirectories(&link.Health{}, foundDirectories(known)).Resources[0]; !strings.Contains(r.Reason,
"`nox node hand-over <node> <directory>`") {
+40
View File
@@ -8,6 +8,8 @@ import (
"testing"
"time"
meshasker "git.novox.be/novox/mesh-sdk/go/asker"
"git.novox.be/novox/mesh-sdk/go/asks"
"git.novox.be/novox/mesh-sdk/go/declaration"
"github.com/novox/mesh-host/internal/apply"
"github.com/novox/mesh-host/internal/link"
@@ -53,4 +55,42 @@ func TestALibrarySaysItsWaitWithoutWordsToParse(t *testing.T) {
if strings.Contains(media.Reason, dir) || len(books.AccessWaits) != 1 || books.AccessWaits[0].Reason != link.WaitDisk {
t.Fatalf("a missing root does not wait for its disk, or a reason names a path: %+v %+v", books, media)
}
// 774 (ADR 0563 §4): the library's home engine asks its take of the operator itself, once: no second ask while one is
// open, nor after it ended; and a library waiting for its disk is not asked.
t.Run("the home engine asks the take once", func(t *testing.T) {
var asked []asks.Ask
state := &link.MemState{}
h := &link.HandActs{Node: "n1", State: state,
JudgeAccess: func(act, access string) (link.Judged, error) {
return link.Judged{Subject: access, Path: root, Sets: "the group mesh-media", Disk: "0:0 755"}, nil
},
AccessAct: func(link.AccessAct) (string, error) { return "taken", nil }}
h.Asker = &meshasker.Client{Module: asks.EngineAsker("n1"), State: state,
Publish: func(subject string, body any, id string) error {
if a, ok := body.(asks.Ask); ok {
asked = append(asked, a)
}
return nil
}}
waiting := keepWaiting(vs)
askTakes(h, waiting, nil)
askTakes(h, waiting, nil)
if len(waiting) != 1 || waiting[0] != "library:media" || len(asked) != 1 {
t.Fatalf("waiting %v asked %d times", waiting, len(asked))
}
h.Heard(asks.Warrant{Ask: asked[0].ID, Asker: asks.EngineAsker("n1"), Outcome: asks.OutcomeExpired, At: time.Now()})
askTakes(h, waiting, nil)
if len(asked) != 1 {
t.Fatalf("the take was asked again after its ask ended: %d", len(asked))
}
// Kept in a key of its own (rev-l3 round 1): the acts gone from the record never make it ask again.
keys, _ := state.Keys("act.")
for _, k := range keys {
_ = state.Delete(k)
}
askTakes(h, waiting, nil)
if len(asked) != 1 {
t.Fatalf("the take was asked again with the acts gone: %d", len(asked))
}
})
}
+72 -8
View File
@@ -516,7 +516,7 @@ func runApply(ctx context.Context, opts options, d *declaration.Declaration, raw
// A file given by digest is had before the plan is drawn from it (novox/hq ADR 0486 §1, issue 1695). This command has
// no link to the bus, so a file already here that verifies is used, and any other is refused: never written empty.
if err := apply.ResolveFileReferences(ctx, d, link.FilesFrom(ctx)); err != nil {
if err := apply.ResolveFileReferencesKept(ctx, d, link.FilesFrom(ctx), objectKeep(opts)); err != nil {
return fmt.Errorf("a file of this declaration could not be had by its digest: %w", err)
}
@@ -1105,10 +1105,17 @@ func runLink(ctx context.Context, opts options) error {
// the bus, answer what the controller asks of this machine, and say its facts: when a link opens, and when the
// controller asks.
proof := &proving{statePath: opts.state, version: runningVersion(), say: say}
// What this engine trusts a mesh signature from: the controller's key until its lease is handed away, and the vault's
// once read from its signers (novox/hq ADR 0550 §2).
trust := newMeshTrust(mine.Membership.Signer)
// And the way out a directory used as found names follows the lease it hears (novox/hq ADR 0563 §5).
foundTrust.Store(trust)
queue := &link.Session{
Membership: membership,
// And the public halves and certificate its record says (novox/hq ADR 0555 §3), read at each facts report.
Keys: recordKeys(opts.state),
Keys: recordKeys(opts.state, trust.Trusts),
// And the controller's key, read per signed ask: none once its lease is handed away (novox/hq ADR 0563 §7).
ControllerKey: trust.Controller,
// And the first have this build writes is what proves it to the launcher (ADR 0539 §9, ADR 0227 §8).
HaveWritten: proof.haveWritten,
// And the operator's hand-over of a directory used as found, carried by the controller from the
@@ -1129,6 +1136,10 @@ func runLink(ctx context.Context, opts options) error {
Say: say,
Timeout: opts.timeout,
}
// And the node seat node-engine this engine holds (novox/hq ADR 0563): its verbs on every link, its own asks to the
// operator under node-engine-<node>, and its acts kept beside its state.
queue.Acts = engineActs(opts, mine.Node, queue, setuidSearch)
go reconcileActs(ctx, queue.Acts)
current := &agentNow{}
setRunsHere(runsFromAgent(current))
factsOfNow := func(ctx context.Context, ag *agent.Agent) {
@@ -1173,7 +1184,10 @@ func runLink(ctx context.Context, opts options) error {
}
// The node follows its wants: one module at a time from the composition its want names (novox/hq ADR 0539), and the
// busy containers are asked for again soon (ADR 0360 §11).
go nodeAgent(aside, opts, mine.Membership.Signer, queue, sched, say, afterApply, current, factsOfNow)
// And what it trusts, followed on the bus, said in its record whenever it changes (ADR 0550 §2).
trust.onChange(func() { factsOfNow(aside, current.get()) })
go followTrust(aside, queue, trust, say)
go nodeAgent(aside, opts, mine.Membership.Signer, trust, queue, sched, say, afterApply, current, factsOfNow)
go busyPasses(aside, func() {
if ag := current.get(); ag != nil {
ag.Retry(busyModules()...)
@@ -1186,7 +1200,7 @@ func runLink(ctx context.Context, opts options) error {
go serveMeshCLI(aside, queue, say)
// And a filled file is written on the provisioner client's request, one at a time with the apply (novox/hq ADR
// 0349 §4).
go serveFilling(aside, opts, mine.Membership.Signer, say)
go serveFilling(aside, opts, mine.Membership.Signer, queue, mine.Node, say)
held := link.HoldRoused(aside, membership, queue, say, opts.timeout, rousedBySignal(ctx))
// **Cleanly**, or the launcher counts standing aside as a crash and rolls the new host back
@@ -1339,6 +1353,8 @@ func judgeWhatRuns(ctx context.Context, j *liveness.Judge, queue *link.Session,
var lastSaid time.Time
owed := true
spaced := 1.0
// healthUnwritten is why the health report was last not written, said once each time it changes.
var healthUnwritten string
// A module whose change ended unjudged is looked at and said at once (review of fix 15): a module with nothing
// long-running is then ok within a look of its apply, not at the five-minute statement.
var nudged <-chan struct{}
@@ -1370,6 +1386,7 @@ func judgeWhatRuns(ctx context.Context, j *liveness.Judge, queue *link.Session,
if a := accountJudge; a != nil {
as, accountsChanged := a.Look(ctx)
accountSt = &as
lastAccounts.Store(&as)
owed = owed || accountsChanged
if accountsChanged {
for _, v := range as.Accounts {
@@ -1384,6 +1401,9 @@ func judgeWhatRuns(ctx context.Context, j *liveness.Judge, queue *link.Session,
accessSt, accessesChanged := accesses.look(time.Now())
keeping.poke(ctx)
owed = owed || accessesChanged
// And a library whose root holds data outside its group is asked of the operator by this engine, once (novox/hq
// ADR 0563 §4).
askTakes(queue.Acts, keepWaiting(accessSt), say)
if accessesChanged {
for _, v := range accessSt {
if v.State != link.StateHealthy {
@@ -1418,6 +1438,11 @@ func judgeWhatRuns(ctx context.Context, j *liveness.Judge, queue *link.Session,
healthy := st.Healthy() && (netSt == nil || netSt.State != network.Unhealthy) &&
(unitSt == nil || len(unitSt.Failed) == 0) && (accountSt == nil || accountSt.Healthy()) && len(found) == 0 &&
accessesHealthy(accessSt)
h := *withAccesses(withFoundDirectories(withAccounts(withUnits(healthAsReported(st, netSt),
unitSt), accountSt), found), accessSt)
// The node's own health report in its key on the bus, compared on every look and written only when what it says
// changed (issue 774): what the watchman raises the machine's alerts from, without the controller.
writeHealthReport(ctx, queue, h, netSt, &healthUnwritten, say)
if !owed && !(!healthy && since >= sayUnhealthyAgain) && since < sayAnyway {
continue
}
@@ -1444,13 +1469,40 @@ func judgeWhatRuns(ctx context.Context, j *liveness.Judge, queue *link.Session,
reconcile(inactive...)
}
}
if queue.SayHealth(ctx, *withAccesses(withFoundDirectories(withAccounts(withUnits(healthAsReported(st, netSt),
unitSt), accountSt), found), accessSt)) {
if queue.SayHealth(ctx, h) {
lastSaid, owed = time.Now(), false
}
}
}
// writeHealthReport writes this node's health report to its key (link.WriteNodeHealth) when it changed. Why it could
// not is said once per reason, and said again when it is written after all.
func writeHealthReport(ctx context.Context, queue *link.Session, h link.Health, ns *network.Statement, unwritten *string,
say link.Announce) {
if queue == nil {
return
}
var peers []link.PeerSeen
if ns != nil {
for _, p := range ns.Peers {
peers = append(peers, link.PeerSeen{Key: p.Key, Addresses: p.Addresses, Seen: p.Seen})
}
}
var machines link.Machines
if c := machinesNow.Load(); c != nil {
machines.ByAddress, machines.Hub = c.Machines()
}
_, err := queue.WriteNodeHealth(ctx, link.NodeHealthOf(queue.Membership.Node, h, peers, machines))
switch {
case err != nil && err.Error() != *unwritten:
*unwritten = err.Error()
say("this machine's health report is not written to the bus: " + err.Error())
case err == nil && *unwritten != "":
*unwritten = ""
say("this machine's health report is written to the bus again")
}
}
// healthAsReported is a statement as the report and the event carry it.
func healthAsReported(st liveness.Statement, ns *network.Statement) *link.Health {
// LibraryContract: this engine reads a resource's declared `health` and judges it (ADR 0240 Phase B), judges a
@@ -1646,6 +1698,15 @@ func announceOr(say link.Announce) func(string) {
// run by hand beside this service — the lock beside the state does the same (store.Lock).
var applying sync.Mutex
// objectKeep is the node's own keep of the objects its compositions name by digest, beside its state (apply.ObjectKeep,
// issue 924): a composition the node once applied never again needs the controller's bucket to hold its files.
func objectKeep(opts options) *apply.ObjectKeep {
if opts.state == "" {
return nil
}
return apply.KeepObjectsIn(filepath.Join(filepath.Dir(opts.state), "objects"))
}
// appliedModule is how the apply of one module's declaration came out.
type appliedModule struct {
// Refused is why nothing was applied, when nothing was.
@@ -1685,7 +1746,7 @@ func applyModuleHeld(ctx context.Context, opts options, declared *declaration.De
// **A file given by digest is fetched and verified before anything of the composition is applied** (novox/hq ADR
// 0486 §1, issue 1695): one object missing or differing refuses the module whole.
if err := apply.ResolveFileReferences(ctx, declared, link.FilesFrom(ctx)); err != nil {
if err := apply.ResolveFileReferencesKept(ctx, declared, link.FilesFrom(ctx), objectKeep(opts)); err != nil {
return appliedModule{Refused: "a file of this composition could not be had by its digest: " + err.Error()}
}
@@ -2167,9 +2228,12 @@ func withMissing(st liveness.Statement) liveness.Statement {
// recordKeys reads the public halves of the node's keys and the certificate the mesh issued for its internal name from
// its state directory (novox/hq ADR 0555 §3): what the node's record says of them. A key that cannot be read is left
// empty, never invented; the certificate is there only after the first declaration that carried it.
func recordKeys(state string) func() link.RecordKeys {
func recordKeys(state string, trusts func() []string) func() link.RecordKeys {
return func() link.RecordKeys {
var k link.RecordKeys
if trusts != nil {
k.Trusts = trusts()
}
if id, err := identity.Load(identity.Path(state)); err == nil {
k.TunnelKey = id.Overlay.Public
}
+85 -22
View File
@@ -5,12 +5,14 @@ import (
"crypto/ed25519"
"encoding/base64"
"encoding/json"
"errors"
"fmt"
"os"
"path/filepath"
"sort"
"strings"
"sync"
"sync/atomic"
"time"
"git.novox.be/novox/mesh-sdk/go/compose/frombus"
@@ -343,13 +345,14 @@ func resourcesOf(cs []delivery.Composed) []declaration.Resource {
// controller's key the node was enrolled with (novox/hq ADR 0539 §3, §11); a node adopted later has it at its adoption.
const operatorKeysFile = "operator-keys"
// trustedKeys are the keys a composition may be signed by: the controller's, which the node was enrolled with, and the
// operator's, read from operatorKeysFile each time it is asked, so a key that arrives while the agent runs is trusted at
// the next composition. A file that cannot be read trusts the controller's alone, and says so.
func trustedKeys(opts options, controller ed25519.PublicKey) func() []ed25519.PublicKey {
// trustedKeys are the keys a composition may be signed by: the controller's, which the node was enrolled with, while the
// controller holds its lease (ADR 0550 §2: trusted no more once it is handed away), and the operator's, read from
// operatorKeysFile each time it is asked, so a key that arrives while the agent runs is trusted at the next composition.
// A file that cannot be read trusts the controller's alone, and says so.
func trustedKeys(opts options, controllerNow func() ed25519.PublicKey) func() []ed25519.PublicKey {
return func() []ed25519.PublicKey {
keys := []ed25519.PublicKey{}
if len(controller) == ed25519.PublicKeySize {
if controller := controllerNow(); len(controller) == ed25519.PublicKeySize {
keys = append(keys, controller)
}
raw, err := os.ReadFile(filepath.Join(filepath.Dir(opts.state), operatorKeysFile))
@@ -373,13 +376,16 @@ func trustedKeys(opts options, controller ed25519.PublicKey) func() []ed25519.Pu
}
// keptOnce reads the kept declaration once and splits it by module: its resources (the node's own are those that say
// none, ADR 0539 §9) and its filled paths.
// none, ADR 0539 §9) and its filled paths. Once the read is retired (keptRetired) nothing is read.
func keptOnce(opts options, signer ed25519.PublicKey) func() (map[string][]json.RawMessage, map[string][]string) {
var once sync.Once
var by map[string][]json.RawMessage
var filled map[string][]string
return func() (map[string][]json.RawMessage, map[string][]string) {
once.Do(func() {
if keptRetired.Load() {
return
}
raw, err := store.LoadDeclared(store.DeclaredPath(opts.state), signer)
if err != nil {
return
@@ -429,7 +435,7 @@ func splitKept(raw []byte) (map[string][]json.RawMessage, map[string][]string) {
// nodeAgent runs the node agent under the session's links for as long as ctx lasts: it waits for a link, follows the
// node's wants and compositions on it, and begins again with the next link. A watch it cannot make (a bucket this node's
// account may not read) is said, and tried again with a growing wait, as the link is.
func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, session *link.Session, sched *apply.Scheduler,
func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, trust *meshTrust, session *link.Session, sched *apply.Scheduler,
say link.Announce, afterApply func(), current *agentNow, sayFacts func(ctx context.Context, ag *agent.Agent)) {
wait := 2 * time.Second
for ctx.Err() == nil {
@@ -446,6 +452,18 @@ func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, sess
return
}
var ag *agent.Agent
// The kept release-5 declaration is no longer read once every module the seat's placement puts here has a have of its
// own (ADR 0550 §2: its signature is the controller's, trusted no more after the lease write).
// Nor before this engine keeps the node's own as it runs (applied/node.json): until then the kept declaration is the one
// record of what the node's own composition carries, its folded modules among them (issue 774, review of the
// re-land, round 1: novox's first start would compose its own with nats running nothing).
if retired, why := placedAllHave(linked, session, node); retired && !keptRetired.Load() {
if !appliedStoreFor(opts.state).holds(agent.NodeModule) {
say("the kept declaration is still read: " + why + ", but this engine keeps no node's own composition yet")
} else if !keptRetired.Swap(true) {
say("the kept declaration is no longer read: " + why)
}
}
kept := keptOnce(opts, signer)
// The mesh's own certificate the declaration kept from before carries, kept for the record before this machine
// composes its own (ADR 0555 §3): without it a module that asks for a certificate would be stood in for.
@@ -458,7 +476,7 @@ func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, sess
}
// The bus's user list, written alone from the newest composition of the bus module (ADR 0547 §8): on every
// node, and a node whose compositions carry no list writes nothing.
users := userListFollower(opts, session, composedBucket, trustedKeys(opts, signer), say, func() {
users := userListFollower(opts, session, composedBucket, trustedKeys(opts, trust.Controller), say, func() {
// This engine wrote the list, so the bus reloads its accounts and answers no JetStream request for a few
// seconds: what failed in them is tried again once that is over, not at its next wait.
time.AfterFunc(reloadSettles, func() {
@@ -483,7 +501,10 @@ func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, sess
// The same kept declaration's filled paths, so a module still running from it keeps its files filled.
KeptFilled: func() map[string][]string { _, filled := kept(); return filled },
Recorded: recordedIn(opts.state, say),
Inactive: inactiveIn(&liveness.Exec{Run: apply.ExecRunner, System: builtFor}, heldBy(apply.WindowsIn(filepath.Dir(opts.state)))),
// What this engine applied, kept on the machine and read at the start, so what runs is known before the node
// composes its own (issue 774).
Applied: appliedStoreFor(opts.state),
Inactive: inactiveIn(&liveness.Exec{Run: apply.ExecRunner, System: builtFor}, heldBy(apply.WindowsIn(filepath.Dir(opts.state)))),
Revisions: func(ctx context.Context, key string) ([]agent.Update, error) {
got, err := session.KVHistory(ctx, composedBucket, key)
if err != nil {
@@ -495,12 +516,15 @@ func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, sess
}
return out, nil
},
Trusted: trustedKeys(opts, signer),
Trusted: trustedKeys(opts, trust.Controller),
Backup: backupNode(session),
// What the filling refused before this apply may be named now: the provisioner's client is told to fill
// again (ace, 2026-10-25 21:17).
Settled: func(module string) { knockFilling(module, say) },
Say: say,
Settled: func(module string) { knockFilling(module+" settled", say) },
// And once the haves are read at the start, and whenever a module is taken as running without an apply, so
// a refusal from the engine's start does not wait for some module to settle (issue 924).
RunsChanged: func(why string) { knockFilling(why, say) },
Say: say,
// A failure of the bus not answering for a moment is tried again: 2 s, 5 s, 15 s, 60 s, then every 5 min.
Backoff: link.Backoff,
Apply: func(ctx context.Context, c delivery.Composed, others []delivery.Composed) agent.Result {
@@ -512,7 +536,7 @@ func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, sess
if herr != nil {
res = failedWith(herr.Error())
} else {
res = applyComposition(actx, opts, held, others, sched, say)
res = applyComposition(actx, opts, held, others, ag.Placed(scopeName(c)), ag.RemovedByWant(), sched, say)
}
done()
// The mesh's own certificate it carries is kept for this machine's record (ADR 0555 §3).
@@ -531,7 +555,7 @@ func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, sess
},
Remove: func(ctx context.Context, module string, others []delivery.Composed) agent.Result {
actx, done := applyContext(session.WithLink(ctx))
res := removeComposition(actx, opts, module, others, sched, say)
res := removeComposition(actx, opts, module, others, ag.Placed(module), sched, say)
done()
afterApply()
sayFacts(ctx, ag)
@@ -549,7 +573,7 @@ func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, sess
hooks := composerHooks{
running: func() []delivery.Composed {
if now := own.get(); now != nil {
return now.Everything()
return runsAsWritten(now.Everything(), opts.state)
}
return nil
},
@@ -572,11 +596,17 @@ func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, sess
now.NotComposedEach(run, why, held) // a module held alone marks that module, never the others (issue 1126)
}
},
// And the vault's authority beside the controller's, and this node's certificate asked of the vault when a
// module here asks for one (ADR 0550 §2).
vaultAuthority: trust.VaultAuthority,
certify: func(ctx context.Context) error {
return certifyThisNode(ctx, opts.state, node, trust, session.AskVault, func() { sayFacts(ctx, ag) })
},
}
// And its answer to every placement candidate, placed or not, applying nothing (novox/hq ADR 0559 §4): written into
// its own key, `<node>.node`, under candidates.
shadowHooks := hooks
shadowHooks.refused = nil
shadowHooks.refused, shadowHooks.certify = nil, nil
shadowEnv := composerEnv(session, node, say, shadowHooks)
// And, after each answer, what runs of an own secret the answer says the vault holds another value of, handed to
// the vault while the node applies the controller's compositions (issue 1126, cause C4; ownresync).
@@ -596,8 +626,10 @@ func nodeAgent(ctx context.Context, opts options, signer ed25519.PublicKey, sess
continue
}
if composer != nil {
machinesNow.Store(composer) // the machines a failing network part points at, by the placements (issue 774)
composes.set(delivery.ComposesOwn)
env.Composed, env.SelfComposed, env.Trusted, env.Policy = composer, true, nil, composer.Policy
env.HoldsData = composer.HoldsData // what runs from the kept declaration says Data as its manifest does
session.NoteComposes(linked, delivery.ComposesOwn, why)
// And once the catalogue no longer places it, the agent begins again on the controller's compositions: the
// placement key deleted ends this run, the way back (issue 1126, the seed's revert).
@@ -663,7 +695,13 @@ func (c *agentNow) get() *agent.Agent {
}
// applyComposition applies one composition: a module's, or the node's own.
func applyComposition(ctx context.Context, opts options, c delivery.Composed, others []delivery.Composed,
//
// placed is the modules a want still places here besides this one: none of their resources is removed by this apply,
// which is refused instead (issue 774).
//
// removed is the modules whose want says removed: in the node's own apply, a resource of any other module goes only when
// its module is among them (issue 774).
func applyComposition(ctx context.Context, opts options, c delivery.Composed, others []delivery.Composed, placed, removed map[string]bool,
sched *apply.Scheduler, say link.Announce) agent.Result {
name := scopeName(c)
if c.Module == "" {
@@ -679,7 +717,13 @@ func applyComposition(ctx context.Context, opts options, c delivery.Composed, ot
if err := refuseEmptyWrites(d); err != nil {
return failedWith(err.Error())
}
scope := apply.Scope{Module: name, Others: resourcesOf(others)}
scope := apply.Scope{Module: name, Others: resourcesOf(others), Placed: placed}
if name == agent.NodeModule {
scope.RemovedByWant = removed
if scope.RemovedByWant == nil {
scope.RemovedByWant = map[string]bool{}
}
}
got := applyModuleHeld(ctx, opts, d, &scope, sched, say)
if got.Refused != "" {
return failedWith(got.Refused)
@@ -698,9 +742,9 @@ func applyComposition(ctx context.Context, opts options, c delivery.Composed, ot
}
// removeComposition removes one module's own resources and nothing of another's.
func removeComposition(ctx context.Context, opts options, module string, others []delivery.Composed,
func removeComposition(ctx context.Context, opts options, module string, others []delivery.Composed, placed map[string]bool,
sched *apply.Scheduler, say link.Announce) agent.Result {
scope := apply.Scope{Module: module, Others: resourcesOf(others)}
scope := apply.Scope{Module: module, Others: resourcesOf(others), Placed: placed}
got := applyModuleHeld(ctx, opts, &declaration.Declaration{Version: declaration.Version}, &scope, sched, say)
switch {
case got.Refused != "":
@@ -725,8 +769,23 @@ func busyModules() []string {
return out
}
// keptDeclaration is the declaration the node kept before the switch, verified with the mesh's key when it is given.
// machinesNow is the composer whose placements name the mesh's machines by address, for the health report's targets
// (writeHealthReport); nil on a node that does not compose its own, whose parts then name no target.
var machinesNow atomic.Pointer[selfcompose.Composer]
// keptRetired says the kept declaration is read no more: every module the seat's placement puts on this node has a have of
// its own (placedAllHave), so what runs is the agent's alone (ADR 0550 §2, design 47's L3 s9).
var keptRetired atomic.Bool
// errKeptRetired is keptDeclaration's answer once the read is retired.
var errKeptRetired = errors.New("the kept declaration is read no more: every module placed on this node has a have of its own")
// keptDeclaration is the declaration the node kept before the switch, verified with the mesh's key when it is given;
// once the read is retired, nothing is read.
func keptDeclaration(statePath string, signer ed25519.PublicKey) (*declaration.Declaration, error) {
if keptRetired.Load() {
return nil, errKeptRetired
}
var raw []byte
if len(signer) > 0 {
var err error
@@ -903,6 +962,9 @@ type composerHooks struct {
unseal func(string) ([]byte, error)
refused func(why string, held map[string]string)
held func(module string) (delivery.Seals, []string, error)
// vaultAuthority and certify are selfcompose.Env's VaultAuthority and Certify.
vaultAuthority func() string
certify func(ctx context.Context) error
}
func selfComposer(ctx context.Context, session *link.Session, node string, say func(string), hooks composerHooks) (*selfcompose.Composer, string) {
@@ -966,7 +1028,8 @@ func composerEnv(session *link.Session, node string, say func(string), hooks com
Hear: session.Hear,
// And a moved module's templates archive, in the artifact store, as every archive is fetched (issue 1126).
Fetch: apply.Fetch,
Running: hooks.running, Unseal: hooks.unseal, Refused: hooks.refused, Held: hooks.held}
Running: hooks.running, Unseal: hooks.unseal, Refused: hooks.refused, Held: hooks.held,
VaultAuthority: hooks.vaultAuthority, Certify: hooks.certify}
}
// endWhenUnplaced calls stop once this node's placement key is deleted from the seat's placement (after the watch
+2 -1
View File
@@ -159,7 +159,8 @@ func (b *itBed) start(trustOperator bool) (stop func()) {
done := make(chan struct{})
go func() {
defer close(done)
nodeAgent(ctx, b.opts, b.controll.Public().(ed25519.PublicKey), b.session, nil, b.say, func() {},
controller := b.controll.Public().(ed25519.PublicKey)
nodeAgent(ctx, b.opts, controller, newMeshTrust(controller), b.session, nil, b.say, func() {},
&agentNow{}, func(context.Context, *agent.Agent) {})
}()
return func() { cancel(); <-done }
+168
View File
@@ -0,0 +1,168 @@
package main
import (
"context"
"encoding/json"
"fmt"
"regexp"
"sort"
"strings"
"git.novox.be/novox/mesh-sdk/go/catalogue"
"git.novox.be/novox/mesh-sdk/go/provisions"
)
// provisionSecrets fills a provision's value into a file the provisioner's client hands the node-engine to write
// (novox/hq ADR 0570 clause 6): the client fills the file but for each `${secret:<provision>}`, and the node-engine
// fills those at the write, so the client never holds the value. A whole `secrets` file of a handed consumer is the
// placeholder alone.
//
// A placeholder names the credential as the module calls it, `${secret:<name>}` (catalogue.SecretLocal, ADR 0094): the
// local name where the module keeps several values of one provision, else the provision. For each, it finds the one
// provision key `<node>.<consumer>.<provision>[.<local>]` of the seat mesh-provisioner's state `provisions` (clause 1)
// going by that name, which must be `provided` and carry the vault's reference; asks the vault for that pair
// credential sealed to this machine's key (the verb `own`, ADR 0349 §11 with the node-engine standing for the client,
// ADR 0550 §5); and opens it with this machine's sealing key. Anything short of that refuses the write, naming the
// provision and why: a file written with a placeholder left in it, or with no value, is a module that starts and cannot
// connect.
type provisionSecrets struct {
node string
// list is every key of mesh-provisioner's `provisions` under prefix (`<node>.<consumer>.`), by key: what this
// engine may read, its own node's (`<node>.>`).
list func(ctx context.Context, prefix string) (map[string][]byte, error)
// own asks the vault's verb `own` in this engine's name.
own func(ctx context.Context, args map[string]any) (json.RawMessage, error)
// open opens a value sealed to this machine.
open func(sealed string) ([]byte, error)
}
// secretPlaceholder is `${secret:<name>}`.
var secretPlaceholder = regexp.MustCompile(`\$\{secret:([A-Za-z0-9_-]+)\}`)
// fill is content with each provision placeholder of module's filled, and whether any was.
func (p provisionSecrets) fill(ctx context.Context, module string, content []byte) ([]byte, bool, error) {
found := secretPlaceholder.FindAllSubmatch(content, -1)
if len(found) == 0 {
return content, false, nil
}
if module == "" {
return nil, false, fmt.Errorf("the file names %d provision secret(s) and belongs to no module, so whose they are is not known", len(found))
}
if p.list == nil || p.own == nil || p.open == nil {
return nil, false, fmt.Errorf("%s's provisions cannot be read on this machine: the node-engine is not linked to the mesh", module)
}
prefix := p.node + "." + module + "."
keys, err := p.list(ctx, prefix)
if err != nil {
return nil, false, fmt.Errorf("%s's provisions cannot be read: %w", module, err)
}
out := string(content)
done := map[string]bool{}
for _, f := range found {
whole, name := string(f[0]), string(f[1])
if done[whole] {
continue
}
done[whole] = true
pr, key, err := byName(keys, name)
if err != nil {
return nil, false, fmt.Errorf("%s's secret %q: %w", module, name, err)
}
value, err := p.value(ctx, key, pr)
if err != nil {
return nil, false, err
}
out = strings.ReplaceAll(out, whole, string(value))
}
return []byte(out), true, nil
}
// byName is the one provision of keys the module calls name (catalogue.SecretLocal), and its key.
func byName(keys map[string][]byte, name string) (provisions.Provision, string, error) {
var hit []string
var got provisions.Provision
for key, raw := range keys {
pr, err := provisions.Decode(raw)
if err != nil {
return provisions.Provision{}, "", fmt.Errorf("the provision %s: %w", key, err)
}
if catalogue.SecretLocal(pr.Provision, pr.Local) == name {
hit, got = append(hit, key), pr
}
}
switch len(hit) {
case 0:
return provisions.Provision{}, "", fmt.Errorf("no provision of %s goes by that name: nothing asks for it", provisions.ProvisionsRead)
case 1:
return got, hit[0], nil
}
sort.Strings(hit)
return provisions.Provision{}, "", fmt.Errorf("the provisions %s all go by that name", strings.Join(hit, ", "))
}
// value is one provision's value on this node, opened.
func (p provisionSecrets) value(ctx context.Context, key string, pr provisions.Provision) ([]byte, error) {
if pr.State != provisions.Provided || pr.Reference == nil {
why := pr.State
if pr.Why != "" {
why += ": " + pr.Why
}
if pr.State == provisions.Provided {
why = "provided with no reference to the vault's value"
}
return nil, fmt.Errorf("the provision %s is %s, so its value is not written yet", key, why)
}
// The reference names a credential of this provision, as the vault names it (provisions.VaultName).
ours := false
for _, form := range []string{provisions.FormPair, provisions.FormShared, provisions.FormIssued} {
ours = ours || provisions.VaultName(form, pr) == pr.Reference.ID
}
if !ours {
return nil, fmt.Errorf("the provision %s names the vault's %s, which is no credential of it", key, pr.Reference.ID)
}
module, provision, local := pr.Consumer.Module, pr.Provision, pr.Local
type asked struct {
Role string `json:"role"`
Provision string `json:"provision"`
Consumer string `json:"consumer"`
ConsumerModule string `json:"consumer_module"`
Provider string `json:"provider"`
Local string `json:"local,omitempty"`
}
ask := asked{Role: "consumer", Provision: provision, Consumer: p.node, ConsumerModule: module, Provider: pr.Provider.Node, Local: local}
answer, err := p.own(ctx, map[string]any{"credentials": []asked{ask}})
if err != nil {
return nil, fmt.Errorf("the vault's value of %s: %w", key, err)
}
var a struct {
Node string `json:"node"`
Credentials []struct {
asked
Name string `json:"name"`
Generation int `json:"generation"`
Sealed json.RawMessage `json:"sealed"`
} `json:"credentials"`
}
if err := json.Unmarshal(answer, &a); err != nil {
return nil, fmt.Errorf("the vault's answer for %s is not one: %w", key, err)
}
if a.Node != p.node {
return nil, fmt.Errorf("the vault answered %s's credentials for %s on %s", a.Node, key, p.node)
}
for _, c := range a.Credentials {
if c.asked != ask {
continue
}
// The value the provision names, by the vault's credential name and its generation (clause 6): another is refused.
if c.Name != pr.Reference.ID || c.Generation != pr.Reference.Generation {
return nil, fmt.Errorf("the vault answered %s generation %d for %s, and the provision names %s generation %d: "+
"nothing is written until they agree", c.Name, c.Generation, key, pr.Reference.ID, pr.Reference.Generation)
}
value, err := p.open(string(c.Sealed))
if err != nil {
return nil, fmt.Errorf("the vault's value of %s does not open on this machine: %w", key, err)
}
return value, nil
}
return nil, fmt.Errorf("the vault answered no value of %s for this machine", key)
}
+88
View File
@@ -0,0 +1,88 @@
package main
import (
"crypto/sha256"
"encoding/hex"
"encoding/json"
"os"
"git.novox.be/novox/mesh-sdk/go/declaration"
"git.novox.be/novox/mesh-sdk/go/delivery"
"github.com/novox/mesh-host/internal/store"
)
// runsAsWritten is what the node runs as the composer reads it (selfcompose.Env.Running), with every file given by digest
// named as this node last wrote it, where the two differ (issue 924, novox 2026-10-29 01:16). What runs of a module the
// agent has not applied since its start can come from the kept declaration, older than the compositions applied since:
// its export (mesh-vault.kept) then named a digest the controller's sweep had removed, while the file on disk was the
// newer export the node wrote. A node composing its own carries such a file as it runs (mesh-sdk frombus keptOf), so it
// names what is on disk: **only when the node's record says it wrote exactly that content there** (store Applied.Wrote,
// Target), never a file someone else changed. A file that cannot be read, or that the record does not vouch for, is
// left as the composition names it.
func runsAsWritten(running []delivery.Composed, statePath string) []delivery.Composed {
if len(running) == 0 || statePath == "" {
return running
}
known, err := store.Load(statePath)
if err != nil {
return running
}
out := make([]delivery.Composed, len(running))
for i, c := range running {
out[i] = c
var changed []json.RawMessage
for j, raw := range c.Resources {
if fixed, ok := asWritten(raw, known); ok {
if changed == nil {
changed = append([]json.RawMessage(nil), c.Resources...)
}
changed[j] = fixed
}
}
if changed != nil {
out[i].Resources = changed
}
}
return out
}
// asWritten is the file resource raw named by the digest and size of what this node last wrote at its path, when raw is a
// file given by another digest and the file there is what the record says the node wrote.
func asWritten(raw json.RawMessage, known store.State) (json.RawMessage, bool) {
var fields map[string]json.RawMessage
if json.Unmarshal(raw, &fields) != nil {
return nil, false
}
var kind, id, path, digest string
_ = json.Unmarshal(fields["type"], &kind)
_ = json.Unmarshal(fields["id"], &id)
_ = json.Unmarshal(fields["path"], &path)
_ = json.Unmarshal(fields["digest"], &digest)
if kind != string(declaration.TypeFile) || digest == "" || path == "" {
return nil, false
}
rec, ok := known.Find(id)
if !ok || rec.Wrote == "" || rec.Target != path {
return nil, false
}
body, err := os.ReadFile(path)
if err != nil {
return nil, false
}
sum := sha256.Sum256(body)
if hex.EncodeToString(sum[:]) != rec.Wrote {
return nil, false // changed since the node wrote it: not the node's to vouch for
}
if now := declaration.DigestOf(body); now == digest {
return nil, false
} else {
fields["digest"], _ = json.Marshal(now)
}
fields["size"], _ = json.Marshal(len(body))
fixed, err := json.Marshal(fields)
if err != nil {
return nil, false
}
return fixed, true
}
+317
View File
@@ -0,0 +1,317 @@
package main
import (
"context"
"crypto/ed25519"
"encoding/base64"
"encoding/json"
"fmt"
"os"
"path/filepath"
"strings"
"sync"
"time"
"git.novox.be/novox/mesh-sdk/go/buscompose"
"git.novox.be/novox/mesh-sdk/go/delivery"
"git.novox.be/novox/mesh-sdk/go/overlay"
"github.com/novox/mesh-host/internal/identity"
"github.com/novox/mesh-host/internal/link"
)
// **What this engine trusts a mesh signature from** (novox/hq ADR 0550 §2; design 47's L3 item d). Two signers: the
// controller's key, which the node was enrolled with (its membership's signer), and the vault's key, read from the vault's
// state `signers`, which only the vault may write: the bus grants are its anchor, as the wants' and the placement's (the
// lead's design limit, 2026-10-28). The controller's key is trusted until its lease is handed away, heard on the lease key
// (buscompose.LeaseHandedAway) and never on a clock; from then the vault's alone. What it trusts now the node states in its
// record (NodeRecord.Trusts), so the gate before the lease write reads every node's from the bus.
type meshTrust struct {
mu sync.Mutex
controller ed25519.PublicKey
handedAway bool
vault ed25519.PublicKey
authority string
// changed is told when what it trusts changed, so the record is written again.
changed func()
}
// newMeshTrust trusts the controller's key alone until the vault's is read.
func newMeshTrust(controller ed25519.PublicKey) *meshTrust {
t := &meshTrust{}
if len(controller) == ed25519.PublicKeySize {
t.controller = controller
}
return t
}
// onChange names who is told when what it trusts changed.
func (t *meshTrust) onChange(f func()) {
t.mu.Lock()
t.changed = f
t.mu.Unlock()
}
// Controller is the controller's key while the controller holds its lease, and nil once the lease is handed away.
func (t *meshTrust) Controller() ed25519.PublicKey {
if t == nil {
return nil
}
t.mu.Lock()
defer t.mu.Unlock()
if t.handedAway {
return nil
}
return t.controller
}
// HandedAway says the controller's lease was handed away, as heard on the lease key.
func (t *meshTrust) HandedAway() bool {
if t == nil {
return false
}
t.mu.Lock()
defer t.mu.Unlock()
return t.handedAway
}
// Signers are the keys a mesh signature is verified against now: the controller's while it holds the lease, and the
// vault's once read.
func (t *meshTrust) Signers() []delivery.Signer {
t.mu.Lock()
defer t.mu.Unlock()
var out []delivery.Signer
if t.controller != nil && !t.handedAway {
out = append(out, delivery.Signer{Name: delivery.SignerController, Key: t.controller})
}
if t.vault != nil {
out = append(out, delivery.Signer{Name: delivery.SignerVault, Key: t.vault})
}
return out
}
// Verify is the signer that signed msg as kind, or why none it trusts did.
func (t *meshTrust) Verify(kind string, msg, sig []byte) (delivery.Signer, error) {
context, err := delivery.SignContext(kind)
if err != nil {
return delivery.Signer{}, err
}
return delivery.VerifyBy(t.Signers(), context, msg, sig)
}
// VaultAuthority is the vault's authority certificate as read from `signers`, or "".
func (t *meshTrust) VaultAuthority() string {
if t == nil {
return ""
}
t.mu.Lock()
defer t.mu.Unlock()
return t.authority
}
// Trusts is what the record says it trusts: each signer's key and the vault's authority, by fingerprint.
func (t *meshTrust) Trusts() []string {
if t == nil {
return nil
}
authorities := map[string]string{}
if a := t.VaultAuthority(); a != "" {
authorities[delivery.SignerVault] = a
}
return delivery.Trusts(t.Signers(), authorities)
}
// change applies f under the lock and tells who follows when Trusts differs after it.
func (t *meshTrust) change(f func()) {
before := strings.Join(t.Trusts(), "\n")
t.mu.Lock()
f()
changed := t.changed
t.mu.Unlock()
if strings.Join(t.Trusts(), "\n") != before && changed != nil {
changed()
}
}
// heardSigner takes one key of the vault's `signers`: its key or its authority, refused (and said) when it does not
// check; a deleted key is no longer trusted.
func (t *meshTrust) heardSigner(key string, value []byte, deleted bool, say link.Announce) {
switch key {
case delivery.SignersKeyKey:
var pub ed25519.PublicKey
if !deleted && len(value) > 0 {
k, err := delivery.DecodeSignerKey(value)
if err == nil {
pub, err = k.PublicKey()
}
if err != nil {
say("the vault's signing key in " + delivery.SignersRead + " is not trusted: " + err.Error())
return
}
}
t.change(func() { t.vault = pub })
case delivery.SignersAuthorityKey:
pem := ""
if !deleted && len(value) > 0 {
a, err := delivery.DecodeSignerAuthority(value)
if err != nil {
say("the vault's authority in " + delivery.SignersRead + " is not trusted: " + err.Error())
return
}
pem = a.Certificate
}
t.change(func() { t.authority = pem })
}
}
// heardLease takes the controller's lease key as read: once handed away to a holder that is no controller, the
// controller's key is trusted no more. A lease that cannot be read changes nothing.
func (t *meshTrust) heardLease(raw []byte, found bool, readErr error) {
if readErr != nil {
return
}
handed, _ := buscompose.LeaseHandedAway(raw, found, nil)
t.change(func() { t.handedAway = handed })
}
// trustFollowAgain bounds the wait before a watch that could not be made is tried again on the same link: the vault may
// not serve `sign` yet, and until it does this node is not granted the reads.
var trustFollowAgain = 2 * time.Minute
// followTrust follows the vault's `signers` and the controller's lease on every link, as watches (never a poll), for as
// long as ctx lasts. A watch that cannot be made is said once per reason and tried again.
func followTrust(ctx context.Context, session *link.Session, trust *meshTrust, say link.Announce) {
said := ""
sayOnce := func(line string) {
if line != said {
said = line
if line != "" {
say(line)
}
}
}
for ctx.Err() == nil {
linked, cancel, err := session.WhenLinked(ctx)
if err != nil {
return
}
watching, stop := context.WithCancel(linked)
signers, serr := session.KVWatch(watching, delivery.SignersBucket, ">")
lease, lerr := session.KVWatch(watching, buscompose.LeaseBucket, buscompose.ControllerLeaseKey)
if serr != nil || lerr != nil {
stop()
sayOnce(fmt.Sprintf("what this node trusts a mesh signature from is not followed yet (the vault's signers: %v; "+
"the controller's lease: %v): it trusts what it last read, and looks again in %s", serr, lerr, trustFollowAgain))
signers, lease = nil, nil
} else {
sayOnce("")
}
for signers != nil || lease != nil {
select {
case u, ok := <-signers:
if !ok {
signers = nil
continue
}
if !u.Caught {
trust.heardSigner(u.Key, u.Value, u.Deleted, say)
}
case u, ok := <-lease:
if !ok {
lease = nil
continue
}
if !u.Caught {
trust.heardLease(u.Value, !u.Deleted && len(u.Value) > 0, nil)
}
}
}
stop()
select {
case <-linked.Done():
case <-time.After(trustFollowAgain):
}
cancel()
}
}
// placedAllHave says whether every module the seat mesh-placement places on node has a have of its own, read from the
// bus: then the kept declaration is read no more (keptRetired). A node the seat does not place, or whose placement or haves
// cannot be read, keeps reading it.
func placedAllHave(ctx context.Context, session *link.Session, node string) (bool, string) {
key, err := delivery.PlacementKey(node)
if err != nil {
return false, err.Error()
}
h, err := session.KVHistory(ctx, delivery.PlacementBucket, key)
if err != nil || len(h) == 0 || h[len(h)-1].Deleted {
return false, "its placement is not read"
}
p, err := delivery.DecodePlacement(h[len(h)-1].Value)
if err != nil {
return false, err.Error()
}
return allHave(p, func(module string) (bool, error) {
_, ok, err := session.ReadHaveErr(ctx, module)
return ok, err
})
}
// allHave says whether has finds a have for every module p places, and why not.
func allHave(p delivery.Placement, has func(module string) (bool, error)) (bool, string) {
if len(p.Modules) == 0 {
return false, "the seat mesh-placement places no module here"
}
for _, m := range p.Modules {
ok, err := has(m.Module)
if err != nil {
return false, "the have of " + m.Module + " is not read: " + err.Error()
}
if !ok {
return false, m.Module + " has no have yet"
}
}
return true, fmt.Sprintf("each of the %d modules placed here has a have of its own", len(p.Modules))
}
// certifyThisNode asks the vault's `sign` for this node's own certificate (kind node-certificate, ADR 0550 §2), checks
// it is for this node's internal name and serving key and signed by the authority the answer names, which must be the
// vault's this node trusts, keeps it where the record reads it, and writes the record again (written).
func certifyThisNode(ctx context.Context, state, node string, trust *meshTrust,
ask func(ctx context.Context, verb string, args any) (json.RawMessage, error), written func()) error {
serving, err := identity.LoadServingKey(identity.ServingKeyPath(state))
if err != nil {
return err
}
key, err := base64.StdEncoding.DecodeString(serving.Public)
if err != nil || len(key) != ed25519.PublicKeySize {
return fmt.Errorf("this node's serving key is not an ed25519 key")
}
raw, err := ask(ctx, delivery.SignVerb, map[string]any{"kind": delivery.KindNodeCertificate, "node": node})
if err != nil {
return err
}
var a delivery.SignAnswer
if err := json.Unmarshal(raw, &a); err != nil {
return fmt.Errorf("the vault's answer is not a signing: %w", err)
}
trusted := trust.VaultAuthority()
if trusted == "" {
return fmt.Errorf("this node has read no authority of the vault's to check the certificate against")
}
if err := delivery.CheckNodeCertificate(a.Certificate, []string{trusted}, overlay.InternalName(node),
ed25519.PublicKey(key), time.Now()); err != nil {
return err
}
path := identity.CertificatePath(state)
if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil {
return err
}
if err := os.WriteFile(path, []byte(a.Certificate), 0o644); err != nil {
return err
}
if written != nil {
written()
}
return nil
}
+59
View File
@@ -1,6 +1,9 @@
package main
import (
"errors"
"os"
"path/filepath"
"strings"
"testing"
"time"
@@ -9,6 +12,7 @@ import (
"github.com/novox/mesh-host/internal/accounts"
"github.com/novox/mesh-host/internal/link"
"github.com/novox/mesh-host/internal/liveness"
"github.com/novox/mesh-host/internal/store"
"github.com/novox/mesh-host/internal/units"
)
@@ -103,6 +107,61 @@ func TestTheStatementSaysAnAccountJudgedForRoot(t *testing.T) {
if got["claude-code.agent"] != "never" || got["openrazer.account"] != "" {
t.Fatalf("root said: %v", got)
}
// 774 (ADR 0563 §6): the node seat's root-free passes only on the engine's own positive verdict at most 15 minutes old
// with execute not served here; a stale, missing or negative verdict, or execute served, is no pass and says why.
t.Run("this machine's root-free", func(t *testing.T) {
agent := accounts.Verdict{Account: accounts.Account{Module: "claude-code", ID: "claude-code.agent", Name: "agent", Root: true},
State: accounts.Healthy, Since: at}
free := &accounts.Statement{At: at, Accounts: []accounts.Verdict{agent}}
for _, row := range []struct {
name string
st *accounts.Statement
after time.Duration
served bool
free bool
why string
}{
{name: "a positive verdict", st: free, after: 14 * time.Minute, free: true, why: "the account agent cannot become root"},
{name: "a stale verdict", st: free, after: 16 * time.Minute, why: "more than 15 minutes ago"},
{name: "no verdict", st: nil, why: "judged no account yet"},
{name: "a way to root", st: as, why: "in the group docker, which grants root"},
{name: "no agent account", st: &accounts.Statement{At: at}, why: "agents run as the operator's account"},
{name: "execute served", st: free, served: true, why: "the login shell runs any command"},
} {
v := judgeRootFree("ace", row.st, at.Add(row.after), row.served, "the bus hears execute answered here")
if v.Free != row.free || !strings.Contains(v.Why, row.why) || v.Machine != "ace" {
t.Errorf("%s: %+v", row.name, v)
}
}
// execute is served while the holder's setting file says anything but withhold, or the bus hears it.
dir := t.TempDir()
statePath, conf := filepath.Join(dir, "state.json"), filepath.Join(dir, "tools.conf")
if err := store.Save(statePath, store.State{Resources: []store.Applied{{ID: "zsh.tools-conf", Type: "file", Target: conf}}}); err != nil {
t.Fatal(err)
}
quiet := func() (bool, error) { return false, nil }
for _, row := range []struct {
name, setting string
heard func() (bool, error)
served bool
}{
{name: "withheld and not heard", setting: "# the mesh's\nexecute=withhold\n", heard: quiet},
{name: "served", setting: "execute=serve\n", heard: quiet, served: true},
{name: "withheld and heard", setting: "execute=withhold\n", heard: func() (bool, error) { return true, nil }, served: true},
{name: "the bus not asked", setting: "execute=withhold\n", heard: func() (bool, error) { return false, errors.New("no link") }, served: true},
{name: "no setting file", heard: quiet, served: true},
} {
_ = os.Remove(conf)
if row.setting != "" {
if err := os.WriteFile(conf, []byte(row.setting), 0o644); err != nil {
t.Fatal(err)
}
}
if served, why := executeServedHere(statePath, row.heard); served != row.served {
t.Errorf("%s: served %v (%s)", row.name, served, why)
}
}
})
}
// A verdict waiting for the setuid search says how far it got, in the reason the root-free verb repeats and as
+37
View File
@@ -4,6 +4,7 @@ import (
"crypto/ed25519"
"encoding/json"
"errors"
"path/filepath"
"strings"
"testing"
@@ -11,6 +12,7 @@ import (
"git.novox.be/novox/mesh-sdk/go/delivery"
"github.com/novox/mesh-host/internal/agent"
"github.com/novox/mesh-host/internal/store"
)
func aFileOf(module, id string) json.RawMessage {
@@ -59,6 +61,41 @@ func TestDeclaredNowReadsWhatRunsWhenTheAgentIsRunning(t *testing.T) {
if err != nil || d.Provisioner != "from-agent" {
t.Fatalf("declaredNow read %+v (%v), not what runs", d, err)
}
// L3 s9 (ADR 0550 §2): a node whose placed modules all have a have reads the kept declaration no more, though the
// file is there and readable and the controller's key is dropped; until then it reads it.
state := filepath.Join(t.TempDir(), "state", "state.json")
if err := store.SaveDeclared(store.DeclaredPath(state), store.Declared{Declaration: []byte(`{"declaration":1,` +
`"resources":[{"id":"zsh.rc","type":"file","module":"zsh","path":"/etc/zshrc","content":"x"}]}`)}); err != nil {
t.Fatal(err)
}
if d, err := keptDeclaration(state, nil); err != nil || len(d.Resources) != 1 {
t.Fatalf("before the read is retired: %+v, %v", d, err)
}
placed := delivery.Placement{Node: "n1", Modules: []delivery.PlacedModule{{Module: "zsh"}, {Module: "lemurs"}}}
for _, row := range []struct {
name string
has map[string]bool
want bool
}{
{"every placed module has a have", map[string]bool{"zsh": true, "lemurs": true}, true},
{"one has none yet", map[string]bool{"zsh": true}, false},
} {
if got, why := allHave(placed, func(m string) (bool, error) { return row.has[m], nil }); got != row.want {
t.Errorf("%s: %v (%s)", row.name, got, why)
}
}
if got, _ := allHave(delivery.Placement{Node: "n1"}, nil); got {
t.Error("a node with nothing placed retires the read")
}
keptRetired.Store(true)
defer keptRetired.Store(false)
if _, err := keptDeclaration(state, nil); !errors.Is(err, errKeptRetired) {
t.Errorf("after the read is retired: %v", err)
}
if by, filled := keptOnce(options{state: state}, nil)(); by != nil || filled != nil {
t.Errorf("after the read is retired the kept declaration was split: %v %v", by, filled)
}
}
// 1945/AC17 (unit): a resource the node's own composition shares with a running module's composition is declared once,
+1 -1
View File
@@ -3,7 +3,7 @@ module github.com/novox/mesh-host
go 1.26.0
require (
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261028234900-c2999a4e13cc
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261030001116-2da0de75c3f0
github.com/nats-io/nats-server/v2 v2.11.17
github.com/nats-io/nats.go v1.54.0
golang.org/x/crypto v0.57.0
+2 -2
View File
@@ -1,5 +1,5 @@
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261028234900-c2999a4e13cc h1:mOk2tAKTiuvnap/BUIYz91lLVbHi/EB6JPAD6KxTPvc=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261028234900-c2999a4e13cc/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261030001116-2da0de75c3f0 h1:lKU3HjNXGiVbLW2hUgFjU/TMTUYMlNnnmZxG5T7biQA=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261030001116-2da0de75c3f0/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
github.com/antithesishq/antithesis-sdk-go v0.7.0-default-no-op h1:Z/MZK75wC/NSrkgqeNIa7jexam9uWzhLmFTSCPI/kn0=
github.com/antithesishq/antithesis-sdk-go v0.7.0-default-no-op/go.mod h1:FQyySiasQQM8735Ddel3MRojmy4dA1IqCeyJ5jmPMbI=
github.com/google/go-tpm v0.9.8 h1:slArAR9Ft+1ybZu0lBwpSmpwhRXaa85hWtMinMyRAWo=
+11
View File
@@ -240,6 +240,17 @@ func (c *SetuidCache) SearchAgain() {
c.forgetProgress()
}
// Running says whether a search walks now (novox/hq ADR 0563 §2: the node seat starts no second while one does).
// Nil c runs none.
func (c *SetuidCache) Running() bool {
if c == nil {
return false
}
c.mu.Lock()
defer c.mu.Unlock()
return c.running
}
// SearchNow throws away the last complete search, kept or in memory, and any walk under way or kept, and has
// the next look start a full search at once, past AgainAfter and any back-off (novox/hq issue 361): the
// operator changed by hand what the last one found, which no apply says. Until the new one completes the
+348 -14
View File
@@ -109,7 +109,9 @@ type Env struct {
Kept func() map[string][]json.RawMessage
// KeptFilled is the paths the kept declaration names as filled by the provisioner, by the module whose file is at
// each (the node's own under NodeModule). A module that still runs from the kept declaration keeps its filled files
// filled: without them the filling refused each one as "not named as filled" (issue 1945). Nil names none.
// filled: without them the filling refused each one as "not named as filled" (issue 1945). Nil names none. A module
// whose have names a composition the bucket keeps runs from that composition, never from the kept declaration
// (issue 924): the kept declaration is older than every composition applied since the switch.
KeptFilled func() map[string][]string
// Trusted is the public keys a composition may be signed by: the controller's and the operator's (ADR 0539 §3). A
// composition unsigned, signed by another key, or changed after it was signed is refused and nothing of it is
@@ -131,6 +133,12 @@ type Env struct {
// them. The node-engine's filling tells the provisioner's client then to fill again what it refused (mesh-host
// internal/filling Knock; ace, 2026-10-25 21:17). Nil tells nothing.
Settled func(module string)
// RunsChanged is told, outside the agent's lock, that what runs (Everything) changed other than by an apply or a
// removal: once at the start, when the haves and the compositions they name are read, and each time a module is
// taken as running from its have or the kept declaration. Like Settled, the filling then tells the provisioner's
// client to fill again what it refused (issue 924, the second meeting: ace, g14 and shanks on 2026-10-29 at 00:09
// UTC refused power's three new drop-ins from an engine start on, as no module settled after it). Nil tells nothing.
RunsChanged func(why string)
// Backoff is how long to wait before the try-th retry (1, 2, …) of an apply that failed transiently. Nil retries a
// transient failure only on RetryTransient or a change of either key.
Backoff func(try int) time.Duration
@@ -143,6 +151,27 @@ type Env struct {
// move of a module of policy record that recreates a container or moves its data waits for a person's yes in its
// want (ADR 0547 §3). Nil judges no move.
Policy func(ctx context.Context, module string) string
// HoldsData says whether a module's manifest declares data (catalogue.Manifest.DataItems), and whether its manifest
// could be read: what a composition's Data is, for what runs of a module from the kept declaration, which carries no
// Data of its own. A composition says Data, so without it a data module from the kept declaration is judged to move
// its data against its first composition after a restart (mongodb and mssql on novox, issue 924). Nil reads none.
HoldsData func(ctx context.Context, module string) (data, known bool)
// Applied keeps every composition the agent applied, by module (the node's own under NodeModule), beside the engine's
// state on the machine: read at the start before anything is watched, so what runs is known before the node composes
// its own for the first time, whatever the bucket and the kept declaration still hold (issue 774: on 2026-10-29 the
// bus node, its kept declaration retired and its own composition in no bucket, composed its own as if its folded
// modules ran nothing and removed the bus server and four held modules). Nil keeps none.
Applied AppliedStore
}
// AppliedStore is where the engine keeps the compositions it applied (Env.Applied): each module's last, as its bytes.
type AppliedStore interface {
// Load is every composition kept, by module; a store that cannot be read says so and loads none.
Load() (map[string][]byte, error)
// Keep keeps the composition the module was applied from, in place of the one before.
Keep(module string, raw []byte) error
// Drop forgets the module's, once it is removed.
Drop(module string) error
}
// transientFor is how many transient retries say the phase applying before the have says failed: past them the bus has
@@ -160,6 +189,13 @@ type Agent struct {
// applied is the composition each module was last applied from: a revision a want named from the key's history is
// not the key's latest value, and what it holds is still held out of the others' applies.
applied map[string]compNow
// had is, for each module whose have names a composition the bucket keeps (its latest value or a revision), that
// composition, read once at the start (loadHad): what runs of a module the agent has not taken yet, ahead of the kept
// declaration (issue 924).
had map[string]compNow
// keptData is, for each module of the kept declaration whose manifest could be read, whether it declares data
// (Env.HoldsData): the Data of what runs of it from the kept declaration.
keptData map[string]bool
// attempt is the last apply of a module that failed or is held, with the revisions it was made from.
attempt map[string]attempt
checked map[string]bool
@@ -218,7 +254,7 @@ func New(env Env) *Agent {
if env.Say == nil {
env.Say = func(string) {}
}
return &Agent{env: env, wants: map[string]wantNow{}, comps: map[string]compNow{}, running: map[string]delivery.Have{}, applied: map[string]compNow{},
return &Agent{env: env, wants: map[string]wantNow{}, comps: map[string]compNow{}, running: map[string]delivery.Have{}, applied: map[string]compNow{}, had: map[string]compNow{}, keptData: map[string]bool{},
attempt: map[string]attempt{}, checked: map[string]bool{}, prior: map[string]bool{}, said: map[string]string{}, notComposed: map[string]notComposedMark{},
retry: make(chan []string, 8), reconcile: make(chan []string, 64), tries: map[string]int{}, again: make(chan againAsk, 64),
transient: make(chan struct{}, 1), stopped: make(chan struct{})}
@@ -228,6 +264,9 @@ func New(env Env) *Agent {
// before any module's, then each change as it comes.
func (a *Agent) Run(ctx context.Context) error {
defer close(a.stopped)
// What this engine last applied, before any composition is watched: the node composing its own reads what runs
// (Everything) from its first pass on, and that pass comes before the haves are read below (issue 774).
a.loadApplied(ctx)
wants, err := a.env.Wants.Watch(ctx, "want."+a.env.Node+".>")
if err != nil {
return fmt.Errorf("the agent cannot watch its wants: %w", err)
@@ -236,7 +275,7 @@ func (a *Agent) Run(ctx context.Context) error {
if err != nil {
return fmt.Errorf("the agent cannot watch its compositions: %w", err)
}
var wantsCaught, compsCaught bool
var wantsCaught, compsCaught, hadRead bool
dirty := map[string]bool{}
for {
select {
@@ -303,7 +342,17 @@ func (a *Agent) Run(ctx context.Context) error {
}
// Nothing is applied until what was there at the start is all in: a module's composition may arrive after its
// want, and the node's own composition comes before any module's.
if !(wantsCaught && compsCaught) || len(dirty) == 0 {
if !(wantsCaught && compsCaught) {
continue
}
if !hadRead {
// Before the first settle: what runs of each module with a have, from the composition it names, so the
// filling judges by it at once and not by the kept declaration until some module settles (issue 924).
hadRead = true
a.loadHad(ctx)
a.runsChanged("the haves were read at the start")
}
if len(dirty) == 0 {
continue
}
a.settle(ctx, dirty)
@@ -726,7 +775,12 @@ func (a *Agent) revisionNamed(ctx context.Context, module, fp, digest string) (c
if err != nil {
continue
}
dig, err := delivery.VerifyComposed(c, trusted)
var dig string
if a.env.SelfComposed {
dig, err = delivery.ComposedDigestOf(c) // the node's own: unsigned (ADR 0547 §2)
} else {
dig, err = delivery.VerifyComposed(c, trusted)
}
if err != nil || dig != digest {
continue
}
@@ -755,22 +809,33 @@ func (a *Agent) others(except string) []delivery.Composed {
out = append(out, c.C)
} else if c, ok := a.applied[m]; ok && c.Dig == h.Composed {
out = append(out, c.C)
} else if c, ok := a.had[m]; ok && c.Dig == h.Composed {
out = append(out, c.C) // a revision its have named, replaced since under the same key
}
}
hadNames := make([]string, 0, len(a.had))
for m := range a.had {
if _, held := a.running[m]; !held && m != except {
hadNames = append(hadNames, m)
}
}
sort.Strings(hadNames)
for _, m := range hadNames {
out = append(out, a.had[m].C)
}
if a.env.Kept != nil {
kept := a.env.Kept()
names := make([]string, 0, len(kept))
for m := range kept {
if _, held := a.running[m]; !held && m != except {
_, held := a.running[m]
_, had := a.had[m]
if !held && !had && m != except {
names = append(names, m)
}
}
sort.Strings(names)
for _, m := range names {
c := delivery.Composed{Node: a.env.Node, Module: m, Resources: kept[m]}
if a.env.KeptFilled != nil {
c.Filled = a.env.KeptFilled()[m]
}
c, _ := a.keptComposed(m, kept[m])
if m == NodeModule {
c.Module = ""
}
@@ -780,6 +845,188 @@ func (a *Agent) others(except string) []delivery.Composed {
return out
}
// loadHad reads, once at the start, the have of every module with a want or in the kept declaration and, where it names
// a composition, that composition: the bucket's value at its key when its digest is the have's, else the revision the
// bucket keeps with that digest. A have naming none, or one the bucket no longer keeps, leaves the module to the kept
// declaration, said once (issue 924).
func (a *Agent) loadHad(ctx context.Context) {
a.mu.Lock()
seen := map[string]bool{}
for m, w := range a.wants {
if m != NodeModule && w.Present && !w.Want.Removed {
seen[m] = true
}
}
if a.env.Kept != nil {
for m := range a.env.Kept() {
if m != NodeModule {
seen[m] = true
}
}
}
a.mu.Unlock()
names := make([]string, 0, len(seen))
for m := range seen {
names = append(names, m)
}
sort.Strings(names)
a.readKeptData(ctx, names...)
for _, m := range names {
if ctx.Err() != nil {
return
}
h, ok := a.env.Haves.Read(ctx, m)
if !ok || h.Composed == "" || !haveRuns(h) {
continue
}
a.mu.Lock()
if c, kept := a.had[m]; kept && c.Dig == h.Composed {
a.mu.Unlock()
continue // what this engine applied and kept itself (loadApplied)
}
c, found := a.compositionFor(m, h.FP)
a.mu.Unlock()
if !found || c.Err != nil || c.Dig != h.Composed {
c, found = a.revisionNamed(ctx, m, h.FP, h.Composed)
}
a.mu.Lock()
if found {
a.had[m] = c
} else if _, applied := a.had[m]; applied {
// Kept as this engine applied it (loadApplied), never composed as nothing.
} else if _, kept := a.keptHas(m); kept {
a.sayOnce("had:"+m, fmt.Sprintf("%s's have names the composition %s, which the bucket no longer keeps: it runs "+
"from the kept declaration until it is applied or taken", m, short(h.Composed)))
}
a.mu.Unlock()
}
}
// loadApplied takes, once at the start, what this engine last applied of each module as what runs of it, ahead of the
// bucket and the kept declaration (Env.Applied, issue 774): it is what the machine holds, whatever the bus still keeps.
// One whose have names another composition runs as applied until loadHad reads that one from the bucket; one whose
// have cannot be read, or that has none, is kept as applied, so a module is never composed as running nothing.
func (a *Agent) loadApplied(ctx context.Context) {
if a.env.Applied == nil {
return
}
kept, err := a.env.Applied.Load()
if err != nil {
a.env.Say(fmt.Sprintf("the compositions this engine applied cannot be read, so what runs is read from the haves alone: %v", err))
}
// At an engine's first start nothing is kept yet: the node's own runs as the kept declaration has it, which the store
// keeps now, so the node never composes its own as if its folded modules ran nothing once the declaration retires
// (issue 774, review of the re-land, round 1). Taken whatever digest the node's have names: the composition that names
// is in no bucket, and the declaration is the one record of what runs on disk.
seeded := false
if _, own := kept[NodeModule]; !own && err == nil && a.env.Kept != nil {
if rs, ok := a.env.Kept()[NodeModule]; ok && len(rs) > 0 {
c, _ := a.keptComposed(NodeModule, rs)
c.Module = ""
raw, eerr := delivery.EncodeComposed(c)
if eerr == nil {
eerr = a.env.Applied.Keep(NodeModule, raw)
}
if eerr != nil {
a.env.Say(fmt.Sprintf("the node's own as the kept declaration has it cannot be kept: %v", eerr))
} else {
if kept == nil {
kept = map[string][]byte{}
}
kept[NodeModule], seeded = raw, true
a.env.Say("the node's own runs as the kept declaration has it, kept now as what this engine runs")
}
}
}
names := make([]string, 0, len(kept))
for m := range kept {
names = append(names, m)
}
sort.Strings(names)
for _, m := range names {
if ctx.Err() != nil {
return
}
c, err := delivery.DecodeComposed(kept[m])
var dig string
if err == nil {
dig, err = delivery.ComposedDigestOf(c)
}
if err != nil {
a.env.Say(fmt.Sprintf("the composition this engine kept of %s cannot be read, so what runs of it is read from its have: %v", m, err))
continue
}
h, found, unread := a.readHave(ctx, m)
if seeded && m == NodeModule {
found, unread = true, false // what the kept declaration has runs, whatever the have names
} else if found && h.Composed != "" && h.Composed != dig && haveRuns(h) {
// The composition the have names replaces this once loadHad finds it in the bucket; until then, and when
// the bucket keeps it no more, what this engine applied runs (issue 774, review of the re-land, round 1).
a.env.Say(fmt.Sprintf("%s's have names the composition %s, not %s this engine applied last: it runs as applied "+
"until the one its have names is read", m, short(h.Composed), short(dig)))
}
if m == NodeModule {
c.Module, c.FP = "", ""
}
a.mu.Lock()
a.had[m] = compNow{Raw: kept[m], C: c, Dig: dig}
a.mu.Unlock()
switch {
case unread:
a.env.Say(fmt.Sprintf("%s's have cannot be read: it runs as this engine applied it, %s", m, short(dig)))
case !found:
a.env.Say(fmt.Sprintf("%s has no have: it runs as this engine applied it, %s", m, short(dig)))
}
}
}
// keepApplied keeps the composition a module was applied from (Env.Applied); one that cannot be kept is said, and the
// apply stands.
func (a *Agent) keepApplied(module string, comp compNow) {
if a.env.Applied == nil {
return
}
raw := comp.Raw
if len(raw) == 0 {
var err error
if raw, err = delivery.EncodeComposed(comp.C); err != nil {
a.env.Say(fmt.Sprintf("the composition %s was applied from cannot be kept: %v", module, err))
return
}
}
if err := a.env.Applied.Keep(module, raw); err != nil {
a.env.Say(fmt.Sprintf("the composition %s was applied from cannot be kept: %v", module, err))
}
}
// haveRuns says a have names what runs, in the phases baseline takes as running: running (or empty), held for a failed
// backup, or a composer's mark over what still runs. A have backing up, applying, checking, failed or waiting names a
// composition not applied, or not yet (issue 924, review of 031e106).
func haveRuns(h Have) bool {
switch delivery.Phase(h.Phase) {
case "", delivery.PhaseRunning, delivery.PhaseHeld:
return true
}
return composerMark(h)
}
// keptHas says whether the kept declaration has resources for the module. Called holding a.mu or not: Kept takes no lock
// of the agent's.
func (a *Agent) keptHas(module string) ([]json.RawMessage, bool) {
if a.env.Kept == nil {
return nil, false
}
rs, ok := a.env.Kept()[module]
return rs, ok
}
// runsChanged tells RunsChanged, outside the agent's lock.
func (a *Agent) runsChanged(why string) {
if a.env.RunsChanged != nil {
a.env.RunsChanged(why)
}
}
// unreadableModules are the keys of the module compositions in the node's bucket that cannot be read or are refused, of a
// module the assigner has not removed: for each module only the composition its want names, else its newest, since a
// stale one nobody applies (written before a rule the SDK now holds) must not hold the node's own back. Called holding
@@ -901,7 +1148,8 @@ func (a *Agent) heldOut(except string) []delivery.Composed {
}
}
if len(rs) > 0 {
out = append(out, delivery.Composed{Node: a.env.Node, Module: m, Resources: rs})
// Held out, never judged: its Data from its manifest where known, as every other composition says it.
out = append(out, delivery.Composed{Node: a.env.Node, Module: m, Resources: rs, Data: a.keptData[m]})
}
}
return out
@@ -916,6 +1164,34 @@ func (a *Agent) Everything() []delivery.Composed {
return a.others("\x00")
}
// Placed is the modules a want still places on the node (present and not removed), held ones included, but except:
// what no apply but the module's own may remove (apply.Scope.Placed, issue 774). Called without a.mu held.
func (a *Agent) Placed(except string) map[string]bool {
a.mu.Lock()
defer a.mu.Unlock()
out := map[string]bool{}
for m, w := range a.wants {
if m != NodeModule && m != except && w.Present && !w.Want.Removed {
out[m] = true
}
}
return out
}
// RemovedByWant is the modules whose want says removed (apply.Scope.RemovedByWant, issue 774): in the node's own apply,
// only their resources may go. Called without a.mu held.
func (a *Agent) RemovedByWant() map[string]bool {
a.mu.Lock()
defer a.mu.Unlock()
out := map[string]bool{}
for m, w := range a.wants {
if m != NodeModule && w.Present && w.Want.Removed {
out[m] = true
}
}
return out
}
// HoldsNode says whether the agent runs the node's own composition, so the node's parts are its and not the kept
// declaration's.
func (a *Agent) HoldsNode() bool {
@@ -1048,6 +1324,7 @@ func (a *Agent) baseline(ctx context.Context, module string, w wantNow, comp com
a.running[module] = delivery.Have{FP: h.FP, Composed: h.Composed, Phase: delivery.PhaseRunning}
}
a.mu.Unlock()
a.runsChanged(module + " was taken as held")
return true
}
if h.Composed != "" {
@@ -1068,6 +1345,7 @@ func (a *Agent) baseline(ctx context.Context, module string, w wantNow, comp com
a.mu.Lock()
a.running[module] = delivery.Have{FP: h.FP, Composed: h.Composed, Phase: delivery.PhaseRunning}
a.mu.Unlock()
a.runsChanged(module + " was taken from its have")
// Said again, so the session holds the module and its key carries the health this engine judges,
// not the one an older engine left there (xorg's unknown on g14 and shanks, 2026-10-25).
a.writeHave(ctx, Have{Module: module, FP: h.FP, Composed: h.Composed,
@@ -1096,6 +1374,7 @@ func (a *Agent) baseline(ctx context.Context, module string, w wantNow, comp com
a.mu.Lock()
a.running[module] = delivery.Have{FP: w.Want.FP, Composed: w.Want.Composed, Phase: delivery.PhaseRunning}
a.mu.Unlock()
a.runsChanged(module + " was taken from the kept declaration")
a.writeHave(ctx, Have{Module: module, FP: w.Want.FP, Composed: w.Want.Composed, Phase: string(delivery.PhaseRunning)})
return true
}
@@ -1184,6 +1463,7 @@ func (a *Agent) apply(ctx context.Context, module string, w wantNow, comp compNo
a.running[module] = delivery.Have{FP: fp, Composed: dig, Phase: delivery.PhaseRunning}
a.applied[module] = comp
a.mu.Unlock()
a.keepApplied(module, comp)
if a.env.Settled != nil {
a.env.Settled(module)
}
@@ -1213,8 +1493,14 @@ func (a *Agent) remove(ctx context.Context, module string, w wantNow) {
return
}
delete(a.attempt, module)
delete(a.had, module)
a.running[module] = delivery.Have{Phase: delivery.PhaseRemoved}
a.mu.Unlock()
if a.env.Applied != nil {
if err := a.env.Applied.Drop(module); err != nil {
a.env.Say(fmt.Sprintf("the composition %s was applied from cannot be forgotten now it is removed: %v", module, err))
}
}
if a.env.Settled != nil {
a.env.Settled(module)
}
@@ -1286,6 +1572,7 @@ func (a *Agent) waitsForYes(ctx context.Context, module string, w wantNow, comp
if module == NodeModule || a.env.Policy == nil || a.env.Policy(ctx, module) != catalogue.PolicyRecord || w.Want.Yes == comp.Dig {
return false
}
a.readKeptData(ctx, module) // the manifest may read now though it did not at the start
a.mu.Lock()
r, ran := a.running[module]
if ran && r.Composed == comp.Dig && r.Phase != delivery.PhaseRemoved {
@@ -1331,14 +1618,61 @@ func (a *Agent) waitsForYes(ctx context.Context, module string, w wantNow, comp
// runningComposition is what the module runs now: the composition last applied, else its resources in the declaration
// the node kept from before the switch. Called with a.mu held.
//
// The composition its have named at the start comes before the kept declaration (issue 924). From the kept declaration,
// what runs carries Data as its manifest declares it; a module whose manifest could not be read cannot be judged against
// it (false), and that is said, rather than taking it to hold no data and judging a false move of its data.
func (a *Agent) runningComposition(module string) (delivery.Composed, bool) {
if c, ok := a.applied[module]; ok && c.Err == nil {
return c.C, true
}
if a.env.Kept != nil {
if rs, ok := a.env.Kept()[module]; ok {
return delivery.Composed{Node: a.env.Node, Module: module, Resources: rs}, true
if c, ok := a.had[module]; ok {
return c.C, true
}
if rs, ok := a.keptHas(module); ok {
c, known := a.keptComposed(module, rs)
if !known {
a.sayOnce("kept-data:"+module, fmt.Sprintf("%s runs from the kept declaration, and whether it holds data cannot be "+
"read from its manifest: a move of it cannot be judged", module))
}
return c, known
}
return delivery.Composed{}, false
}
// keptComposed is what runs of a module from the kept declaration: its resources there, the paths it names as filled,
// and Data as the module's manifest declares it; false when the manifest could not be read, so Data is not known. Called
// with a.mu held.
func (a *Agent) keptComposed(module string, rs []json.RawMessage) (delivery.Composed, bool) {
c := delivery.Composed{Node: a.env.Node, Module: module, Resources: rs}
if a.env.KeptFilled != nil {
c.Filled = a.env.KeptFilled()[module]
}
data, known := a.keptData[module]
c.Data = data
return c, known || module == NodeModule
}
// readKeptData reads, outside the agent's lock, whether each module of the kept declaration not known yet declares data
// (Env.HoldsData), and keeps the answers it could read.
func (a *Agent) readKeptData(ctx context.Context, modules ...string) {
if a.env.HoldsData == nil {
return
}
for _, m := range modules {
if m == NodeModule {
continue
}
a.mu.Lock()
_, done := a.keptData[m]
a.mu.Unlock()
if _, kept := a.keptHas(m); done || !kept {
continue
}
if data, known := a.env.HoldsData(ctx, m); known {
a.mu.Lock()
a.keptData[m] = data
a.mu.Unlock()
}
}
}
+251 -11
View File
@@ -769,6 +769,162 @@ func TestEverythingIsWhatTheAgentRunsAndWhatIsKeptForTheRest(t *testing.T) {
if a.HoldsNode() {
t.Fatal("the agent holds no node composition")
}
// Issue 774 (novox, 2026-10-29 07:54 UTC): the kept declaration retired and the node's own composition in no bucket, a
// fresh engine's first own composition was composed from what runs with the node's own missing, so its folded modules
// (the bus server among them) were composed as nothing and removed. What this engine applied is kept on the machine
// and is what runs from before the first composition is watched; an apply keeps, a removal forgets.
t.Run("what this engine applied is what runs before the node composes its own", func(t *testing.T) {
b := newBed(t)
b.env.SelfComposed, b.env.Trusted, b.kept = true, nil, nil
store := &memApplied{kept: map[string][]byte{}}
b.env.Applied = store
// A folded module's resource in the node's own composition: its id names the module, as the controller folds it.
natsConf, _ := json.Marshal(map[string]any{"id": "nats.conf", "type": "file", "path": "/tmp/r6/nats/conf", "content": "listen 4222"})
own := delivery.Composed{Node: node, Resources: []json.RawMessage{natsConf}}
ownRaw, err := delivery.EncodeComposed(own)
if err != nil {
t.Fatal(err)
}
ownDig, _ := delivery.ComposedDigestOf(own)
web := delivery.Composed{Node: node, Module: "web", FP: fpOf("a"), Resources: []json.RawMessage{file("web", "x", "1")}}
webRaw, _ := delivery.EncodeComposed(web)
webDig, _ := delivery.ComposedDigestOf(web)
store.kept[NodeModule], store.kept["web"] = ownRaw, webRaw
b.haves.Write(context.Background(), Have{Module: NodeModule, Composed: ownDig})
b.haves.Write(context.Background(), Have{Module: "web", FP: fpOf("a"), Composed: webDig, Phase: string(delivery.PhaseRunning)})
// The composer's first pass reads what runs when its watch begins, before the agent reads any have.
firstCh := make(chan []delivery.Composed, 1)
comps := b.comps
var ag *Agent
b.env.Composed = watchFunc(func(ctx context.Context, p string) (<-chan Update, error) {
firstCh <- ag.Everything()
return comps.Watch(ctx, p)
})
ag = New(b.env)
b.ag = ag
ctx, cancel := context.WithCancel(context.Background())
done := make(chan struct{})
go func() { defer close(done); _ = ag.Run(ctx) }()
defer func() { cancel(); <-done }()
var first []delivery.Composed
select {
case first = <-firstCh:
case <-time.After(5 * time.Second):
t.Fatal("the composer was not watched")
}
var ranOwn, ranWeb bool
for _, c := range first {
ranOwn = ranOwn || (c.Module == "" && len(c.Resources) == 1 && strings.Contains(string(c.Resources[0]), "4222"))
ranWeb = ranWeb || c.Module == "web"
}
if !ranOwn || !ranWeb {
t.Fatalf("what runs at the node's first own composition: own %v, web %v (%d compositions); want both as this engine applied them",
ranOwn, ranWeb, len(first))
}
// An apply keeps the composition it applied; a removal forgets it.
web2 := delivery.Composed{Node: node, Module: "web", FP: fpOf("b"), Resources: []json.RawMessage{file("web", "x", "2")}}
ck, _ := delivery.ComposedKey(node, "web", fpOf("b"))
raw2, _ := delivery.EncodeComposed(web2)
comps.put(ck, raw2)
wk, _ := delivery.WantKey(node, "web")
wraw, _ := delivery.EncodeWant(delivery.Want{FP: fpOf("b"), Self: true})
b.wants.put(wk, wraw)
eventually(t, "the applied composition was not kept", func() bool { return string(store.get("web")) == string(raw2) })
rraw, _ := delivery.EncodeWant(delivery.Want{Removed: true})
b.wants.put(wk, rraw)
eventually(t, "the removed module's composition was not forgotten", func() bool { return store.get("web") == nil })
})
// Issue 774, review of the re-land, round 1: at the re-landed engine's first start nothing is kept yet, and the node's
// have names a composition the node composed itself, in no bucket. The node's own runs as the kept declaration has it
// and is kept then, so the next start, the declaration retired, still runs nats.
t.Run("at the first start nothing is kept: the node's own runs as the kept declaration has it", func(t *testing.T) {
natsConf, _ := json.Marshal(map[string]any{"id": "nats.conf", "type": "file", "path": "/tmp/r6/nats/conf", "content": "listen 4222"})
store := &memApplied{kept: map[string][]byte{}}
h := &haves{}
h.Write(context.Background(), Have{Module: NodeModule, Composed: "sha256:" + strings.Repeat("c", 64)})
firstPass := func(kept map[string][]json.RawMessage) []delivery.Composed {
b := newBed(t)
b.env.SelfComposed, b.env.Trusted, b.kept, b.env.Applied, b.env.Haves = true, nil, kept, store, h
firstCh := make(chan []delivery.Composed, 1)
comps := b.comps
var ag *Agent
b.env.Composed = watchFunc(func(ctx context.Context, p string) (<-chan Update, error) {
firstCh <- ag.Everything()
return comps.Watch(ctx, p)
})
ag = New(b.env)
ctx, cancel := context.WithCancel(context.Background())
done := make(chan struct{})
go func() { defer close(done); _ = ag.Run(ctx) }()
defer func() { cancel(); <-done }()
select {
case got := <-firstCh:
return got
case <-time.After(5 * time.Second):
t.Fatal("the composer was not watched")
}
return nil
}
runsNats := func(cs []delivery.Composed) bool {
for _, c := range cs {
if c.Module == "" && len(c.Resources) == 1 && strings.Contains(string(c.Resources[0]), "4222") {
return true
}
}
return false
}
if got := firstPass(map[string][]json.RawMessage{NodeModule: {natsConf}}); !runsNats(got) {
t.Fatalf("the first start does not run nats in the node's own (%d compositions)", len(got))
}
if store.get(NodeModule) == nil {
t.Fatal("the first start kept no node's own composition")
}
if got := firstPass(nil); !runsNats(got) {
t.Fatalf("the next start, the kept declaration retired, does not run nats in the node's own (%d compositions)", len(got))
}
})
}
// watchFunc is a Watcher from a function.
type watchFunc func(ctx context.Context, pattern string) (<-chan Update, error)
func (f watchFunc) Watch(ctx context.Context, pattern string) (<-chan Update, error) {
return f(ctx, pattern)
}
// memApplied is an AppliedStore in memory.
type memApplied struct {
mu sync.Mutex
kept map[string][]byte
}
func (m *memApplied) Load() (map[string][]byte, error) {
m.mu.Lock()
defer m.mu.Unlock()
out := map[string][]byte{}
for k, v := range m.kept {
out[k] = v
}
return out, nil
}
func (m *memApplied) Keep(module string, raw []byte) error {
m.mu.Lock()
defer m.mu.Unlock()
m.kept[module] = raw
return nil
}
func (m *memApplied) Drop(module string) error {
m.mu.Lock()
defer m.mu.Unlock()
delete(m.kept, module)
return nil
}
func (m *memApplied) get(module string) []byte {
m.mu.Lock()
defer m.mu.Unlock()
return m.kept[module]
}
// A failed apply is applied again after a restart even when the kept declaration holds the composition: once this
@@ -1026,18 +1182,102 @@ func TestTheRevisionAWantNamesIsAppliedFromHistory(t *testing.T) {
}
}
// 1945/AC18 (unit): a module that still runs from the kept declaration carries the kept declaration's filled paths of
// its files, so the filling writes them ("not named as filled" for claude-code and screen-lock, 2026-10-25).
func TestAModuleRunningFromTheKeptDeclarationKeepsItsFilledPaths(t *testing.T) {
b := newBed(t)
b.kept = map[string][]json.RawMessage{"claude-code": {file("claude-code", "settings", "")}}
b.env.KeptFilled = func() map[string][]string {
return map[string][]string{"claude-code": {"/tmp/r6/claude-code/settings"}}
// 1945/AC18 and issue 924 (unit): what runs of a module the agent has not taken yet carries the filled paths the
// filling judges by. From the kept declaration while the module has no have naming a composition ("not named as filled"
// for claude-code and screen-lock, 2026-10-25); from the composition its have names once it has one, the kept
// declaration being older, read at the start before any module settles and told to the filling then, so a file the
// newer composition fills is accepted on the first fill (power's three drop-ins on ace, g14 and shanks, 2026-10-29).
// Only a have in a phase baseline takes as running names what runs; and what runs carries Data, from the composition
// or, from the kept declaration, from the module's manifest, so a data module is not judged to move its data against
// its first composition after a restart (mongodb and mssql on novox).
func TestWhatRunsBeforeTheAgentTakesAModuleNamesItsFilledPaths(t *testing.T) {
const old, added = "/tmp/r6/power/settings", "/tmp/r6/power/bound"
filledFile := func(id string) json.RawMessage {
return json.RawMessage(`{"id":"power.` + id + `","type":"file","module":"power","path":"/tmp/r6/power/` + id +
`","mode":"0644","owner-at-runtime":"owned"}`)
}
a := New(b.env)
got := a.Everything()
if len(got) != 1 || got[0].Module != "claude-code" || strings.Join(got[0].Filled, ",") != "/tmp/r6/claude-code/settings" {
t.Fatalf("what runs from the kept declaration: %+v", got)
fp := fpOf("a")
ck, _ := delivery.ComposedKey(node, "power", fp)
applied := delivery.Composed{Node: node, Module: "power", FP: fp, Filled: []string{old, added}, Data: true,
Resources: []json.RawMessage{filledFile("settings"), filledFile("bound")}}
_, appliedRaw, appliedDig := signedBy(t, applied, ck, controllerKey)
newer := applied
newer.Resources = append([]json.RawMessage{}, applied.Resources...)
newer.Resources = append(newer.Resources, file("power", "later", "1"))
_, newerRaw, _ := signedBy(t, newer, ck, controllerKey)
for _, tc := range []struct {
name string
// have is the module's have; bucket the value its composition key holds now, history the revisions kept.
have *Have
bucket []byte
history []Update
filled, told string
// data is what runs' Data; holdsData the manifest's answer (nil: none read).
data bool
holdsData func(context.Context, string) (bool, bool)
}{
{name: "no have: the kept declaration's", filled: old},
{name: "a have naming the bucket's value", have: &Have{Module: "power", FP: fp, Composed: appliedDig},
bucket: appliedRaw, filled: old + "," + added, told: "the haves were read at the start", data: true},
{name: "a have applying names nothing that runs yet", have: &Have{Module: "power", FP: fp, Composed: appliedDig,
Phase: string(delivery.PhaseApplying)}, bucket: appliedRaw, filled: old, told: "the haves were read at the start"},
{name: "a data module from the kept declaration says Data as its manifest does", filled: old,
told: "the haves were read at the start", data: true,
holdsData: func(context.Context, string) (bool, bool) { return true, true }},
{name: "a have naming a revision replaced since", have: &Have{Module: "power", FP: fp, Composed: appliedDig},
bucket: newerRaw, history: []Update{{Key: ck, Value: appliedRaw, Revision: 3}, {Key: ck, Value: newerRaw, Revision: 7}},
filled: old + "," + added, told: "the haves were read at the start", data: true},
{name: "a have naming a composition the bucket no longer keeps", have: &Have{Module: "power", FP: fp, Composed: appliedDig},
bucket: newerRaw, filled: old, told: "the haves were read at the start"},
} {
t.Run(tc.name, func(t *testing.T) {
b := newBed(t)
b.kept = map[string][]json.RawMessage{"power": {filledFile("settings")}}
b.env.KeptFilled = func() map[string][]string { return map[string][]string{"power": {old}} }
if tc.have != nil {
b.haves.stored = map[string]Have{"power": *tc.have}
}
if tc.bucket != nil {
b.comps.before(ck, tc.bucket)
}
b.env.Revisions = func(_ context.Context, key string) ([]Update, error) { return tc.history, nil }
b.env.HoldsData = tc.holdsData
var told []string
var mu sync.Mutex
b.env.RunsChanged = func(why string) { mu.Lock(); told = append(told, why); mu.Unlock() }
filledNow := func() string {
for _, c := range b.ag.Everything() {
if c.Module == "power" {
if c.Data != tc.data {
t.Errorf("what runs says Data %v, want %v", c.Data, tc.data)
}
return strings.Join(c.Filled, ",")
}
}
return ""
}
if tc.told == "" { // no want, no have: nothing to start, and the kept declaration is what runs
a := New(b.env)
b.ag = a
if got := filledNow(); got != tc.filled {
t.Fatalf("filled %q, want %q", got, tc.filled)
}
return
}
stop := b.run()
defer stop()
eventually(t, "the filling was not told what runs at the start", func() bool {
mu.Lock()
defer mu.Unlock()
return len(told) > 0 && told[0] == tc.told
})
if got := filledNow(); got != tc.filled {
t.Fatalf("filled %q, want %q", got, tc.filled)
}
if got := b.appliedList(); len(got) != 0 {
t.Fatalf("applied %v: nothing has a want", got)
}
})
}
}
+57
View File
@@ -312,6 +312,38 @@ type Scope struct {
Module string
// Others are the resources of every other module the node has a composition for.
Others []declaration.Resource
// Placed are the modules a want still places on the node (present, not removed), held ones included: no apply but
// the module's own removes a resource of one of them (issue 774: on 2026-10-29 the node's own apply on the bus node,
// composed without what its folded modules ran, removed the bus server and four held modules). Nil guards nothing.
Placed map[string]bool
// RemovedByWant are the modules whose want says removed. In the node's own apply (Module NodeModule) with this set
// non-nil, a resource of any other module is removed only when that module is here: a module the node runs folded
// in its own composition has no want, so Placed does not name it (issue 774, review of the re-land, round 1: nats on
// the bus node has none, and its nats.conf was guessed as nats's and removed). Nil guards by Placed alone.
RemovedByWant map[string]bool
}
// placedOwner is the module a want still places on the node that the record belongs to, other than the scope's own:
// the module recorded with it, else the one its id names (a resource the node's own composition carried for a folded
// module is recorded as the node's, its id naming the module). Empty when there is none.
func (sc *Scope) placedOwner(a store.Applied) string {
if sc == nil || (len(sc.Placed) == 0 && sc.RemovedByWant == nil) {
return ""
}
m := a.Module
if m == "" || m == store.NodeModule {
m, _ = declaration.ModuleGuess(a.ID)
}
if m == "" || m == sc.Module || m == store.NodeModule {
return ""
}
if sc.Placed[m] {
return m
}
if sc.Module == store.NodeModule && sc.RemovedByWant != nil && !sc.RemovedByWant[m] {
return m // a module the node's own composition carried, which no want removed
}
return ""
}
// owns reports whether a record belongs to the scope's module.
@@ -441,6 +473,31 @@ func applyScoped(
}
}
// **A module a want still places here is never removed by another's apply** (issue 774): its resources go only with
// its own apply or its want's removal. One that would is refused whole, before anything is changed, naming them: a
// composition that leaves them out lost what runs, and removing what runs is the worst a node can do with that.
if scope != nil {
var refused []string
for _, orphan := range known.Orphans(declared, origin) {
if !scope.owns(orphan) {
continue
}
if _, kept := heldByOthers[orphan.ID]; kept {
continue
}
if m := scope.placedOwner(orphan); m != "" {
refused = append(refused, orphan.ID+" of "+m)
}
}
if len(refused) > 0 {
sort.Strings(refused)
err := fmt.Errorf("the apply of %s would remove %s, of modules a want still places here; a module's resources "+
"go only with its own apply or its want's removal, so nothing was applied", scope.Module, strings.Join(refused, ", "))
log(" refused: " + err.Error())
return report, known, &Error{Resource: "the resources of modules placed here", Err: err, Done: report}
}
}
removeOrphan := func(orphan store.Applied) error {
var action, detail string
var err error
+27 -3
View File
@@ -45,6 +45,30 @@ func WriteFilled(
log func(string),
keep Keep,
windows *Windows,
) (report Report, _ store.State, _ error) {
return WriteFilledWith(ctx, sys, d, known, path, content, run, log, keep, windows, Filled{})
}
// Filled is what the node-engine did to a filled file's content before its write.
type Filled struct {
// Secret says the node-engine filled a value into the content, a provision's `${secret:}` (novox/hq ADR 0570
// clause 6): the file is then written 0600 when the declaration gives no mode, as a file carrying a secret is.
Secret bool
}
// WriteFilledWith is WriteFilled for content the node-engine filled before the write (Filled).
func WriteFilledWith(
ctx context.Context,
sys system.System,
d *declaration.Declaration,
known store.State,
path string,
content []byte,
run Runner,
log func(string),
keep Keep,
windows *Windows,
filled Filled,
) (report Report, _ store.State, _ error) {
if log == nil {
log = func(string) {}
@@ -65,7 +89,7 @@ func WriteFilled(
return report, known, &Error{Resource: path, Err: fmt.Errorf("%d resources declare %s, and a filled file "+
"is written for exactly one, so nothing was written", len(declared), path)}
}
file := filledAs(declared[0], content)
file := filledAs(declared[0], content, filled.Secret)
if held, isHeld := known.HeldAt(file.ID); isHeld {
return report, known, &Error{Resource: file.ID, Err: fmt.Errorf("%s is held as found for a module not "+
"yet taken (%s), so nothing was written", path, held.Why)}
@@ -163,10 +187,10 @@ func WriteFilled(
}
// filledAs is the declared file with the filled content in place of the declaration's.
func filledAs(declared *declaration.File, content []byte) *declaration.File {
func filledAs(declared *declaration.File, content []byte, secret bool) *declaration.File {
f := *declared
f.Content, f.Sealed, f.VaultSealed, f.Bytes, f.Secrets = string(content), "", nil, "", nil
if f.Mode == "" && (declared.Secret() || len(declared.Secrets) > 0) {
if f.Mode == "" && (secret || declared.Secret() || len(declared.Secrets) > 0) {
// Written as the declaration would have written it: a file carrying a credential is not world-readable.
f.Mode = "0600"
}
+72 -25
View File
@@ -3,6 +3,7 @@ package apply
import (
"context"
"path/filepath"
"strings"
"testing"
"git.novox.be/novox/mesh-sdk/go/declaration"
@@ -116,32 +117,78 @@ func TestARecordWithoutAnOwnerBelongsToTheModuleItsIdNames(t *testing.T) {
// leaves it as it is, whatever its type, until the module's own apply takes it (issue 1945, live on 2026-10-25 at 02:32:
// shanks and g14 lost systemd-resolved's guard, its resolver files and sudo's operator file, ace qbittorrent's server,
// because the node's own apply removed every record no longer in it, the others holding the very same resources or not).
//
// And a resource the node's own composition carried for a module a want still places here (a folded module of the old
// form, held) is never removed by the node's own apply when a composition leaves it out: the apply is refused whole and
// says why (issue 774, live on 2026-10-29 at 07:54 UTC on the bus node: the node's own composition, composed without what
// it ran once the kept declaration was read no more, removed the bus server, the proxy and three held modules). Only its
// want's removal lets it go.
func TestAResourceThatMovedToItsModulesCompositionIsLeftByTheNodesOwnApply(t *testing.T) {
dir := t.TempDir()
fact := filepath.Join(dir, "suffix")
gone := filepath.Join(dir, "gone.conf")
own := declare(t, `{"id":"systemd-resolved.fact-suffix","type":"file","path":"`+fact+`","content":"mesh\n"},
{"id":"old.gone","type":"file","path":"`+gone+`","content":"g\n"}`)
_, state, err := ApplyModule(context.Background(), archHost(t), own, Scope{Module: store.NodeModule},
state0(), store.OriginDeclared, noServices, nil, nil, nil, nil)
if err != nil {
t.Fatal(err)
}
// systemd-resolved now has a composition of its own, not applied yet; old.gone is no module's any more.
moved := declare(t, `{"id":"systemd-resolved.fact-suffix","type":"file","path":"`+fact+`","content":"mesh\n","module":"systemd-resolved"}`)
report, state, err := ApplyModule(context.Background(), archHost(t), &declaration.Declaration{},
Scope{Module: store.NodeModule, Others: moved.Resources}, state, store.OriginDeclared, noServices, nil, nil, nil, nil)
if err != nil {
t.Fatal(err)
}
if !exists(fact) {
t.Fatalf("the moved resource was removed: %+v", outcomeOf(report, "systemd-resolved.fact-suffix"))
}
if exists(gone) {
t.Fatal("a resource no composition holds any more was kept")
}
if _, still := state.Find("systemd-resolved.fact-suffix"); !still {
t.Fatal("the moved resource's record was dropped, so its module's apply would not know it")
for _, tc := range []struct {
name string
id string // the resource the node's own composition carried, and no longer does
moved bool // its module's own composition now declares it (Others)
placed bool // a want still places its module here
refused bool // the node's own apply is refused whole
stays bool // the resource is still on the machine and recorded
// wants names the modules whose want says removed (Scope.RemovedByWant); nil leaves the guard to Placed alone.
wants []string
}{
{name: "moved to its module's own composition", id: "systemd-resolved.fact-suffix", moved: true, placed: true, stays: true},
{name: "a held folded module, in no composition", id: "nats.server-conf", placed: true, refused: true, stays: true},
{name: "its module's want removed", id: "nats.server-conf"},
// Issue 774, review of the re-land, round 1: a module folded in the node's own has no want, so Placed does not name
// it; in the node's own apply only a module a want removed loses its resources.
{name: "a folded module with no want", id: "nats.server-conf", wants: []string{}, refused: true, stays: true},
{name: "a folded module a want removed", id: "nats.server-conf", wants: []string{"nats", "old"}},
} {
t.Run(tc.name, func(t *testing.T) {
dir := t.TempDir()
fact := filepath.Join(dir, "carried")
gone := filepath.Join(dir, "gone.conf")
own := declare(t, `{"id":"`+tc.id+`","type":"file","path":"`+fact+`","content":"mesh\n"},
{"id":"old.gone","type":"file","path":"`+gone+`","content":"g\n"}`)
_, state, err := ApplyModule(context.Background(), archHost(t), own, Scope{Module: store.NodeModule},
state0(), store.OriginDeclared, noServices, nil, nil, nil, nil)
if err != nil {
t.Fatal(err)
}
module, _ := declaration.ModuleGuess(tc.id)
scope := Scope{Module: store.NodeModule}
if tc.moved {
moved := declare(t, `{"id":"`+tc.id+`","type":"file","path":"`+fact+`","content":"mesh\n","module":"`+module+`"}`)
scope.Others = moved.Resources
}
if tc.placed {
scope.Placed = map[string]bool{module: true}
}
if tc.wants != nil {
scope.RemovedByWant = map[string]bool{}
for _, m := range tc.wants {
scope.RemovedByWant[m] = true
}
}
report, after, err := ApplyModule(context.Background(), archHost(t), &declaration.Declaration{}, scope, state,
store.OriginDeclared, noServices, nil, nil, nil, nil)
if tc.refused != (err != nil) {
t.Fatalf("refused %v, want %v: %v", err != nil, tc.refused, err)
}
if tc.refused {
if !strings.Contains(err.Error(), tc.id+" of "+module) {
t.Fatalf("the refusal does not name what it kept: %v", err)
}
after = state
}
if exists(fact) != tc.stays {
t.Fatalf("the carried resource exists %v, want %v: %+v", exists(fact), tc.stays, outcomeOf(report, tc.id))
}
if _, still := after.Find(tc.id); still != tc.stays {
t.Fatalf("the carried resource's record kept %v, want %v", still, tc.stays)
}
if exists(gone) == !tc.refused {
t.Fatalf("a resource no composition holds any more exists %v, with the apply refused %v", exists(gone), tc.refused)
}
})
}
}
+111
View File
@@ -0,0 +1,111 @@
package apply
import (
"fmt"
"os"
"path/filepath"
"time"
"git.novox.be/novox/mesh-sdk/go/declaration"
)
// ObjectKeep is the node-engine's own keep of the content of files given by digest (novox/hq ADR 0486 §1, issue 924): one
// file per object, named as the object is, beside the node's state. **What a node once had it never asks the bus for
// again**, so neither the controller's sweep of the bucket (an object is removed a day after no composition of the
// controller names it) nor a bus that does not answer can refuse a composition the node already ran. The node's account
// may only read its bucket (ADR 0486 §2), so the keep is the node's, not the bucket's: nothing is written back.
//
// An object unused for KeptObjectsFor is let go when another is kept, so the keep holds what the node applied lately and
// not every export it ever ran. Every failure to keep is silent and costs only a fetch: the keep is a copy.
type ObjectKeep struct {
Dir string
// Now is the clock; nil is time.Now.
Now func() time.Time
}
// KeptObjectsFor is how long an object of the keep is held after it was last used.
const KeptObjectsFor = 14 * 24 * time.Hour
// KeepObjectsIn is the keep at dir; an empty dir keeps nothing.
func KeepObjectsIn(dir string) *ObjectKeep {
if dir == "" {
return nil
}
return &ObjectKeep{Dir: dir}
}
func (k *ObjectKeep) now() time.Time {
if k.Now != nil {
return k.Now()
}
return time.Now()
}
// get is the kept content of the object name when it is what f names, and marks it used.
func (k *ObjectKeep) get(name string, f *declaration.File) ([]byte, bool) {
if k == nil {
return nil, false
}
path := filepath.Join(k.Dir, name)
body, err := os.ReadFile(path)
if err != nil || int64(len(body)) != f.Size || f.Verify(body) != nil {
return nil, false
}
now := k.now()
_ = os.Chtimes(path, now, now)
return body, true
}
// put keeps body under name, unless it is kept already (then it is marked used), and lets go of what was unused too long.
func (k *ObjectKeep) put(name string, body []byte) {
if k == nil {
return
}
path := filepath.Join(k.Dir, name)
now := k.now()
if info, err := os.Stat(path); err == nil && info.Size() == int64(len(body)) {
_ = os.Chtimes(path, now, now)
return
}
if err := os.MkdirAll(k.Dir, 0o700); err != nil {
return
}
tmp, err := os.CreateTemp(k.Dir, ".put-*")
if err != nil {
return
}
_, werr := tmp.Write(body)
cerr := tmp.Close()
if werr != nil || cerr != nil || os.Chmod(tmp.Name(), 0o600) != nil || os.Rename(tmp.Name(), path) != nil {
_ = os.Remove(tmp.Name())
return
}
_ = os.Chtimes(path, now, now)
k.prune(now)
}
// prune lets go of every object unused for KeptObjectsFor.
func (k *ObjectKeep) prune(now time.Time) {
entries, err := os.ReadDir(k.Dir)
if err != nil {
return
}
for _, e := range entries {
info, err := e.Info()
if err != nil || !info.Mode().IsRegular() {
continue
}
if now.Sub(info.ModTime()) >= KeptObjectsFor {
_ = os.Remove(filepath.Join(k.Dir, e.Name()))
}
}
}
// never says, in a failed fetch's error before its cause (a transient one's detail keeps the tail), that the node's own
// keep did not hold the object either.
func (k *ObjectKeep) never() string {
if k == nil {
return ""
}
return fmt.Sprintf(" (this node's own keep of objects, %s, never held it)", k.Dir)
}
+24 -8
View File
@@ -27,13 +27,23 @@ type FetchObject func(ctx context.Context, name string, limit int64) ([]byte, er
// unchanged declaration, and a node that is asleep and has no bus, apply as they always did. One object is fetched
// once however many files name its digest.
func ResolveFileReferences(ctx context.Context, d *declaration.Declaration, fetch FetchObject) error {
return ResolveFileReferencesKept(ctx, d, fetch, nil)
}
// ResolveFileReferencesKept is ResolveFileReferences with the node's own keep of objects (ObjectKeep): every content had
// (a file here that verifies, or an object of the bus) is kept, and a kept one is used before the bus is asked. **A node
// depends on its bucket only for content it never had** (issue 924, 2026-10-29: novox's own composition named the
// export mesh-vault runs by a digest whose object the controller's sweep had removed, so the vault's apply failed while
// it ran; the bucket's objects are kept only while a composition of the controller names them, and a node composing its
// own names objects the controller no longer knows of). A nil keep keeps nothing.
func ResolveFileReferencesKept(ctx context.Context, d *declaration.Declaration, fetch FetchObject, keep *ObjectKeep) error {
fetched := map[string][]byte{}
for i, r := range d.Resources {
f, ok := r.(*declaration.File)
if !ok || !f.IsReference() {
continue
}
body, err := referencedContent(ctx, f, fetch, fetched)
body, err := referencedContent(ctx, f, fetch, fetched, keep)
if err != nil {
return err
}
@@ -50,31 +60,37 @@ func ResolveFileReferences(ctx context.Context, d *declaration.Declaration, fetc
}
func referencedContent(ctx context.Context, f *declaration.File, fetch FetchObject,
fetched map[string][]byte) ([]byte, error) {
fetched map[string][]byte, keep *ObjectKeep) ([]byte, error) {
if body, ok := fetched[f.Digest]; ok {
return body, nil
}
if body, ok := alreadyThere(f); ok {
fetched[f.Digest] = body
return body, nil
}
name, err := declaration.ObjectName(f.Digest)
if err != nil {
return nil, fmt.Errorf("file %s (%s): %w", f.ID, f.Path, err)
}
if body, ok := alreadyThere(f); ok {
fetched[f.Digest] = body
keep.put(name, body)
return body, nil
}
if body, ok := keep.get(name, f); ok {
fetched[f.Digest] = body
return body, nil
}
if fetch == nil {
return nil, fmt.Errorf("file %s (%s) is given by digest %s and is not on this machine, and this node-engine "+
"has no link to the bus to fetch it from", f.ID, f.Path, f.Digest)
}
body, err := fetch(ctx, name, f.Size)
if err != nil {
return nil, fmt.Errorf("file %s (%s): fetching the object %s of this node's own object bucket: %w",
f.ID, f.Path, name, err)
return nil, fmt.Errorf("file %s (%s): fetching the object %s of this node's own object bucket%s: %w",
f.ID, f.Path, name, keep.never(), err)
}
if err := f.Verify(body); err != nil {
return nil, fmt.Errorf("%s (%s): %w; nothing was applied", f.ID, f.Path, err)
}
fetched[f.Digest] = body
keep.put(name, body)
return body, nil
}
+110
View File
@@ -6,6 +6,10 @@ import (
"strings"
"testing"
"time"
stdio "git.novox.be/novox/mesh-sdk/go"
"git.novox.be/novox/mesh-sdk/go/asker"
"git.novox.be/novox/mesh-sdk/go/asks"
)
// The take act's ask (novox/hq ADR 0325 §8): a hand-over's fields and the act's own in one signed object, under a
@@ -68,6 +72,112 @@ func TestTheTakeActRunsOnlyOnAWarrant(t *testing.T) {
if a.Said != "started" || len(given) != 1 || given[0].Warrant != "w-1" || given[0].Path != "/srv/media/movies" {
t.Fatalf("a warranted ask did not reach the act whole: %+v %+v", a, given)
}
// 774 (ADR 0563 §3): a hand-over asked through the node seat is the engine's own ask, performed only on the
// operator's warrant to it: never on the call, not on a warrant for another digest, once on a warrant heard twice,
// and not when the disk changed since it was asked.
t.Run("the node seat's hand-over acts only on the operator's warrant", func(t *testing.T) {
for _, row := range []struct {
name string
warrant func(asked asks.Ask, w asks.Warrant) asks.Warrant
diskAfter string
twice bool
done int
state string
}{
{name: "no warrant", done: 0, state: HandAsked},
{name: "the operator's yes", warrant: yes, done: 1, state: HandDone},
{name: "the yes heard twice", warrant: yes, twice: true, done: 1, state: HandDone},
{name: "a warrant for another digest", warrant: func(a asks.Ask, w asks.Warrant) asks.Warrant {
w = yes(a, w)
w.AskDigest = "sha256:another"
return w
}, done: 0, state: HandAsked},
{name: "a no", warrant: func(a asks.Ask, w asks.Warrant) asks.Warrant {
w = yes(a, w)
w.Option, w.Label, w.Level = "no", "No, leave it", asks.Acknowledge
return w
}, done: 0, state: HandRefused},
{name: "the disk changed", warrant: yes, diskAfter: "0:0 700 9:2", done: 0, state: HandRefused},
{name: "expired", warrant: func(a asks.Ask, w asks.Warrant) asks.Warrant {
return asks.Warrant{Ask: a.ID, Asker: w.Asker, Outcome: asks.OutcomeExpired, At: w.At}
}, done: 0, state: HandRefused},
} {
t.Run(row.name, func(t *testing.T) {
h, asked, done := anEngine(t)
disk := "1000:1000 755 9:1"
h.JudgeHandOver = func(path string) (Judged, error) {
return Judged{Subject: path, Path: path, Sets: "owner jo, group jo, mode 0750", Disk: disk,
Words: "A directory used as found."}, nil
}
if _, err := h.Call(EngineCall{Verb: ActHandOver, Caller: "shanks.node-tools",
Args: map[string]any{"path": "/srv/x", "why": "a test"}}); err != nil {
t.Fatal(err)
}
if len(*asked) != 1 || *done != 0 {
t.Fatalf("the call asked %d and performed %d", len(*asked), *done)
}
// The act is one key, and the open ask one key made by a create (ADR 0563 §3).
if acts, _ := h.State.Keys("act."); len(acts) != 1 || acts[0] != "act."+(*asked)[0].ID {
t.Fatalf("the act is kept as %v", acts)
}
if open, _ := h.State.Keys("open."); len(open) != 1 || !strings.HasPrefix(open[0], "open.hand-over.") {
t.Fatalf("the open ask is kept as %v", open)
}
if _, err := h.Call(EngineCall{Verb: ActHandOver, Caller: "x", Args: map[string]any{"path": "/srv/x"}}); err == nil {
t.Fatal("a second ask of the same path was asked while one is open")
}
if row.diskAfter != "" {
disk = row.diskAfter
}
if row.warrant != nil {
w := row.warrant((*asked)[0], asks.Warrant{Asker: asks.EngineAsker("n1"), At: time.Now()})
h.Heard(w)
if row.twice {
h.Heard(w)
}
}
got, _ := h.Acts(0)
acts := got.(map[string]any)["acts"].([]HandAct)
if *done != row.done || len(acts) != 1 || acts[0].State != row.state {
t.Fatalf("performed %d (want %d), acts %+v (want %s)", *done, row.done, acts, row.state)
}
if open, _ := h.State.Keys("open."); (len(open) == 1) != (row.state == HandAsked) {
t.Fatalf("an ask %s leaves its open key as %v", row.state, open)
}
if row.state == HandDone && (acts[0].By != "operator" || acts[0].Via == "" || acts[0].Ask == "") {
t.Fatalf("the act does not record the warrant: %+v", acts[0])
}
})
}
})
}
// yes is the operator's yes to an ask, as the router writes it.
func yes(a asks.Ask, w asks.Warrant) asks.Warrant {
w.Ask, w.Outcome, w.Option, w.Label, w.Level = a.ID, asks.OutcomeChosen, "yes", "Yes", asks.Approve
w.By = &asks.Person{Who: asks.Operator, Kind: "telegram", Verified: "user id verified"}
w.AskDigest, w.Proofs = a.Digest(), []string{"P1"}
return w
}
// anEngine is a node seat with its own state in a temporary folder, a router that keeps what it was asked, and a
// hand-over that counts what it performed.
func anEngine(t *testing.T) (*HandActs, *[]asks.Ask, *int) {
t.Helper()
var asked []asks.Ask
done := 0
state := &MemState{}
h := &HandActs{Node: "n1", State: state,
HandOver: func(path, by string) (string, error) { done++; return path + " handed over by " + by, nil }}
h.Asker = &asker.Client{Module: asks.EngineAsker("n1"), State: state,
Publish: func(subject string, body any, id string) error {
if a, ok := body.(asks.Ask); ok && subject == asks.AskSubject(asks.EngineAsker("n1")) {
asked = append(asked, a)
}
return nil
},
Record: func(string) (*stdio.StateEntry, error) { return nil, nil }}
return h, &asked, &done
}
// The controller names an access by its declared resource id exactly, which may hold a `/`: it reaches the act,
+382
View File
@@ -0,0 +1,382 @@
package link
import (
"context"
"encoding/json"
"errors"
"fmt"
"sort"
"strings"
"time"
stdio "git.novox.be/novox/mesh-sdk/go"
"git.novox.be/novox/mesh-sdk/go/asks"
"git.novox.be/novox/mesh-sdk/go/catalogue"
"github.com/nats-io/nats.go"
"github.com/nats-io/nats.go/micro"
)
// The node seat node-engine, served on the engine's own link (novox/hq ADR 0563 §1).
//
// Each verb is answered on this node's subject alone, `mesh.seat.node-engine.tool.<verb>.<node>`, and on the same
// address naming its caller, `mesh.seat.node-engine.call.<verb>.<node>.<caller>`, whose last token the bus lets each user
// publish with its own name alone (novox/hq issue 365): the caller is that token, and a call on the subject naming none is
// from an unproven caller, recorded so. The body is the call's arguments; the answer is {result, error, node}, as every
// tool's (mesh-tools node-tools). One service, `node-engine`, answers discovery with each verb's seat, tool and node.
// EngineSeat is the seat.
const EngineSeat = catalogue.EngineSeat
// EngineToolSubject is where a verb of this node's seat is asked naming no caller.
func EngineToolSubject(verb, node string) string {
return "mesh.seat." + EngineSeat + ".tool." + verb + "." + node
}
// EngineCallPattern is where a verb of this node's seat is asked naming its caller, any caller.
func EngineCallPattern(verb, node string) string {
return "mesh.seat." + EngineSeat + ".call." + verb + "." + node + ".*"
}
// EngineEventSubject is where the seat says one of its events about this node.
func EngineEventSubject(event, node string) string {
return "mesh.seat." + EngineSeat + ".event." + event + "." + node
}
// EngineCall is one call of the seat: its verb, its caller as the bus proved it, and its arguments.
type EngineCall struct {
Verb string
Caller string
Args map[string]any
}
// EngineFunc answers one call: its result, or why it is refused.
type EngineFunc func(EngineCall) (any, error)
// CallerOfSubject is the caller a call's subject names for this node's verb: the last token of a caller-named subject,
// each `~` a dot, or stdio.Unproven on the subject that names none or on any other.
func CallerOfSubject(subject, verb, node string) string {
token, ok := strings.CutPrefix(subject, "mesh.seat."+EngineSeat+".call."+verb+"."+node+".")
if !ok || token == "" || strings.ContainsAny(token, ".*> ") {
return stdio.Unproven
}
return strings.ReplaceAll(token, "~", ".")
}
// MaxEngineCall bounds a call's arguments: a path, an access and a few words.
const MaxEngineCall = 16 << 10
// AnswerEngineCall is the envelope one call is answered with: the verb's result, or its refusal in words.
func AnswerEngineCall(handle EngineFunc, verb, node, subject string, body []byte) []byte {
envelope := map[string]any{"node": node}
say := func() []byte {
raw, err := json.Marshal(envelope)
if err != nil {
raw, _ = json.Marshal(map[string]any{"node": node, "error": "the answer could not be written: " + err.Error()})
}
return raw
}
args := map[string]any{}
switch {
case handle == nil:
envelope["error"] = "this engine serves no verb of " + EngineSeat
return say()
case len(body) > MaxEngineCall:
envelope["error"] = fmt.Sprintf("the call is %d bytes, more than a call of %s takes", len(body), EngineSeat)
return say()
case len(strings.TrimSpace(string(body))) > 0:
if err := json.Unmarshal(body, &args); err != nil {
envelope["error"] = "the call's arguments are not an object: " + err.Error()
return say()
}
}
result, err := handle(EngineCall{Verb: verb, Caller: CallerOfSubject(subject, verb, node), Args: args})
if err != nil {
envelope["error"] = err.Error()
} else {
envelope["result"] = result
}
return say()
}
// EngineSeats is a link that serves this node's seat. Optional, as Asked is.
type EngineSeats interface {
ServeEngineSeat(EngineFunc)
}
// ServeEngineSeat answers every verb of the seat on this node's subjects, as the service node-engine, which says each
// verb's seat, tool and node to discovery (novox/hq ADR 0197). A bus whose user list is older than the grant refuses the
// subscriptions, said by the client; a caller then hears nobody answer.
func (l *natsLink) ServeEngineSeat(handle EngineFunc) {
if handle == nil || l.conn == nil {
return
}
svc, err := micro.AddService(l.conn, micro.Config{Name: EngineSeat, Version: "1.0.0",
Description: "the node seat node-engine of " + l.node + " (novox/hq ADR 0563)", QueueGroupDisabled: true})
if err != nil {
return
}
for _, v := range catalogue.EngineVerbs() {
verb := v.Name
answer := micro.HandlerFunc(func(r micro.Request) {
_ = r.Respond(AnswerEngineCall(handle, verb, l.node, r.Subject(), r.Data()))
})
meta := map[string]string{"seat": EngineSeat, "tool": verb, "node": l.node}
_ = svc.AddEndpoint(verb, answer, micro.WithEndpointSubject(EngineToolSubject(verb, l.node)),
micro.WithEndpointMetadata(meta))
_ = svc.AddEndpoint(verb+"-called", answer, micro.WithEndpointSubject(EngineCallPattern(verb, l.node)),
micro.WithEndpointMetadata(meta))
}
l.engine = svc
}
// SayEngine says one of the seat's events about this node, on the events stream.
func (l *natsLink) SayEngine(event string, body any) error {
raw, err := json.Marshal(body)
if err != nil {
return err
}
ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
defer cancel()
_, err = l.js.Publish(EngineEventSubject(event, l.node), raw, nats.Context(ctx))
return err
}
// The engine asks the operator under its own asker name (asks.EngineAsker) on the router's seat (ADR 0259 §3, as ADR
// 0563 §3 changes it): its ask and cancel on the seat's work stream, its warrants heard on its own decided subject, and
// its asks' state read from the router's records under its own name.
// RouterRecords is the router's bucket of asks, which each asker reads under its own name: the state `asks` of the
// operator-channel's holder, the messenger (mesh-catalog modules/messenger).
const RouterRecords = "messenger_asks"
// OperatorAsks is a link that asks the operator as this node's engine. Optional.
type OperatorAsks interface {
PublishAsk(subject string, body any, id string) error
RouterRecord(key string) (*stdio.StateEntry, error)
HearWarrants(asker string, heard func(asks.Warrant)) error
}
// PublishAsk publishes an ask or a cancel on the router's seat, once per id within the stream's duplicate window.
func (l *natsLink) PublishAsk(subject string, body any, id string) error {
raw, err := json.Marshal(body)
if err != nil {
return err
}
ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
defer cancel()
_, err = l.js.Publish(subject, raw, nats.MsgId(id), nats.Context(ctx))
return err
}
// RouterRecord reads one key of the router's records, `<asker>.<id>`; nil when it holds none.
func (l *natsLink) RouterRecord(key string) (*stdio.StateEntry, error) {
ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
defer cancel()
msg, err := l.conn.RequestWithContext(ctx, "$JS.API.DIRECT.GET.KV_"+RouterRecords+".$KV."+RouterRecords+"."+key, nil)
if err != nil {
return nil, err
}
if status := msg.Header.Get("Status"); status == "404" {
return nil, nil
} else if status != "" {
return nil, fmt.Errorf("the router's record %s could not be read: %s %s", key, status, msg.Header.Get("Description"))
}
return &stdio.StateEntry{Key: key, Value: msg.Data}, nil
}
// HearWarrants hears the warrants for asker on its decided subject, each handed to heard as it arrives.
func (l *natsLink) HearWarrants(asker string, heard func(asks.Warrant)) error {
sub, err := l.conn.Subscribe(asks.DecidedSubject(asker), func(msg *nats.Msg) {
var w asks.Warrant
if json.Unmarshal(msg.Data, &w) == nil {
heard(w)
}
})
if err != nil {
return err
}
l.warrants = sub
return nil
}
// errNoAsking is a link that cannot ask the operator.
var errNoAsking = errors.New("the link open now cannot ask the operator")
// discoveryQuiet is how long the engine waits after the last answer to the bus's discovery, and discoveryMost the
// longest it waits in all.
const (
discoveryQuiet = 300 * time.Millisecond
discoveryMost = 2 * time.Second
)
// HeardServed asks the bus's discovery (novox/hq ADR 0197) whether a seat's verb is answered on this machine: an
// endpoint naming the seat, the verb as its tool and this node.
func (l *natsLink) HeardServed(seat, verb string) (bool, error) {
inbox := l.conn.NewRespInbox()
sub, err := l.conn.SubscribeSync(inbox)
if err != nil {
return false, err
}
defer func() { _ = sub.Unsubscribe() }()
if err := l.conn.PublishRequest("$SRV.INFO", inbox, nil); err != nil {
return false, fmt.Errorf("asking the bus who serves what: %w", err)
}
heard := false
deadline := time.Now().Add(discoveryMost)
for time.Now().Before(deadline) {
msg, err := sub.NextMsg(discoveryQuiet)
if err != nil {
break
}
var info micro.Info
if json.Unmarshal(msg.Data, &info) != nil {
continue
}
for _, e := range info.Endpoints {
node := e.Metadata["node"]
if node == "" {
node = info.ID
}
if e.Metadata["seat"] == seat && e.Metadata["tool"] == verb && node == l.node {
heard = true
}
}
}
return heard, nil
}
// HeardServed is the link open now asking the bus's discovery whether a seat's verb is answered on this machine.
func (q *Session) HeardServed(seat, verb string) (bool, error) {
q.init()
q.mu.Lock()
bus := q.bus
q.mu.Unlock()
asker, ok := bus.(interface {
HeardServed(string, string) (bool, error)
})
if bus == nil || !ok {
return false, fmt.Errorf("%w: this machine has no link to the bus open now", ErrNotAsked)
}
return asker.HeardServed(seat, verb)
}
// ---- the engine's acts in the bus's own storage (ADR 0563 §2, §3) -------------------------------------------------
// actsStore is this node's keys of ActsBucket over the link open now: bound, never created (the grants part creates
// it, as the node-engine's declared state).
func (l *natsLink) actsStore() (nats.KeyValue, error) {
kv, err := l.js.KeyValue(ActsBucket)
if err != nil {
return nil, fmt.Errorf("the node-engine's state %s cannot be bound (it is created from the node-engine's "+
"manifest; asked again once it is): %w", ActsBucket, err)
}
return kv, nil
}
// BusState is the engine's ActStore: its own node's keys of ActsBucket, `<node>.<key>`, over whichever link the session
// has open; with none open every read and write is refused, and nothing is kept anywhere else.
type BusState struct {
Session *Session
Node string
}
func (b BusState) kv() (nats.KeyValue, error) {
b.Session.init()
b.Session.mu.Lock()
bus := b.Session.bus
b.Session.mu.Unlock()
l, ok := bus.(interface{ actsStore() (nats.KeyValue, error) })
if bus == nil || !ok {
return nil, fmt.Errorf("%w: this machine has no link to the bus open now, where its acts are kept", ErrNotAsked)
}
return l.actsStore()
}
func (b BusState) key(k string) string { return b.Node + "." + k }
// changed is a write that lost its compare-and-set, as stdio.ErrChanged.
func changed(err error) error {
var api *nats.APIError
if errors.Is(err, nats.ErrKeyExists) || (errors.As(err, &api) && api.ErrorCode == nats.JSErrCodeStreamWrongLastSequence) {
return stdio.ErrChanged
}
return err
}
func (b BusState) Get(key string) (*stdio.StateEntry, error) {
kv, err := b.kv()
if err != nil {
return nil, err
}
e, err := kv.Get(b.key(key))
if errors.Is(err, nats.ErrKeyNotFound) {
return nil, nil
}
if err != nil {
return nil, err
}
return &stdio.StateEntry{Key: key, Value: e.Value(), Revision: e.Revision()}, nil
}
func (b BusState) Create(key string, value any) (uint64, error) {
kv, err := b.kv()
if err != nil {
return 0, err
}
raw, err := json.Marshal(value)
if err != nil {
return 0, err
}
rev, err := kv.Create(b.key(key), raw)
return rev, changed(err)
}
func (b BusState) Update(key string, value any, revision uint64) (uint64, error) {
kv, err := b.kv()
if err != nil {
return 0, err
}
raw, err := json.Marshal(value)
if err != nil {
return 0, err
}
rev, err := kv.Update(b.key(key), raw, revision)
return rev, changed(err)
}
func (b BusState) Delete(key string) error {
kv, err := b.kv()
if err != nil {
return err
}
return kv.Delete(b.key(key))
}
// Keys are this node's keys under prefix, without the node: a watch of `<node>.>` alone, the one filter the node's
// grant names, its values left out.
func (b BusState) Keys(prefix string) ([]string, error) {
kv, err := b.kv()
if err != nil {
return nil, err
}
w, err := kv.Watch(b.Node+".>", nats.MetaOnly(), nats.IgnoreDeletes())
if err != nil {
return nil, err
}
defer func() { _ = w.Stop() }()
var out []string
for {
select {
case e := <-w.Updates():
if e == nil {
sort.Strings(out)
return out, nil
}
if k := strings.TrimPrefix(e.Key(), b.Node+"."); strings.HasPrefix(k, prefix) {
out = append(out, k)
}
case <-time.After(10 * time.Second):
return nil, errors.New("the node-engine's acts could not be listed within 10s")
}
}
}
+769
View File
@@ -0,0 +1,769 @@
package link
import (
"crypto/rand"
"crypto/sha256"
"encoding/hex"
"encoding/json"
"errors"
"fmt"
"sort"
"strconv"
"strings"
"sync"
"time"
stdio "git.novox.be/novox/mesh-sdk/go"
"git.novox.be/novox/mesh-sdk/go/asker"
"git.novox.be/novox/mesh-sdk/go/asks"
)
// The node seat node-engine's acts (novox/hq ADR 0563).
//
// A hand-over and an access act are the engine's own asks, performed only on the warrant (§3): a call performs nothing.
// The engine judges every value as it does for the controller's signed ask, then refuses in words or asks the operator
// under its own asker name (asks.EngineAsker) at the level approve, binding the digest of the verb, the node, every
// argument and the owner, group and mode it will set. A warrant is claimed once by compare-and-set in the engine's own
// state (asker.Client.Take), held to the ask it answers, and to the act bound and the disk as judged, judged again now;
// only then is the act performed and recorded with the warrant. A decline, an expiry or a changed disk says refused and
// changes nothing. One ask is open per path or access.
//
// The read-only verbs need no warrant (§2): judge-directories changes nothing; setuid-search is bounded here, one at a
// time and at most one started in an hour. Every act, with who asked as the runtime proved it, is kept in the engine's own
// declared state acts (ActsBucket, created by the grants part), under its own node's keys alone, read by the verb acts.
// The acts of the seat, as its verbs name them.
const (
ActHandOver = "hand-over"
ActAccess = "access-act"
ActJudgeDirectories = "judge-directories"
ActSetuidSearch = "setuid-search"
)
// The states of an act.
const (
HandAsked = "asked"
HandDone = "done"
HandRefused = "refused"
HandAnswered = "answered"
)
// SearchEvery is how often a setuid search may be started on a node (ADR 0563 §2): one in any hour.
const SearchEvery = time.Hour
// askMargin is how long past an ask's expiry the router is given to say it ended, before the engine closes it itself.
const askMargin = 5 * time.Minute
// ActsBucket is the node-engine's declared state acts (its manifest's state, ADR 0563 §2, §3), the bucket the grants
// part creates (mesh-sdk buscompose EngineActsBucket): keys `<node>.act.<id>`, `<node>.open.<act>.<digest>` and
// `<node>.ask.<id>`, each written by this node's engine alone.
const ActsBucket = "mesh-host_acts"
// Judged is what the engine judged of one act before it asks: what it is about, what it will set (the owner, group and
// mode, in words), and what the disk holds now (Disk), compared again when the warrant arrives.
type Judged struct {
// Subject is the path of a hand-over, or the access of an access act.
Subject string
// Path is the directory the act changes, as the engine states it.
Path string
// Sets is the owner, group and mode the act will set, in words; bound in the ask's digest.
Sets string
// Disk is what the engine read on disk when it judged: owner, group, mode and the file it is.
Disk string
// Words is what the operator is shown of it, without a path (ADR 0234 §6): the path is in the whole words alone.
Words string
}
// HandAct is one act asked through the seat, as acts reads it.
type HandAct struct {
ID string `json:"id"`
Act string `json:"act"`
Subject string `json:"subject,omitempty"`
Caller string `json:"caller"`
Why string `json:"why,omitempty"`
At time.Time `json:"at"`
State string `json:"state"`
Said string `json:"said,omitempty"`
Ended time.Time `json:"ended,omitempty"`
// Ask is the operator's ask of a hand-over or an access act; Sets and Disk what was judged when it was asked.
Ask string `json:"ask,omitempty"`
Sets string `json:"sets,omitempty"`
Disk string `json:"disk,omitempty"`
Path string `json:"path,omitempty"`
// Expires is when the ask ends unanswered.
Expires time.Time `json:"expires,omitempty"`
// By, Via, Proofs are the operator's warrant, recorded with the act (ADR 0563 §3).
By string `json:"by,omitempty"`
Via string `json:"via,omitempty"`
Proofs []string `json:"proofs,omitempty"`
// Auto says the engine asked it itself, on finding a library's root holding data outside its group (§4).
Auto bool `json:"auto,omitempty"`
}
// RootVerdict is the engine's answer to root-free (ADR 0563 §6), in the controller's words.
type RootVerdict struct {
Machine string `json:"machine"`
Free bool `json:"free"`
Why string `json:"why"`
Judged time.Time `json:"judged"`
}
// HandActs is the engine's side of the seat. Every func is the engine's own judgement or act; nil refuses.
type HandActs struct {
Node string
// State is where the acts and the asks are kept: this node's keys of ActsBucket (BusState), each key one act.
State ActStore
// Asker asks the operator as asks.EngineAsker(Node), with State as its store.
Asker *asker.Client
// JudgeHandOver judges a hand-over's path, refusing in words; HandOver records it on the warrant.
JudgeHandOver func(path string) (Judged, error)
HandOver HandOverFunc
// JudgeAccess judges an access act on an access, refusing in words; AccessAct runs it on the warrant.
JudgeAccess func(act, access string) (Judged, error)
AccessAct AccessActFunc
// JudgeDirectories, SetuidSearch and Searching are the read-only verbs' and whether a search runs now.
JudgeDirectories JudgeDirectoriesFunc
SetuidSearch SetuidSearchFunc
Searching func() bool
// RootFree is this machine's root-free verdict now.
RootFree func() RootVerdict
// Say says one of the seat's events, asked, done or refused; nil says nothing.
Say func(event string, body any)
Now func() time.Time
// Nonce keeps two asks of one second apart; nil is random.
Nonce func() string
mu sync.Mutex
}
func (h *HandActs) now() time.Time {
if h.Now != nil {
return h.Now()
}
return time.Now()
}
func (h *HandActs) nonce() string {
if h.Nonce != nil {
return h.Nonce()
}
var b [2]byte
_, _ = rand.Read(b[:])
return hex.EncodeToString(b[:])
}
func (h *HandActs) say(event string, a HandAct) {
if h.Say != nil {
h.Say(event, map[string]any{"act": a.Act, "subject": a.Subject, "caller": a.Caller, "ask": a.Ask,
"state": a.State, "said": a.Said, "node": h.Node})
}
}
// Call is the seat's one entry: a verb, the caller as the runtime proved it, and its arguments.
func (h *HandActs) Call(c EngineCall) (any, error) {
arg := func(k string) string { s, _ := c.Args[k].(string); return strings.TrimSpace(s) }
switch c.Verb {
case ActHandOver:
return h.AskHandOver(c.Caller, arg("path"), arg("why"), false)
case ActAccess:
return h.AskAccess(c.Caller, arg("act"), arg("access"), arg("why"), false)
case ActJudgeDirectories:
return h.Judge(c.Caller)
case ActSetuidSearch:
return h.Search(c.Caller, arg("why"))
case "root-free":
if h.RootFree == nil {
return nil, errors.New("this engine judges no machine root-free")
}
return map[string]any{"machines": []RootVerdict{h.RootFree()}}, nil
case "acts":
if id := arg("id"); id != "" {
return h.Act(id)
}
limit, _ := strconv.Atoi(arg("limit"))
if f, ok := c.Args["limit"].(float64); ok {
limit = int(f)
}
return h.Acts(limit)
}
return nil, fmt.Errorf("the node seat %s serves no verb %q", EngineSeat, c.Verb)
}
// ---- the record --------------------------------------------------------------------------------------------------
// ActStore is the engine's own state, this node's keys alone (BusState prefixes them with the node): each act one key,
// written by compare-and-set on its revision, an open ask one key made by a create.
type ActStore interface {
Get(key string) (*stdio.StateEntry, error)
Create(key string, value any) (uint64, error)
Update(key string, value any, revision uint64) (uint64, error)
Delete(key string) error
// Keys are the keys under prefix.
Keys(prefix string) ([]string, error)
}
func actKey(id string) string { return "act." + id }
// openKey is the one key an open ask of act on subject holds: made by a create, so a second ask of it is refused.
func openKey(act, subject string) string {
sum := sha256.Sum256([]byte(subject))
return "open." + act + "." + hex.EncodeToString(sum[:12])
}
// read is every act kept, oldest first.
func (h *HandActs) read() ([]HandAct, error) {
keys, err := h.State.Keys("act.")
if err != nil {
return nil, fmt.Errorf("the engine's acts cannot be listed: %w", err)
}
var out []HandAct
for _, k := range keys {
a, _, err := h.readOne(k)
if err != nil {
return nil, err
}
if a != nil {
out = append(out, *a)
}
}
sort.SliceStable(out, func(i, j int) bool { return out[i].At.Before(out[j].At) })
return out, nil
}
func (h *HandActs) readOne(key string) (*HandAct, uint64, error) {
e, err := h.State.Get(key)
if err != nil || e == nil {
return nil, 0, err
}
var a HandAct
if err := json.Unmarshal(e.Value, &a); err != nil {
return nil, 0, fmt.Errorf("the engine's act %s cannot be read: %w", key, err)
}
return &a, e.Revision, nil
}
// judgedKey keeps the node's judgements and refused searches, the last keptJudged of them: any caller may ask them, so
// they roll in one key and never add one (ADR 0563 §2: a wrong caller gains nothing, and fills no bucket).
const (
judgedKey = "judged"
keptJudged = 50
)
// roll adds a to the rolling key by compare-and-set, the oldest falling out.
func (h *HandActs) roll(a HandAct) error {
for range 5 {
e, err := h.State.Get(judgedKey)
if err != nil {
return err
}
var kept []HandAct
if e != nil {
if err := json.Unmarshal(e.Value, &kept); err != nil {
kept = nil // an unreadable roll is started again, never a reason to refuse
}
}
kept = append(kept, a)
if len(kept) > keptJudged {
kept = kept[len(kept)-keptJudged:]
}
if e == nil {
_, err = h.State.Create(judgedKey, kept)
} else {
_, err = h.State.Update(judgedKey, kept, e.Revision)
}
if !errors.Is(err, stdio.ErrChanged) {
return err
}
}
return errors.New("the engine's judgements kept changing under its write")
}
func (h *HandActs) rolled() ([]HandAct, error) {
e, err := h.State.Get(judgedKey)
if err != nil || e == nil {
return nil, err
}
var kept []HandAct
if err := json.Unmarshal(e.Value, &kept); err != nil {
return nil, fmt.Errorf("the engine's judgements cannot be read: %w", err)
}
return kept, nil
}
func (h *HandActs) record(a HandAct) error {
_, err := h.State.Create(actKey(a.ID), a)
return err
}
// end marks the open act of an ask ended, once, by compare-and-set on its revision, and frees its open key: false when
// it was not open.
func (h *HandActs) end(ask string, set func(*HandAct)) (HandAct, bool, error) {
for range 5 {
a, rev, err := h.readOne(actKey(ask))
if err != nil || a == nil || a.State != HandAsked {
return HandAct{}, false, err
}
set(a)
a.Ended = h.now().UTC()
if _, err := h.State.Update(actKey(ask), *a, rev); errors.Is(err, stdio.ErrChanged) {
continue
} else if err != nil {
return HandAct{}, false, err
}
_ = h.State.Delete(openKey(a.Act, a.Subject))
return *a, true, nil
}
return HandAct{}, false, errors.New("the engine's act " + ask + " kept changing under its write")
}
// Act is the verb acts with an id: that one act.
func (h *HandActs) Act(id string) (any, error) {
a, _, err := h.readOne(actKey(id))
if err != nil {
return nil, err
}
if a == nil {
return nil, fmt.Errorf("this engine keeps no act %q", id)
}
return map[string]any{"node": h.Node, "acts": []HandAct{*a}}, nil
}
// Acts is the verb acts: every act kept, newest first, at most limit when given.
func (h *HandActs) Acts(limit int) (any, error) {
acts, err := h.read()
if err != nil {
return nil, err
}
judged, err := h.rolled()
if err != nil {
return nil, err
}
acts = append(acts, judged...)
sort.SliceStable(acts, func(i, j int) bool { return acts[i].At.Before(acts[j].At) })
out := make([]HandAct, 0, len(acts))
for i := len(acts) - 1; i >= 0; i-- {
out = append(out, acts[i])
}
if limit > 0 && len(out) > limit {
out = out[:limit]
}
return map[string]any{"node": h.Node, "acts": out}, nil
}
// ---- the read-only verbs (§2) -------------------------------------------------------------------------------------
// Judge is judge-directories: the engine's judgement, read-only, recorded with its caller.
func (h *HandActs) Judge(caller string) (any, error) {
if h.JudgeDirectories == nil {
return nil, errors.New("this engine judges no directory")
}
said, err := h.JudgeDirectories(caller)
a := HandAct{ID: h.id(ActJudgeDirectories), Act: ActJudgeDirectories, Caller: caller, At: h.now().UTC(), State: HandAnswered}
if err != nil {
a.State, a.Said = HandRefused, err.Error()
}
if rerr := h.roll(a); rerr != nil && err == nil {
return nil, rerr
}
if err != nil {
return nil, err
}
return map[string]any{"node": h.Node, "said": said}, nil
}
// Search is setuid-search: refused while one runs and when one was started within the hour, naming when the next may
// start; otherwise the last search is thrown away and a full one starts, recorded with its caller.
func (h *HandActs) Search(caller, why string) (any, error) {
if h.SetuidSearch == nil {
return nil, errors.New("this engine keeps no search for setuid programs")
}
h.mu.Lock()
defer h.mu.Unlock()
now := h.now().UTC()
if h.Searching != nil && h.Searching() {
return nil, errors.New("a search for setuid programs runs on this machine now; no second is started while it does")
}
// The last start is a key of its own, claimed by compare-and-set before the search starts: no number of other acts,
// nor the record of acts, moves it (ADR 0563 §2).
if err := h.claimSearch(now, caller); err != nil {
return nil, err
}
said, err := h.SetuidSearch(caller)
a := HandAct{ID: h.id(ActSetuidSearch), Act: ActSetuidSearch, Caller: caller, Why: why, At: now, State: HandDone, Said: said}
if err != nil {
a.State, a.Said = HandRefused, err.Error()
}
// A search started is bounded to one an hour and kept as an act of its own; a refused one, which any caller may
// cause, rolls with the judgements.
record := h.record
if err != nil {
record = h.roll
}
if rerr := record(a); rerr != nil {
return nil, rerr
}
if err != nil {
return nil, err
}
return map[string]any{"node": h.Node, "said": said}, nil
}
// searchKey keeps the last search's start on this node, and libKey each library's own take ask: keys of their own, which
// no other act's key moves, kept for the bucket's age (ninety days, the node-engine's manifest).
const searchKey = "search.last"
func libKey(access string) string {
sum := sha256.Sum256([]byte(access))
return "lib." + hex.EncodeToString(sum[:12]) + ".asked"
}
// started is the last search's start as kept, and who asked it.
type started struct {
At time.Time `json:"at"`
By string `json:"by"`
}
// claimSearch refuses a search within SearchEvery of the last start, naming when the next may start, and otherwise
// claims this one by compare-and-set: a second claim at the same moment loses.
func (h *HandActs) claimSearch(now time.Time, caller string) error {
e, err := h.State.Get(searchKey)
if err != nil {
return err
}
var rev uint64
if e != nil {
var last started
if err := json.Unmarshal(e.Value, &last); err != nil {
return fmt.Errorf("the last search's start cannot be read, so none is started: %w", err)
}
if now.Sub(last.At) < SearchEvery {
return fmt.Errorf("a search was started at %s by %s, and one is started in any hour: the next may start "+
"at %s", last.At.Format(time.RFC3339), last.By, last.At.Add(SearchEvery).Format(time.RFC3339))
}
rev = e.Revision
}
claim := started{At: now, By: caller}
if rev == 0 {
_, err = h.State.Create(searchKey, claim)
} else {
_, err = h.State.Update(searchKey, claim, rev)
}
if errors.Is(err, stdio.ErrChanged) {
return errors.New("another search was started at this moment; no second is started")
}
return err
}
// ---- the asks (§3, §4) --------------------------------------------------------------------------------------------
func (h *HandActs) id(act string) string {
return act + "-" + h.now().UTC().Format("20060102t150405") + "-" + h.nonce()
}
// bound is the act an ask's yes binds: the verb, the node, every argument and what it will set (ADR 0563 §3).
func (h *HandActs) bound(a HandAct) asks.Act {
act := asks.Act{"verb": a.Act, "node": h.Node, "arg.subject": a.Subject, "arg.why": a.Why, "sets": a.Sets}
if a.Path != "" {
act["path"] = a.Path
}
return act
}
// AskHandOver is hand-over: judged, refused in words or asked of the operator. auto is the engine asking by itself.
func (h *HandActs) AskHandOver(caller, path, why string, auto bool) (any, error) {
if h.JudgeHandOver == nil || h.HandOver == nil {
return nil, errors.New("this engine takes no hand-over")
}
j, err := h.JudgeHandOver(path)
if err != nil {
return nil, err
}
return h.ask(ActHandOver, caller, why, auto, j)
}
// AskAccess is access-act: the act take on an access, judged on the library's home, refused or asked of the operator.
func (h *HandActs) AskAccess(caller, act, access, why string, auto bool) (any, error) {
if h.JudgeAccess == nil || h.AccessAct == nil {
return nil, errors.New("this engine runs no act on an accessed folder")
}
if act != ActTake {
return nil, fmt.Errorf("the act asked is %q, and the one act on an access is %q", act, ActTake)
}
j, err := h.JudgeAccess(act, access)
if err != nil {
return nil, err
}
return h.ask(ActAccess, caller, why, auto, j)
}
// AskedFor says whether an act on subject was ever asked by the engine itself: a library's take is asked by itself once
// (ADR 0563 §4), and again only by hand.
func (h *HandActs) AskedFor(act, subject string) bool {
if act == ActAccess {
e, err := h.State.Get(libKey(subject))
return err != nil || e != nil // unread is never a reason to ask again
}
acts, err := h.read()
if err != nil {
return true
}
for _, a := range acts {
if a.Act == act && a.Subject == subject && (a.Auto || a.State == HandAsked) {
return true
}
}
return false
}
func (h *HandActs) ask(act, caller, why string, auto bool, j Judged) (any, error) {
if h.Asker == nil {
return nil, errors.New("this engine reaches no operator to ask: nothing was asked and nothing changed")
}
h.mu.Lock()
defer h.mu.Unlock()
now := h.now().UTC()
a := HandAct{ID: h.id(act), Act: act, Subject: j.Subject, Path: j.Path, Caller: caller, Why: why, At: now,
State: HandAsked, Sets: j.Sets, Disk: j.Disk, Expires: now.Add(asks.ApproveLasts - time.Minute), Auto: auto}
a.Ask = a.ID
// One ask open per path or access: its open key made by a create, which a second ask cannot make.
if _, err := h.State.Create(openKey(act, j.Subject), map[string]string{"ask": a.Ask}); errors.Is(err, stdio.ErrChanged) {
held := "another"
if e, _ := h.State.Get(openKey(act, j.Subject)); e != nil {
var o struct {
Ask string `json:"ask"`
}
if json.Unmarshal(e.Value, &o) == nil && o.Ask != "" {
held = o.Ask
}
}
return nil, fmt.Errorf("%s of %s is asked already: the ask %s waits for the operator's answer", act, j.Subject, held)
} else if err != nil {
return nil, err
}
if err := h.record(a); err != nil {
_ = h.State.Delete(openKey(act, j.Subject))
return nil, err
}
bind, err := asks.ActDigest(h.bound(a))
if err != nil {
return nil, err
}
headline, does, stays := "Hand a directory on "+h.Node+" to the mesh?", "The mesh gives it "+j.Sets+".",
"The directory stays as it was found."
if act == ActAccess {
headline, does, stays = "Keep the library "+strings.TrimPrefix(j.Subject, "library:")+" on "+h.Node+"?",
"The library's group, group write and setgid are set on everything under its root.",
"The library waits and nothing under its root changes."
}
if len([]rune(headline)) > asks.HeadlineLength {
headline = string([]rune(headline)[:asks.HeadlineLength])
}
asked := asks.Ask{ID: a.Ask, Headline: headline, Who: asks.Operator,
Explanation: j.Words + " Your yes sets " + j.Sets + "; nothing changes until you answer.",
Whole: j.Words + " (" + j.Path + ") Your yes sets " + j.Sets + "; nothing changes until you answer.",
Details: "Asked by " + caller + " on " + h.Node + orWhy(why),
Options: []asks.Option{{ID: "yes", Label: "Yes", Does: does, Level: asks.Approve, Binds: bind},
{ID: "no", Label: "No, leave it", Does: stays, Level: asks.Acknowledge}},
Expires: a.Expires, OnExpiry: stays, About: EngineSeat + "-" + h.Node + "-" + act}
if err := h.Asker.Ask(asked); err != nil {
_, _, _ = h.end(a.Ask, func(x *HandAct) { x.State, x.Said = HandRefused, "the operator could not be asked: "+err.Error() })
return nil, fmt.Errorf("the operator could not be asked, and nothing changed: %w", err)
}
// Once asked, a library's own take ask is kept's own take ask: once, in a key of its own (ADR 0563 §4).
if auto && act == ActAccess {
_, _ = h.State.Create(libKey(j.Subject), map[string]any{"ask": a.Ask, "at": now})
}
h.say(HandAsked, a)
return map[string]any{"asked": a.Ask, "node": h.Node, "says": "the operator is asked on their channel; only " +
"their yes changes anything. acts shows where it stands."}, nil
}
func orWhy(why string) string {
if why == "" {
return ""
}
return ": " + why
}
// Warranted is the act on the operator's answer, handed once by asker.Client.Take after it held the warrant to the
// ask: a no is recorded refused; a yes is held to the act the option bound and the disk as judged, judged again now,
// and only then performed and recorded done with the warrant.
func (h *HandActs) Warranted(o asks.Option, w asks.Warrant) error {
a, _, err := h.readOne(actKey(w.Ask))
if err != nil {
return err
}
if a == nil || a.State != HandAsked {
return fmt.Errorf("the warrant answers the ask %s, which is not open here: nothing is done", w.Ask)
}
refuse := func(why string) error {
ended, ok, err := h.end(w.Ask, func(x *HandAct) { x.State, x.Said = HandRefused, why; warrantOn(x, w) })
if err == nil && ok {
h.say(HandRefused, ended)
}
return err
}
if o.ID != "yes" {
return refuse("the operator said no")
}
var j Judged
switch a.Act {
case ActHandOver:
j, err = h.JudgeHandOver(a.Subject)
case ActAccess:
j, err = h.JudgeAccess(ActTake, a.Subject)
default:
err = fmt.Errorf("the act %s is not performed on a warrant", a.Act)
}
if err != nil {
return refuse("judged again on the operator's yes, it is refused: " + err.Error())
}
now := *a
now.Sets, now.Path = j.Sets, j.Path
if err := o.Performs(h.bound(now)); err != nil {
return refuse("what the yes bound is not what would be done now: " + err.Error())
}
if j.Disk != a.Disk {
return refuse("the disk changed since the operator was asked (" + a.Disk + ", now " + j.Disk + "): ask again")
}
// Recorded with the way it came (ADR 0563 §5): the operator's yes to this engine's ask, not a terminal's line.
by := operatorOf(w) + ", to " + EngineSeat + "'s ask " + w.Ask
var said string
switch a.Act {
case ActHandOver:
said, err = h.HandOver(a.Subject, by)
case ActAccess:
said, err = h.AccessAct(AccessAct{Act: ActTake, Access: a.Subject, Path: j.Path, Warrant: w.Ask, By: by})
}
if err != nil {
return refuse(err.Error())
}
ended, ok, err := h.end(w.Ask, func(x *HandAct) { x.State, x.Said = HandDone, said; warrantOn(x, w) })
if err == nil && ok {
h.say(HandDone, ended)
}
return err
}
func warrantOn(a *HandAct, w asks.Warrant) {
a.Proofs = w.Proofs
if w.By != nil {
a.By, a.Via = w.By.Who, w.By.Kind
if w.By.Verified != "" {
a.Via += " (" + w.By.Verified + ")"
}
}
}
func operatorOf(w asks.Warrant) string {
if w.By == nil {
return "the operator"
}
return "the " + w.By.Who + " via " + w.By.Kind
}
// Heard is a warrant heard on the engine's decided subject: taken once (asker.Client.Take), and an ask that ended with
// no choice recorded refused.
func (h *HandActs) Heard(w asks.Warrant) {
if h.Asker == nil {
return
}
// A warrant the ask kept here does not hold to (another digest, another option, too late) ends nothing: it is no
// answer to the ask, which stays open for the operator's.
if handled, _ := h.Asker.Take(w, h.Warranted); handled == asker.Recorded {
why := "the ask ended " + string(w.Outcome)
if w.Words != "" {
why += ": " + w.Words
}
if ended, ok, _ := h.end(w.Ask, func(x *HandAct) { x.State, x.Said = HandRefused, why; warrantOn(x, w) }); ok {
h.say(HandRefused, ended)
}
}
}
// Reconcile catches up what the engine missed while it was away: a warrant the router holds for an open ask is taken,
// and an ask past its expiry and a margin is closed refused, expired.
func (h *HandActs) Reconcile() {
if h.Asker == nil {
return
}
acts, err := h.read()
if err != nil {
return
}
for _, a := range acts {
if a.State != HandAsked || a.Ask == "" {
continue
}
if _, w, err := h.Asker.Where(a.Ask); err == nil && w != nil {
h.Heard(*w)
continue
}
if !a.Expires.IsZero() && h.now().After(a.Expires.Add(askMargin)) {
if ended, ok, _ := h.end(a.Ask, func(x *HandAct) { x.State, x.Said = HandRefused, "the ask expired unanswered" }); ok {
h.say(HandRefused, ended)
}
}
}
}
// MemState is an ActStore in memory: an engine's acts while no bus is reached in a test.
type MemState struct {
mu sync.Mutex
keys map[string]stdio.StateEntry
top uint64
}
func (m *MemState) Get(key string) (*stdio.StateEntry, error) {
m.mu.Lock()
defer m.mu.Unlock()
if e, ok := m.keys[key]; ok {
return &e, nil
}
return nil, nil
}
func (m *MemState) Create(key string, value any) (uint64, error) { return m.put(key, value, 0, true) }
func (m *MemState) Update(key string, value any, revision uint64) (uint64, error) {
return m.put(key, value, revision, false)
}
func (m *MemState) put(key string, value any, revision uint64, create bool) (uint64, error) {
raw, err := json.Marshal(value)
if err != nil {
return 0, err
}
m.mu.Lock()
defer m.mu.Unlock()
was, kept := m.keys[key]
if (create && kept) || (!create && (!kept || was.Revision != revision)) {
return 0, stdio.ErrChanged
}
if m.keys == nil {
m.keys = map[string]stdio.StateEntry{}
}
m.top++
m.keys[key] = stdio.StateEntry{Key: key, Value: raw, Revision: m.top}
return m.top, nil
}
func (m *MemState) Delete(key string) error {
m.mu.Lock()
defer m.mu.Unlock()
delete(m.keys, key)
return nil
}
func (m *MemState) Keys(prefix string) ([]string, error) {
m.mu.Lock()
defer m.mu.Unlock()
var out []string
for k := range m.keys {
if strings.HasPrefix(k, prefix) {
out = append(out, k)
}
}
sort.Strings(out)
return out, nil
}
+15 -5
View File
@@ -76,9 +76,13 @@ const MaxHandOverBy = 200
// HandOverCheck is what an ask is held to: the mesh's key, this node's name, the clock, and the nonces taken.
type HandOverCheck struct {
Signer ed25519.PublicKey
Node string
Now func() time.Time
Nonces *Nonces
// SignerNow, when set, is the controller's key as the engine trusts it at this ask, read per ask in place of Signer:
// nil once the controller's lease is handed away, which refuses every ask signed by its key (novox/hq ADR 0563 §7,
// ADR 0550 §2).
SignerNow func() ed25519.PublicKey
Node string
Now func() time.Time
Nonces *Nonces
}
// Nonces are the nonces of asks taken, each kept until its ask expires: an ask is taken once. Kept in memory —
@@ -157,8 +161,14 @@ func believedBytesWithin(check HandOverCheck, context string, withPath bool, bod
if err := json.Unmarshal(body, &signed); err != nil {
return HandOverAsk{}, nil, "the ask is not a signed ask: " + err.Error()
}
if len(check.Signer) != ed25519.PublicKeySize ||
!ed25519.Verify(check.Signer, append([]byte(context), signed.Ask...), signed.Signature) {
signer := check.Signer
if check.SignerNow != nil {
if signer = check.SignerNow(); signer == nil {
return HandOverAsk{}, nil, "the controller's lease is handed away, and no ask signed by its key is taken"
}
}
if len(signer) != ed25519.PublicKeySize ||
!ed25519.Verify(signer, append([]byte(context), signed.Ask...), signed.Signature) {
return HandOverAsk{}, nil, "the ask is not signed by the mesh"
}
var ask HandOverAsk
+117 -2
View File
@@ -9,6 +9,9 @@ import (
"strings"
"testing"
"time"
stdio "git.novox.be/novox/mesh-sdk/go"
"github.com/nats-io/nats.go"
)
// The hand-over's ask and answer hold these field names; the controller's side holds the same list
@@ -145,6 +148,21 @@ func TestAHandOverAskIsBelievedOnlySignedForThisNodeUnexpiredAndOnce(t *testing.
if strings.Join(given, "; ") != "/srv/notes by jo through mesh-cli on anchor; /refused by jo through mesh-cli on anchor" {
t.Fatalf("the handler was given %v", given)
}
// ADR 0563 §7, ADR 0550 §2: after the lease write the engine answers no ask signed by the controller's key. The key
// is read per ask from what the engine trusts now; nil (the lease handed away) refuses, and nothing is handed over.
trusted := m.Signer
following := HandOverCheck{Signer: m.Signer, SignerNow: func() ed25519.PublicKey { return trusted }, Node: m.Node,
Now: func() time.Time { return now }, Nonces: &Nonces{}}
before := len(given)
trusted = nil
late := signedAsk(t, key, with(func(a *HandOverAsk) { a.Expires = now.Add(time.Minute); a.Nonce = strings.Repeat("cd", 16) }))
if a := AnswerHandOver(following, handle, late); !strings.Contains(a.Refused, "lease is handed away") || len(given) != before {
t.Fatalf("after the lease write an ask signed by the controller was answered %+v", a)
}
trusted = m.Signer
if a := AnswerHandOver(following, handle, late); a.Refused != "" || len(given) != before+1 {
t.Fatalf("while the controller holds the lease its ask was answered %+v", a)
}
// A nonce is forgotten once its ask has expired, so the set does not grow for ever.
now = now.Add(HandOverGood + time.Second)
check.Nonces.take("ff", now.Add(time.Second), now)
@@ -174,17 +192,23 @@ func TestAHandOverIsAnsweredOnlyWhenItCarriesAReply(t *testing.T) {
// handingLink is a quiet link that hears hand-overs, keeping the handler it was given.
type handingLink struct {
*quietLink
given chan HandOverFunc
given chan HandOverFunc
signer chan func() ed25519.PublicKey
}
func (l handingLink) HearHandOvers(h HandOverFunc) { l.given <- h }
// FollowSigner keeps what the link is told to verify the controller's asks with.
func (l handingLink) FollowSigner(f func() ed25519.PublicKey) { l.signer <- f }
// **The link is given the queue's hand-over** when it is served: what the controller asks reaches the engine's
// judgement, and nothing else.
func TestALinkThatHearsHandOversIsGivenTheQueues(t *testing.T) {
m, _ := aMember(t)
l := handingLink{quietLink: newQuietLink(), given: make(chan HandOverFunc, 1)}
l := handingLink{quietLink: newQuietLink(), given: make(chan HandOverFunc, 1), signer: make(chan func() ed25519.PublicKey, 1)}
q := &Session{Membership: m, Timeout: time.Second}
// The link verifies the controller's asks with the key the engine trusts now (ADR 0563 §7).
q.ControllerKey = func() ed25519.PublicKey { return nil }
q.HandOver = func(path, by string) (string, error) { return "the queue's: " + path + " by " + by, nil }
ctx, stop := context.WithCancel(context.Background())
served := make(chan error, 1)
@@ -197,8 +221,40 @@ func TestALinkThatHearsHandOversIsGivenTheQueues(t *testing.T) {
case <-time.After(5 * time.Second):
t.Fatal("the link was never given the queue's hand-over")
}
select {
case f := <-l.signer:
if f() != nil {
t.Fatal("the link was given another key than the queue's")
}
case <-time.After(5 * time.Second):
t.Fatal("the link was never given the key the engine trusts now")
}
stop()
<-served
// 774 (ADR 0563 §1): a call of the node seat reaches its verb with its caller as the subject proves it, unproven on
// the subject that names none; a verb the seat does not serve is refused, and the answer is a tool's envelope.
t.Run("a call of the node seat reaches its verb with its proven caller", func(t *testing.T) {
h := &HandActs{Node: "n1", State: &MemState{},
RootFree: func() RootVerdict { return RootVerdict{Machine: "n1", Free: true, Why: "judged"} },
JudgeDirectories: func(by string) (string, error) { return "judged for " + by, nil }}
for _, row := range []struct{ verb, subject, want string }{
{"judge-directories", "mesh.seat.node-engine.call.judge-directories.n1.shanks~node-tools", `"said":"judged for shanks.node-tools"`},
{"judge-directories", EngineToolSubject("judge-directories", "n1"), `"said":"judged for unproven"`},
{"judge-directories", "mesh.seat.node-engine.call.judge-directories.n2.shanks~node-tools", `"said":"judged for unproven"`},
{"root-free", "mesh.seat.node-engine.call.root-free.n1.ace~messenger", `"free":true`},
{"unserved", "mesh.seat.node-engine.call.unserved.n1.x", `"error":"the node seat node-engine serves no verb \"unserved\""`},
} {
if got := string(AnswerEngineCall(h.Call, row.verb, "n1", row.subject, []byte(`{}`))); !strings.Contains(got, row.want) ||
!strings.Contains(got, `"node":"n1"`) {
t.Errorf("%s on %s: %s", row.verb, row.subject, got)
}
}
acts, _ := h.Acts(0)
if got := acts.(map[string]any)["acts"].([]HandAct); len(got) != 3 || got[0].Caller != "unproven" ||
got[2].Caller != "shanks.node-tools" || got[2].Act != ActJudgeDirectories {
t.Errorf("the judgements are recorded as %+v", got)
}
})
}
// **Against a real server**: a signed request on the node's own subject is answered; one on another node's is
@@ -237,4 +293,63 @@ func TestNatsAHandOverIsAskedOfTheNode(t *testing.T) {
if len(handled) != 1 {
t.Fatalf("handled %d asks; want only the one with a reply", len(handled))
}
// 774 (ADR 0563 §2, §3): the engine's acts are its node's keys of its declared state on the bus, each written by
// compare-and-set: a second create and a stale revision lose, and another node's keys are not this node's.
t.Run("the acts are kept in the bus's own storage under this node's keys", func(t *testing.T) {
_ = js.DeleteKeyValue(ActsBucket)
if _, err := js.CreateKeyValue(&nats.KeyValueConfig{Bucket: ActsBucket}); err != nil {
t.Fatal(err)
}
q := &Session{Membership: m, Timeout: time.Second}
q.bus = l
mine, theirs := BusState{Session: q, Node: m.Node}, BusState{Session: q, Node: "another"}
rev, err := mine.Create("act.a-1", map[string]string{"state": "asked"})
if err != nil {
t.Fatal(err)
}
if _, err := mine.Create("act.a-1", map[string]string{}); !errors.Is(err, stdio.ErrChanged) {
t.Errorf("a second create: %v", err)
}
if _, err := mine.Update("act.a-1", map[string]string{"state": "done"}, rev+7); !errors.Is(err, stdio.ErrChanged) {
t.Errorf("an update at a stale revision: %v", err)
}
if _, err := mine.Update("act.a-1", map[string]string{"state": "done"}, rev); err != nil {
t.Errorf("an update at its revision: %v", err)
}
_, _ = theirs.Create("act.b-1", map[string]string{})
if keys, err := mine.Keys("act."); err != nil || len(keys) != 1 || keys[0] != "act.a-1" {
t.Errorf("this node's acts are %v, %v", keys, err)
}
if e, _ := js.KeyValue(ActsBucket); e != nil {
if got, err := e.Get(m.Node + ".act.a-1"); err != nil || !strings.Contains(string(got.Value()), "done") {
t.Errorf("the bucket holds %v, %v", got, err)
}
}
if err := mine.Delete("act.a-1"); err != nil {
t.Fatal(err)
}
if e, err := mine.Get("act.a-1"); err != nil || e != nil {
t.Errorf("a deleted act reads %v, %v", e, err)
}
})
// 774 (ADR 0563 §1): the node seat is served on this node's subjects alone, naming its caller, and said to discovery.
t.Run("the node seat is served on this node and discovered", func(t *testing.T) {
l.ServeEngineSeat(func(c EngineCall) (any, error) { return c.Verb + " for " + c.Caller, nil })
reply, err := conn.Request("mesh.seat.node-engine.call.acts."+m.Node+".shanks~node-tools", []byte(`{}`), 5*time.Second)
if err != nil || !strings.Contains(string(reply.Data), `"result":"acts for shanks.node-tools"`) {
t.Fatalf("the call answered %v, %v", reply, err)
}
if _, err := conn.Request(EngineToolSubject("acts", "another"), []byte(`{}`), 300*time.Millisecond); err == nil {
t.Fatal("another node's verb was answered here")
}
for seat, want := range map[string]bool{EngineSeat: true, "node-login-shell": false} {
verb := "acts"
if !want {
verb = "execute"
}
if heard, err := l.HeardServed(seat, verb); err != nil || heard != want {
t.Errorf("discovery hears %s.%s here: %v, %v", seat, verb, heard, err)
}
}
})
}
+3 -2
View File
@@ -274,7 +274,8 @@ func TestTheReportIsSentFirstAndTheNodeKeyCarriesTheNodesFacts(t *testing.T) {
// 1126 (ADR 0555 §3): with its keys, the node key carries the node's record: the profile's present capabilities and
// networks, the outward links, the health's contract and the public keys; a report that changes none of it writes
// nothing, a changed fact writes the key once, and a report without a profile or health leaves them standing.
keys := RecordKeys{TunnelKey: "tk", SealingKey: "sk", ServingKey: "pk", Certificate: "cert"}
keys := RecordKeys{TunnelKey: "tk", SealingKey: "sk", ServingKey: "pk", Certificate: "cert",
Trusts: []string{"controller key sha256:aa", "vault key sha256:bb"}} // and what it trusts (L3 s6, ADR 0550 §2)
q.Keys = func() RecordKeys { return keys }
rep.Profile = map[string]any{"capabilities": []any{map[string]any{"name": "systemd", "present": true},
map[string]any{"name": "docker", "present": false}, map[string]any{"name": "apt", "present": true}},
@@ -286,7 +287,7 @@ func TestTheReportIsSentFirstAndTheNodeKeyCarriesTheNodesFacts(t *testing.T) {
t.Fatalf("the node key holds no record: %+v, %v", got, err)
}
wantRecord := delivery.NodeRecord{V: delivery.Version, Node: "n1", TunnelKey: "tk", SealingKey: "sk", ServingKey: "pk",
Certificate: "cert", Capabilities: []string{"apt", "systemd"}, OutwardLinks: []string{"eth0"}, NetworksSaid: true,
Certificate: "cert", Trusts: keys.Trusts, Capabilities: []string{"apt", "systemd"}, OutwardLinks: []string{"eth0"}, NetworksSaid: true,
Networks: []catalogue.AttachedNetwork{{Link: "eth0", Network: "192.0.2.0/24"}}, Contract: NodeKeyContract}
if !delivery.SameNodeRecord(*got.Record, wantRecord) {
t.Errorf("record = %+v, want %+v", *got.Record, wantRecord)
+98
View File
@@ -3,9 +3,12 @@ package link
import (
"context"
"encoding/json"
"slices"
"strings"
"testing"
"time"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
// A health statement is said on the link open now, as the machine's own word on its own subject, and
@@ -50,6 +53,101 @@ func TestAHealthStatementIsSaidOnTheLinkOpenNow(t *testing.T) {
if HealthSubject("anchor") != "mesh.control.anchor.health" {
t.Fatalf("the subject %s is not inside the host's own grant", HealthSubject("anchor"))
}
// Issue 774: the node's own health report in its key of mesh-host_health, for the watchman's six alerts: written on
// a change only, the streaks and the time of the look not being one, and read first after a restart.
t.Run("the node's health report is written on a change only", func(t *testing.T) {
since := time.Date(2026, 10, 29, 20, 0, 0, 0, time.UTC)
look := func(streak int, writer string) Health {
return Health{Contract: 7, At: time.Now(), Network: &NetworkHealth{State: StateUnhealthy, Since: since, Parts: []NetworkPart{
{Part: "resolv-conf", State: StateUnhealthy, Writer: writer, Owner: "dnsmasq", Since: since, Streak: streak}}},
Units: &UnitsHealth{State: "degraded", Failed: []FailedUnit{{Unit: "cups.service", Scope: "system", Since: since}}},
Resources: []ResourceHealth{
{Module: "brother", Resource: "brother.scanner", Kind: KindAccount, Account: "jochen", State: StateUnhealthy,
Reason: "relogin needed: the session began before scanner", Since: since},
{Module: "photos", Resource: "photos.library", Kind: KindDirectory, Target: "/srv/photos", State: StateUnhealthy,
Reason: ReasonUsedAsFound + " owned by root", Since: since},
{Module: "letta", Resource: "letta.server", Kind: "container", State: StateHealthy, Since: since},
{Module: "claude-code", Resource: "claude-code.agent", Kind: KindAccount, Account: "agent", Root: "never",
State: StateHealthy, Since: since}}}
}
peers := []PeerSeen{{Key: "B=", Seen: true}, {Key: "A=", Addresses: []string{"10.10.0.2/32"}}}
r := NodeHealthOf("anchor", look(1, "NetworkManager"), peers, Machines{})
if r.Network.Parts[0].Writer != "NetworkManager" || r.Network.Parts[0].Streak != 0 || r.Peers[0].Key != "A=" ||
len(r.Units.Failed) != 1 || len(r.Relogin) != 1 || r.Relogin[0].Account != "jochen" ||
len(r.Found) != 1 || r.Found[0].Path != "/srv/photos" ||
len(r.Root) != 1 || r.Root[0].Account != "agent" || r.Root[0].State != StateHealthy || r.Contract != 7 {
t.Fatalf("the report is %+v", r)
}
// The machines an unhealthy part points at, as the controller holds it there; none for the machine's own.
machines := Machines{ByAddress: map[string]string{"10.10.0.1": "novox", "10.10.0.4": "ace", "10.10.0.9": "anchor"}, Hub: "novox"}
for _, tc := range []struct {
name string
toward []string
state string
want []string
}{
{"the tunnel to the hub", []string{"hub"}, StateUnhealthy, []string{"novox"}},
{"two mesh resolvers", []string{"10.10.0.1", "10.10.0.4"}, StateUnhealthy, []string{"novox", "ace"}},
{"a resolver a VPN client wrote in is the machine's own", []string{"10.10.0.1", "192.0.2.53"}, StateUnhealthy, nil},
{"its own resolver is its own", []string{"10.10.0.9"}, StateUnhealthy, nil},
{"a healthy part points at nothing", []string{"hub"}, StateHealthy, nil},
} {
h := Health{Network: &NetworkHealth{State: tc.state, Parts: []NetworkPart{{Part: "names", State: tc.state, Toward: tc.toward}}}}
got := NodeHealthOf("anchor", h, nil, machines).Network.Parts[0].Targets
if !slices.Equal(got, tc.want) {
t.Errorf("%s: targets %v, want %v", tc.name, got, tc.want)
}
}
store := &memHealth{}
var last *NodeHealth
clock := func() time.Time { return since }
for i, step := range []struct {
h NodeHealth
fresh bool // a restart: nothing written in this run yet
wrote bool
}{
{h: r, wrote: true},
{h: NodeHealthOf("anchor", look(2, "NetworkManager"), peers, Machines{})},
{h: NodeHealthOf("anchor", look(3, "NetworkManager"), peers, Machines{}), fresh: true},
{h: NodeHealthOf("anchor", look(3, "openvpn"), peers, Machines{}), wrote: true},
{h: NodeHealthOf("anchor", look(3, "openvpn"), []PeerSeen{{Key: "A=", Seen: true}}, Machines{}), wrote: true},
} {
if step.fresh {
last = nil
}
wrote, err := writeNodeHealth(t.Context(), store, &last, step.h, clock)
if err != nil || wrote != step.wrote {
t.Fatalf("step %d: wrote %v (want %v), %v", i, wrote, step.wrote, err)
}
}
got, err := delivery.DecodeNodeHealth(store.values["anchor"])
if err != nil || got.Node != "anchor" || got.Network.Parts[0].Writer != "openvpn" || !got.Peers[0].Seen || got.At.IsZero() {
t.Fatalf("the key holds %+v, %v", got, err)
}
// The key's bytes are mesh-sdk's layout: decoded and encoded again by it, the same bytes.
if again, err := delivery.EncodeNodeHealth(got); err != nil || string(again) != string(store.values["anchor"]) {
t.Fatalf("the key's bytes are not the sdk's layout: %s\n%s (%v)", store.values["anchor"], again, err)
}
if store.puts != 3 {
t.Fatalf("%d writes, want 3: one per change", store.puts)
}
})
}
// memHealth is a HealthStore in memory.
type memHealth struct {
values map[string][]byte
puts int
}
func (m *memHealth) Get(_ context.Context, key string) ([]byte, error) { return m.values[key], nil }
func (m *memHealth) Put(_ context.Context, key string, value []byte) error {
if m.values == nil {
m.values = map[string][]byte{}
}
m.values[key], m.puts = value, m.puts+1
return nil
}
// A declared tool check reads the node tools' reply envelope: healthy only when the tool says so (novox/hq
+48 -17
View File
@@ -6,9 +6,11 @@ import (
"errors"
"fmt"
"strings"
"sync"
"time"
"github.com/nats-io/nats.go"
"github.com/nats-io/nats.go/micro"
)
// The host's link on the bus: one connection, and what is asked of this node over it. The node hears no declaration
@@ -35,11 +37,14 @@ func natsURL(address string) string {
// natsLink is this node's connection as a JetStream subscription.
type natsLink struct {
conn *nats.Conn
js nats.JetStreamContext
asking *nats.Subscription
asked chan struct{}
handing *nats.Subscription
// mu guards signerNow (FollowSigner).
mu sync.Mutex
signerNow func() ed25519.PublicKey
conn *nats.Conn
js nats.JetStreamContext
asking *nats.Subscription
asked chan struct{}
handing *nats.Subscription
// searching hears the controller's terminal asking for a fresh setuid search (novox/hq issue 361).
searching *nats.Subscription
// judging hears the controller asking for a read-only judgement of this node's directories (issue 358).
@@ -48,7 +53,10 @@ type natsLink struct {
acting *nats.Subscription
// moving hears the controller asking to move a value sealed to this machine into the vault (issue 676).
moving *nats.Subscription
node string
// engine serves the node seat node-engine, and warrants hears the operator's answers to its asks (ADR 0563).
engine micro.Service
warrants *nats.Subscription
node string
// signer is the mesh's key, which a hand-over ask is verified with as a declaration is (issue 356).
signer ed25519.PublicKey
lost chan error
@@ -141,6 +149,21 @@ func (l *natsLink) hearAsks() {
}
}
// FollowSigner has every signed ask heard on this link verified with the key the engine trusts at that ask, the
// controller's while it holds its lease and none from its hand-over (novox/hq ADR 0563 §7, ADR 0550 §2).
func (l *natsLink) FollowSigner(f func() ed25519.PublicKey) {
l.mu.Lock()
l.signerNow = f
l.mu.Unlock()
}
// check is how a signed ask is believed on this link: the static membership signer until FollowSigner names the trust.
func (l *natsLink) check() HandOverCheck {
l.mu.Lock()
defer l.mu.Unlock()
return HandOverCheck{Signer: l.signer, SignerNow: l.signerNow, Node: l.node}
}
// HearHandOvers listens for the controller asking this node to hand a directory used as found to the mesh
// (novox/hq issue 356): a request on this machine's own subject, answered once on the reply it carried. The
// ask is verified with the mesh's key before anything is read out of it (AnswerHandOver), because a responder
@@ -151,7 +174,7 @@ func (l *natsLink) HearHandOvers(handle HandOverFunc) {
if handle == nil {
return
}
check := HandOverCheck{Signer: l.signer, Node: l.node}
check := l.check()
handing, err := l.conn.Subscribe(AskHandOverSubject(l.node), func(msg *nats.Msg) {
if body, answer := HandOverReply(check, handle, msg.Reply, msg.Data); answer {
_ = msg.Respond(body)
@@ -170,7 +193,7 @@ func (l *natsLink) HearSetuidSearches(handle SetuidSearchFunc) {
if handle == nil {
return
}
check := HandOverCheck{Signer: l.signer, Node: l.node}
check := l.check()
searching, err := l.conn.Subscribe(AskSetuidSearchSubject(l.node), func(msg *nats.Msg) {
if body, answer := SetuidSearchReply(check, handle, msg.Reply, msg.Data); answer {
_ = msg.Respond(body)
@@ -189,7 +212,7 @@ func (l *natsLink) HearDirectoryJudgements(handle JudgeDirectoriesFunc) {
if handle == nil {
return
}
check := HandOverCheck{Signer: l.signer, Node: l.node}
check := l.check()
judging, err := l.conn.Subscribe(AskJudgeDirectoriesSubject(l.node), func(msg *nats.Msg) {
if body, answer := JudgeDirectoriesReply(check, handle, msg.Reply, msg.Data); answer {
_ = msg.Respond(body)
@@ -207,7 +230,7 @@ func (l *natsLink) HearAccessActs(handle AccessActFunc) {
if handle == nil {
return
}
check := HandOverCheck{Signer: l.signer, Node: l.node}
check := l.check()
acting, err := l.conn.Subscribe(AskAccessActSubject(l.node), func(msg *nats.Msg) {
if body, answer := AccessActReply(check, handle, msg.Reply, msg.Data); answer {
_ = msg.Respond(body)
@@ -225,7 +248,7 @@ func (l *natsLink) HearMoveSecrets(handle MoveSecretFunc) {
if handle == nil {
return
}
check := HandOverCheck{Signer: l.signer, Node: l.node}
check := l.check()
moving, err := l.conn.Subscribe(AskMoveSecretSubject(l.node), func(msg *nats.Msg) {
if body, answer := MoveSecretReply(check, handle, msg.Reply, msg.Data); answer {
_ = msg.Respond(body)
@@ -258,6 +281,12 @@ func (l *natsLink) Close() {
if l.judging != nil {
_ = l.judging.Unsubscribe()
}
if l.engine != nil {
_ = l.engine.Stop()
}
if l.warrants != nil {
_ = l.warrants.Unsubscribe()
}
if l.conn != nil {
l.conn.Close()
}
@@ -301,10 +330,12 @@ func (l *natsLink) Ask(ctx context.Context, subject string, body []byte) ([]byte
// What the queue asks of a live link beside Bus, each by a type assertion that fails quietly: said here
// so a link that drops one does not build.
var (
_ Bus = (*natsLink)(nil)
_ HealthBus = (*natsLink)(nil)
_ HaveBus = (*natsLink)(nil)
_ ToolBus = (*natsLink)(nil)
_ Asker = (*natsLink)(nil)
_ Asked = (*natsLink)(nil)
_ Bus = (*natsLink)(nil)
_ HealthBus = (*natsLink)(nil)
_ HaveBus = (*natsLink)(nil)
_ ToolBus = (*natsLink)(nil)
_ Asker = (*natsLink)(nil)
_ Asked = (*natsLink)(nil)
_ EngineSeats = (*natsLink)(nil)
_ OperatorAsks = (*natsLink)(nil)
)
+8 -52
View File
@@ -313,58 +313,14 @@ const KindRunner = "runner"
// ReasonUsedAsFound starts the reason of a directory used as found.
const ReasonUsedAsFound = "used as found:"
// UnitsHealth is the machine's service managers in one statement (issue 315).
type UnitsHealth struct {
// State is running, degraded — a unit failed, the modules' or the machine's — or unknown when no
// manager could be read.
State string `json:"state"`
// Failed is every unit failed on two looks in a row that no module places: the machine's own.
Failed []FailedUnit `json:"failed"`
// Unread names a manager that could not be read, with why.
Unread []string `json:"unread,omitempty"`
// Away names an account's own manager not running, so not read: nothing in it is failed (issue 764).
Away []string `json:"away,omitempty"`
}
// FailedUnit is one failed unit no module places.
type FailedUnit struct {
Unit string `json:"unit"`
// Scope is system, or user: an account's own manager.
Scope string `json:"scope"`
// Load is loaded, not-found — a unit whose file is gone and whose failure its manager still holds…
Load string `json:"load,omitempty"`
// Result is how it failed: exit-code, timeout, start-limit-hit…
Result string `json:"result,omitempty"`
// Resource is the mesh's own resource that names it, when the mesh placed it in its own right.
Resource string `json:"resource,omitempty"`
Since time.Time `json:"since"`
}
// NetworkHealth is the machine's networking in one statement (ADR 0241): the worst of its parts, since
// when, and each part.
type NetworkHealth struct {
State string `json:"state"`
Since time.Time `json:"since"`
Parts []NetworkPart `json:"parts"`
}
// NetworkPart is one part of a machine's networking, as its engine judged it on its second look.
type NetworkPart struct {
// Part is resolv-conf, names, tunnel, bus or route.
Part string `json:"part"`
State string `json:"state"`
// Reason is why it is not healthy, in words with no address, path or domain; Said the detail.
Reason string `json:"reason,omitempty"`
Said string `json:"said,omitempty"`
// Writer is the program that rewrote the resolver file, when it can be named; Owner the module whose
// file it is — the uplink's holder.
Writer string `json:"writer,omitempty"`
Owner string `json:"owner,omitempty"`
// Toward is what the failure points at: "hub", or each resolver's address that failed.
Toward []string `json:"toward,omitempty"`
Since time.Time `json:"since"`
Streak int `json:"streak,omitempty"`
}
// The machine's networking and its service managers in a statement are the node's health report's parts (mesh-sdk
// delivery, issue 774): one layout for the message, the report and the watchman that reads it.
type (
NetworkHealth = delivery.NetworkHealth
NetworkPart = delivery.NetworkPart
UnitsHealth = delivery.UnitsHealth
FailedUnit = delivery.FailedUnit
)
// The states a long-running resource is said in (ADR 0240 §4).
const (
+208
View File
@@ -0,0 +1,208 @@
package link
import (
"context"
"errors"
"fmt"
"slices"
"sort"
"strings"
"time"
"git.novox.be/novox/mesh-sdk/go/declaration"
"git.novox.be/novox/mesh-sdk/go/delivery"
"github.com/nats-io/nats.go"
)
// The node-engine's own health report in the bus's own storage (issue 774, the operator's alert table, 2026-10-29): one
// key per node, `<node>`, of the engine's declared state `health` (HealthBucket, created by the grants part from the
// engine's manifest as its state `acts` is), written by this node's engine alone and only when what it says changed.
// The watchman (seat mesh-watchdogs) reads it for six alerts the controller raised from the passing health message:
//
// - machine-network: Network, the machine's route, tunnel, name servers, resolver file and bus, each a part;
// - network-rewritten: the resolver file's part, unhealthy with Writer, the program that rewrote it, and Owner, the
// module whose file it is;
// - network-unreachable: Peers, each peer of the mesh's tunnel and whether this machine saw it within five minutes,
// read across every node's report;
// - machine-units: Units, the failed system units no module owns;
// - relogin-needed: Relogin, each account whose running session lacks a group it was given;
// - directory-used-as-found: Found, each directory a module declares that the engine uses as found.
//
// The report's layout is mesh-sdk's (delivery.NodeHealth): the bucket, its version and its parts.
const (
HealthBucket = delivery.NodeHealthBucket
NodeHealthVersion = delivery.NodeHealthVersion
)
type (
NodeHealth = delivery.NodeHealth
PeerSeen = delivery.PeerSeen
NodeAccount = delivery.NodeAccount
NodeDirectory = delivery.NodeDirectory
)
// ReasonReloginNeeded starts an account's reason when its session needs a new login (accounts.ReasonRelogin).
const ReasonReloginNeeded = "relogin needed"
// Machines names the mesh's machines for a report's targets: each node by its address on the private network, and the
// hub's node. The zero value names none, and every part is then the machine's own.
type Machines struct {
ByAddress map[string]string
Hub string
}
// targets is the machines a part points at, as the controller holds it (mesh-controller machine_network.go): the hub's
// node for TowardHub, each resolver's node by its address; none when it points at nothing, or at this machine or an
// address no other mesh machine has, as one of them makes the part the machine's own.
func (m Machines) targets(node string, toward []string) []string {
if len(toward) == 0 {
return nil
}
var out []string
for _, t := range toward {
x := m.ByAddress[t]
if t == TowardHub {
x = m.Hub
}
if x == "" || x == node {
return nil
}
if !slices.Contains(out, x) {
out = append(out, x)
}
}
return out
}
// TowardHub is what a part failing toward the hub names in Toward (network.TowardHub).
const TowardHub = "hub"
// NodeHealthOf is the report a health statement and the tunnel's peers make: the statement's network without its
// streaks, each unhealthy part naming the machines it points at (machines), its machine units, and each account and
// directory of its resources that the six alerts read.
func NodeHealthOf(node string, h Health, peers []PeerSeen, machines Machines) NodeHealth {
out := NodeHealth{V: NodeHealthVersion, Node: node, Units: h.Units, Contract: h.Contract}
if len(peers) > 0 {
out.Peers = append([]PeerSeen(nil), peers...)
sort.Slice(out.Peers, func(i, k int) bool { return out.Peers[i].Key < out.Peers[k].Key })
}
if h.Network != nil {
n := NetworkHealth{State: h.Network.State, Since: h.Network.Since, Parts: make([]NetworkPart, 0, len(h.Network.Parts))}
for _, p := range h.Network.Parts {
p.Streak = 0 // a look's count moves every look; the report says a part's state, not how often it was seen
p.Targets = nil
if p.State == StateUnhealthy {
p.Targets = machines.targets(node, p.Toward)
}
n.Parts = append(n.Parts, p)
}
out.Network = &n
}
for _, r := range h.Resources {
// Each alert reads its own: an account may need a new login and be declared never to become root at once.
if r.Kind == KindAccount && r.State == StateUnhealthy && strings.HasPrefix(r.Reason, ReasonReloginNeeded) {
out.Relogin = append(out.Relogin, NodeAccount{Account: r.Account, Module: r.Module, Resource: r.Resource,
Reason: r.Reason, Since: r.Since})
}
if r.Kind == KindAccount && r.Root == declaration.RootNever {
// The account declared never to become root without a person, judged (ADR 0266): what DA and D-root read.
out.Root = append(out.Root, delivery.RootVerdict{Account: r.Account, Module: r.Module, Resource: r.Resource,
State: r.State, Reason: r.Reason, Since: r.Since})
}
if r.Kind == KindDirectory && strings.HasPrefix(r.Reason, ReasonUsedAsFound) {
out.Found = append(out.Found, NodeDirectory{Module: r.Module, Resource: r.Resource, Path: r.Target,
Reason: r.Reason, Since: r.Since})
}
}
sort.Slice(out.Relogin, func(i, k int) bool {
return out.Relogin[i].Account+out.Relogin[i].Module < out.Relogin[k].Account+out.Relogin[k].Module
})
sort.Slice(out.Found, func(i, k int) bool { return out.Found[i].Path < out.Found[k].Path })
sort.Slice(out.Root, func(i, k int) bool {
return out.Root[i].Module+out.Root[i].Resource < out.Root[k].Module+out.Root[k].Resource
})
return out
}
// HealthStore is where a node's report is kept: its own key of HealthBucket.
type HealthStore interface {
// Get is the key's value, nil when it has none.
Get(ctx context.Context, key string) ([]byte, error)
Put(ctx context.Context, key string, value []byte) error
}
// healthStore binds HealthBucket over the link open now: bound, never created.
func (l *natsLink) healthStore() (HealthStore, error) {
kv, err := l.js.KeyValue(HealthBucket)
if err != nil {
return nil, fmt.Errorf("the node-engine's state %s cannot be bound (it is created from the node-engine's "+
"manifest; written again once it is): %w", HealthBucket, err)
}
return natsHealth{kv}, nil
}
type natsHealth struct{ kv nats.KeyValue }
func (n natsHealth) Get(_ context.Context, key string) ([]byte, error) {
e, err := n.kv.Get(key)
if errors.Is(err, nats.ErrKeyNotFound) {
return nil, nil
}
if err != nil {
return nil, err
}
return e.Value(), nil
}
func (n natsHealth) Put(_ context.Context, key string, value []byte) error {
_, err := n.kv.Put(key, value)
return err
}
// WriteNodeHealth writes this node's report to its key when it says something other than the key holds, and says
// whether it wrote. The key is read before the first write of a run, so a restart that finds the same report writes
// nothing; a store that cannot be bound or written is an error, and the next look tries again.
func (q *Session) WriteNodeHealth(ctx context.Context, h NodeHealth) (bool, error) {
q.init()
q.mu.Lock()
bus := q.bus
q.mu.Unlock()
l, ok := bus.(interface{ healthStore() (HealthStore, error) })
if bus == nil || !ok {
return false, ErrNoLink
}
store, err := l.healthStore()
if err != nil {
return false, err
}
return writeNodeHealth(ctx, store, &q.healthWritten, h, time.Now)
}
// writeNodeHealth is WriteNodeHealth over a store: last is the report last written or read, nil before the first.
func writeNodeHealth(ctx context.Context, store HealthStore, last **NodeHealth, h NodeHealth, now func() time.Time) (bool, error) {
if *last == nil {
raw, err := store.Get(ctx, h.Node)
if err != nil {
return false, fmt.Errorf("the health report of %s cannot be read: %w", h.Node, err)
}
if raw != nil {
if was, err := delivery.DecodeNodeHealth(raw); err == nil {
*last = &was
}
}
}
if *last != nil && delivery.SameNodeHealth(**last, h) {
return false, nil
}
h.At = now().UTC()
raw, err := delivery.EncodeNodeHealth(h)
if err != nil {
return false, err
}
if err := store.Put(ctx, h.Node, raw); err != nil {
return false, fmt.Errorf("the health report of %s cannot be written: %w", h.Node, err)
}
*last = &h
return true, nil
}
+3 -1
View File
@@ -23,6 +23,8 @@ type RecordKeys struct {
Certificate string
// Authority is the mesh's own certificate as this machine last applied it (public).
Authority string
// Trusts is what this machine trusts a mesh signature from now (delivery.Trusts, ADR 0550 §2).
Trusts []string
}
// recordOf is the node's record from one facts report and its keys; nil for a report that carries no profile (one that
@@ -45,7 +47,7 @@ func recordOf(node string, r Report, keys RecordKeys) *delivery.NodeRecord {
return nil
}
rec := &delivery.NodeRecord{V: delivery.Version, Node: node, TunnelKey: keys.TunnelKey, SealingKey: keys.SealingKey,
ServingKey: keys.ServingKey, Certificate: keys.Certificate, Authority: keys.Authority}
ServingKey: keys.ServingKey, Certificate: keys.Certificate, Authority: keys.Authority, Trusts: keys.Trusts}
for _, c := range p.Capabilities {
if c.Present && c.Name != "" {
rec.Capabilities = append(rec.Capabilities, c.Name)
+21
View File
@@ -7,6 +7,8 @@ import (
"errors"
"fmt"
"time"
meshasks "git.novox.be/novox/mesh-sdk/go/asks"
)
// Membership is what a node needs to reach its mesh again, held by the caller.
@@ -185,6 +187,12 @@ func serve(ctx context.Context, link Link, m Membership, queue *Session, say Ann
}
// And the controller asking this node to hand a directory used as found to the mesh (novox/hq issue
// 356), answered on the link the question came over.
// Every signed ask is verified with the key trusted at that ask, none after the lease write (ADR 0563 §7).
if f, ok := link.(interface {
FollowSigner(func() ed25519.PublicKey)
}); ok && queue.ControllerKey != nil {
f.FollowSigner(queue.ControllerKey)
}
if h, ok := link.(HandOvers); ok {
h.HearHandOvers(queue.HandOver)
}
@@ -200,6 +208,19 @@ func serve(ctx context.Context, link Link, m Membership, queue *Session, say Ann
if s, ok := link.(AccessActs); ok {
s.HearAccessActs(queue.AccessAct)
}
// And its node seat, node-engine, with the operator's answers to its own asks (novox/hq ADR 0563): served on this
// link, and what was answered while no link was open caught up from the router's records.
if queue.Acts != nil {
if s, ok := link.(EngineSeats); ok {
s.ServeEngineSeat(queue.Acts.Call)
}
if a, ok := link.(OperatorAsks); ok {
if err := a.HearWarrants(meshasks.EngineAsker(m.Node), queue.Acts.Heard); err != nil {
say("cannot hear the operator's answers to this engine's asks: " + err.Error())
}
go queue.Acts.Reconcile()
}
}
// And asking to move a value sealed to this machine into the vault (design 57's V3, issue 676).
if s, ok := link.(MoveSecrets); ok {
s.HearMoveSecrets(queue.MoveSecret)
+57
View File
@@ -2,12 +2,14 @@ package link
import (
"context"
"crypto/ed25519"
"encoding/json"
"errors"
"fmt"
"sync"
"time"
stdio "git.novox.be/novox/mesh-sdk/go"
"git.novox.be/novox/mesh-sdk/go/health"
)
@@ -47,6 +49,12 @@ type Session struct {
// MoveSecret moves one value sealed to this machine into the vault, asked by the controller during the move of
// the secrets (design 57's V3, issue 676). Nil answers every such ask with a refusal.
MoveSecret MoveSecretFunc
// ControllerKey is the controller's key as the engine trusts it now: nil once its lease is handed away (novox/hq
// ADR 0563 §7). Each link verifies every signed ask with it, read per ask; nil ControllerKey keeps the membership's.
ControllerKey func() ed25519.PublicKey
// Acts is the node seat node-engine this engine holds (novox/hq ADR 0563): served on every link, and its asks to the
// operator asked and heard over it. Nil serves no seat.
Acts *HandActs
// Keys reads the node's public keys and its certificate for its record (ADR 0555 §3), at each facts report; nil writes
// no record.
Keys func() RecordKeys
@@ -78,6 +86,9 @@ type Session struct {
// haves is what this node-engine writes beside its facts (novox/hq ADR 0429 §4, issue 1132).
haves haveState
// healthWritten is this node's health report as last written or read (WriteNodeHealth): the health loop alone
// writes it, one look at a time.
healthWritten *NodeHealth
}
func (q *Session) init() {
@@ -337,3 +348,49 @@ func (q *Session) AskSeat(ctx context.Context, seat, verb string, args any, with
}
return env.Result, nil
}
// PublishAsk publishes an ask or a cancel to the operator's router over the link open now (novox/hq ADR 0563 §3).
func (q *Session) PublishAsk(subject string, body any, id string) error {
asking, err := q.operatorAsks()
if err != nil {
return err
}
return asking.PublishAsk(subject, body, id)
}
// RouterRecord reads one key of the router's records over the link open now.
func (q *Session) RouterRecord(key string) (*stdio.StateEntry, error) {
asking, err := q.operatorAsks()
if err != nil {
return nil, err
}
return asking.RouterRecord(key)
}
// SayEngine says one of the node seat's events over the link open now; with none open it is not said.
func (q *Session) SayEngine(event string, body any) {
q.init()
q.mu.Lock()
bus := q.bus
q.mu.Unlock()
if sayer, ok := bus.(interface{ SayEngine(string, any) error }); ok {
if err := sayer.SayEngine(event, body); err != nil {
q.Say("could not say the node seat's " + event + ": " + err.Error())
}
}
}
func (q *Session) operatorAsks() (OperatorAsks, error) {
q.init()
q.mu.Lock()
bus := q.bus
q.mu.Unlock()
if bus == nil {
return nil, fmt.Errorf("%w: this machine has no link to the bus open now", ErrNotAsked)
}
asking, ok := bus.(OperatorAsks)
if !ok {
return nil, errNoAsking
}
return asking, nil
}
+78
View File
@@ -57,4 +57,82 @@ func TestASetuidSearchAskIsBelievedOnlySignedForThisNodeAndOnce(t *testing.T) {
if a := AnswerSetuidSearch(check, nil, body); a.Refused == "" {
t.Fatalf("no handler: %+v", a)
}
// 774 (ADR 0563 §2): the node seat's setuid-search starts none while one runs, at most one in any hour, naming when the
// next may start, and records who asked.
t.Run("the node seat's search is bounded", func(t *testing.T) {
at, running, started := now, false, 0
h := &HandActs{Node: "n1", State: &MemState{}, Now: func() time.Time { return at },
Searching: func() bool { return running },
SetuidSearch: func(by string) (string, error) { started++; return "a new search starts", nil }}
for _, row := range []struct {
name string
after time.Duration
running bool
refused string
started int
}{
{name: "one runs", running: true, refused: "runs on this machine now", started: 0},
{name: "the first", started: 1},
{name: "within the hour", after: 59 * time.Minute, refused: "the next may start at 2026-10-10T03:00:00Z", started: 1},
{name: "after the hour", after: time.Hour + time.Second, started: 2},
} {
at, running = now.Add(row.after), row.running
_, err := h.Call(EngineCall{Verb: ActSetuidSearch, Caller: "ace.node-tools"})
if (row.refused == "") != (err == nil) || (err != nil && !strings.Contains(err.Error(), row.refused)) || started != row.started {
t.Errorf("%s: %v, %d started", row.name, err, started)
}
}
acts, _ := h.Acts(1)
if got := acts.(map[string]any)["acts"].([]HandAct); len(got) != 1 || got[0].Caller != "ace.node-tools" || got[0].State != HandDone {
t.Errorf("the last search is recorded as %+v", got)
}
// The bound is kept in a key of its own (rev-l3 round 1): two hundred judgements before a search, or the acts
// gone from the record, never let a second search start within the hour.
h.JudgeDirectories = func(string) (string, error) { return "judged", nil }
for range 200 {
if _, err := h.Call(EngineCall{Verb: ActJudgeDirectories, Caller: "ace.node-tools"}); err != nil {
t.Fatal(err)
}
}
at = now.Add(time.Hour + 2*time.Minute)
if _, err := h.Call(EngineCall{Verb: ActSetuidSearch, Caller: "ace.node-tools"}); err == nil || !strings.Contains(err.Error(), "the next may start") {
t.Errorf("a search after two hundred judgements: %v", err)
}
keys, _ := h.State.Keys("act.")
for _, k := range keys {
_ = h.State.Delete(k)
}
if _, err := h.Call(EngineCall{Verb: ActSetuidSearch, Caller: "ace.node-tools"}); err == nil || !strings.Contains(err.Error(), "the next may start") || started != 2 {
t.Errorf("a search with the acts gone: %v, %d started", err, started)
}
// rev-l3 round 2 (ADR 0563 §2, a wrong caller gains nothing): judgements and refused searches, which any caller
// may ask, keep one rolling key of the node's, so ten thousand of them write a bounded number of bytes and add no
// key to the bucket every node shares.
keysNow := func() (int, int) {
all, _ := h.State.Keys("")
size := 0
for _, k := range all {
if e, _ := h.State.Get(k); e != nil {
size += len(e.Value)
}
}
return len(all), size
}
_, _ = h.Call(EngineCall{Verb: ActJudgeDirectories, Caller: "unproven"})
keysBefore, _ := keysNow()
for i := range 10000 {
verb := ActJudgeDirectories
if i%10 == 0 {
verb = ActSetuidSearch
}
_, _ = h.Call(EngineCall{Verb: verb, Caller: "unproven"})
}
if n, size := keysNow(); n != keysBefore || size > 64<<10 {
t.Errorf("ten thousand judgements and refused searches left %d keys (was %d) of %d bytes", n, keysBefore, size)
}
listed, _ := h.Acts(0)
if got := listed.(map[string]any)["acts"].([]HandAct); len(got) == 0 || got[0].Caller != "unproven" {
t.Errorf("acts does not show the latest judgement first: %d", len(got))
}
})
}
+46 -7
View File
@@ -32,6 +32,7 @@ import (
"fmt"
"os"
"path/filepath"
"slices"
"sort"
"strconv"
"strings"
@@ -94,6 +95,18 @@ type Finding struct {
Writer string
// Toward is what the failure points at: TowardHub, or each resolver's address that failed.
Toward []string
// Peers is, for the tunnel, each peer of the mesh's interface and whether it was seen (Peer).
Peers []Peer
}
// Peer is one peer of the mesh's tunnel as this machine sees it: its public key, the addresses it is allowed, and
// whether it handshook within StaleHandshake. What the watchman reads to tell a machine nobody reaches from one whose
// tunnel alone is down (network-unreachable). Seen, not the handshake's time: the time moves every two minutes, the
// fact only when a peer comes or goes.
type Peer struct {
Key string `json:"key"`
Addresses []string `json:"addresses,omitempty"`
Seen bool `json:"seen"`
}
// Part is one part's state, as the statement says it.
@@ -115,6 +128,8 @@ type Statement struct {
Since time.Time `json:"since"`
At time.Time `json:"at"`
Parts []Part `json:"parts"`
// Peers are the tunnel's peers on the last look that read them, sorted by key; nil where the tunnel is not read.
Peers []Peer `json:"peers,omitempty"`
}
// Machine is what a look reads, replaced in tests.
@@ -266,6 +281,9 @@ func (j *Judge) Look(ctx context.Context) (Statement, bool) {
changed = true
}
st.State, st.Since = worst, j.overall.since
// The tunnel's peers as this look read them; a peer seen or lost is a change of its own (network-unreachable).
st.Peers = findings[PartTunnel].Peers
changed = changed || !samePeers(j.said.Peers, st.Peers)
j.said = st
return st, changed
}
@@ -435,14 +453,22 @@ func (j *Judge) lookTunnel(ctx context.Context, now time.Time) Finding {
}
}
hub := ""
allowed := map[string][]string{}
for _, line := range strings.Split(ips, "\n") {
f := strings.Fields(line)
for _, prefix := range f[min(1, len(f)):] {
allowed[f[0]] = append(allowed[f[0]], prefix)
if _, bits, ok := strings.Cut(prefix, "/"); ok && bits != "32" && bits != "128" {
hub = f[0]
}
}
}
peers := make([]Peer, 0, len(handshakes))
for key, at := range handshakes {
peers = append(peers, Peer{Key: key, Addresses: allowed[key], Seen: at > 0 && now.Sub(time.Unix(at, 0)) <= StaleHandshake})
}
sort.Slice(peers, func(i, k int) bool { return peers[i].Key < peers[k].Key })
found := func(f Finding) Finding { f.Peers = peers; return f }
age := func(at int64) string {
if at == 0 {
return "never"
@@ -452,23 +478,36 @@ func (j *Judge) lookTunnel(ctx context.Context, now time.Time) Finding {
if hub != "" {
at := handshakes[hub]
if at > 0 && now.Sub(time.Unix(at, 0)) <= StaleHandshake {
return Finding{OK: true}
return found(Finding{OK: true})
}
return Finding{Reason: "the tunnel to the hub has not handshaken for over five minutes",
Said: fmt.Sprintf("%s's newest handshake with the hub was %s", Interface, age(at)), Toward: []string{TowardHub}}
return found(Finding{Reason: "the tunnel to the hub has not handshaken for over five minutes",
Said: fmt.Sprintf("%s's newest handshake with the hub was %s", Interface, age(at)), Toward: []string{TowardHub}})
}
if len(handshakes) == 0 {
return Finding{OK: true}
return found(Finding{OK: true})
}
var newest int64
for _, at := range handshakes {
newest = max(newest, at)
}
if newest > 0 && now.Sub(time.Unix(newest, 0)) <= StaleHandshake {
return Finding{OK: true}
return found(Finding{OK: true})
}
return Finding{Reason: "no machine has handshaken with the hub's tunnel for over five minutes",
Said: fmt.Sprintf("%s's newest handshake with any of its %d peers was %s", Interface, len(handshakes), age(newest))}
return found(Finding{Reason: "no machine has handshaken with the hub's tunnel for over five minutes",
Said: fmt.Sprintf("%s's newest handshake with any of its %d peers was %s", Interface, len(handshakes), age(newest))})
}
// samePeers says two looks saw the same peers, each seen or not alike.
func samePeers(a, b []Peer) bool {
if len(a) != len(b) {
return false
}
for i := range a {
if a[i].Key != b[i].Key || a[i].Seen != b[i].Seen || !slices.Equal(a[i].Addresses, b[i].Addresses) {
return false
}
}
return true
}
func (j *Judge) lookBus() Finding {
+19 -1
View File
@@ -302,9 +302,27 @@ func TestOnTheHubAnyFreshPeerIsAHealthyTunnel(t *testing.T) {
f.hs = "A=\t" + itoa(f.now.Add(-time.Hour).Unix()) + "\nB=\t" + itoa(f.now.Unix()) + "\nC=\t0\n"
f.ips = "A=\t10.10.0.2/32\nB=\t10.10.0.3/32\nC=\t10.10.0.4/32\n"
j := f.judge(t)
if st := f.look(t, j); st.State != Healthy {
st := f.look(t, j)
if st.State != Healthy {
t.Fatalf("the hub with one fresh peer is %+v", st.Parts)
}
// Each peer and whether it was seen within five minutes, for the watchman's network-unreachable (issue 774); a
// peer seen again is a change, a handshake renewed alone is not.
want := []Peer{{Key: "A=", Addresses: []string{"10.10.0.2/32"}}, {Key: "B=", Addresses: []string{"10.10.0.3/32"}, Seen: true},
{Key: "C=", Addresses: []string{"10.10.0.4/32"}}}
if !samePeers(st.Peers, want) {
t.Fatalf("the peers are %+v", st.Peers)
}
f.hs = "A=\t" + itoa(f.now.Add(LookEvery).Unix()) + "\nB=\t" + itoa(f.now.Add(LookEvery).Unix()) + "\nC=\t0\n"
f.now = f.now.Add(LookEvery)
if st, changed := j.Look(t.Context()); !changed || !st.Peers[0].Seen {
t.Fatalf("a peer seen again is no change: changed %v, %+v", changed, st.Peers)
}
f.hs = "A=\t" + itoa(f.now.Add(LookEvery).Unix()) + "\nB=\t" + itoa(f.now.Add(LookEvery).Unix()) + "\nC=\t0\n"
f.now = f.now.Add(LookEvery)
if _, changed := j.Look(t.Context()); changed {
t.Fatal("handshakes renewed alone are a change")
}
}
func TestAnUnreadableTunnelIsSaidAndAMissingToolSkipsIt(t *testing.T) {
+116
View File
@@ -12,7 +12,9 @@ package selfcompose
import (
"context"
"crypto/ed25519"
"crypto/sha256"
"encoding/base64"
"encoding/hex"
"encoding/json"
"errors"
@@ -28,6 +30,7 @@ import (
"git.novox.be/novox/mesh-sdk/go/compose"
"git.novox.be/novox/mesh-sdk/go/compose/frombus"
"git.novox.be/novox/mesh-sdk/go/delivery"
"git.novox.be/novox/mesh-sdk/go/overlay"
"github.com/nats-io/nats.go"
@@ -75,6 +78,13 @@ type Env struct {
// Fetch reads an archive at a URL in the artifact store: a moved module's templates archive, whose texts its built
// manifest carries (issue 1126). Read once per digest. Nil reads none, and a moved module does not compose.
Fetch func(ctx context.Context, url string) ([]byte, error)
// VaultAuthority is the vault's authority as this node read it from the vault's `signers` (ADR 0550 §2), composed
// beside the controller's as the mesh's authority (frombus.SignersReader); "" or nil composes the controller's alone.
VaultAuthority func() string
// Certify asks the vault's `sign` for this node's own certificate, keeps it and writes the node's record again (ADR
// 0550 §2): asked only when a module placed here asks for a certificate and the record holds none from an authority
// the node trusts. Nil asks nothing, and such a module stands in, as before.
Certify func(ctx context.Context) error
}
// VaultSecretChanged is the vault's event that a secret changed: a name and its generation, never a value.
@@ -99,6 +109,8 @@ type Composer struct {
// sealedOwn is each own secret's sealed value as first handed, by module/name and generation: a value sealed again
// is other bytes, so the one first had is kept while the generation stands, and the composition stays the same.
sealedOwn map[string]ownSealed
// certified is the serving key a certificate was last asked for successfully, so it is asked once per key.
certified string
said string
refusal string
// held is what the composer's last pass held: each module refused alone, and the node's own ("node") when it could
@@ -411,6 +423,24 @@ func (c *Composer) takeModuleHave(u agent.Update) bool {
}
// inOwnOf is, of the modules p places, those the node runs inside its own composition (c.inOwn), sorted; c.mu held.
// Machines is, from the placements this composer holds, each node by its address on the private network, and the hub's
// node: what names the machine a failing network part points at (delivery.NetworkPart.Targets, issue 774). Empty
// before any placement is read.
func (c *Composer) Machines() (byAddress map[string]string, hub string) {
c.mu.Lock()
defer c.mu.Unlock()
byAddress = map[string]string{}
for node, p := range c.placements {
if a, _, _ := strings.Cut(p.Decided.Address, "/"); a != "" {
byAddress[a] = node
}
if p.Decided.Hub {
hub = node
}
}
return byAddress, hub
}
func (c *Composer) inOwnOf(p delivery.Placement) []string {
var out []string
for _, m := range p.Modules {
@@ -501,6 +531,8 @@ func (c *Composer) compose(ctx context.Context) []agent.Update {
if c.env.Running != nil {
running = c.env.Running()
}
// Its own certificate, asked of the vault when a module here asks for one and the record holds none it trusts.
c.certifyIfWanted(ctx, r, own)
scope := frombus.Scope{Node: c.env.Node, KeptAt: kept, InOwn: inOwn, Running: running}
missed := c.missed.Load()
compositions, _, held, err := frombus.Compose(ctx, r, c.env.Mesh, scope, fps)
@@ -743,9 +775,93 @@ func (c *Composer) Policy(ctx context.Context, module string) string {
return policy
}
// HoldsData says whether a module's manifest declares data (catalogue.Manifest.DataItems), as a composition's Data says
// it, and whether the manifest could be read: the agent's Data for what runs of a module from the kept declaration
// (agent.Env.HoldsData, issue 924).
func (c *Composer) HoldsData(ctx context.Context, module string) (data, known bool) {
c.mu.Lock()
m, ok := c.modules[module]
c.mu.Unlock()
if !ok || m.Manifest == "" {
return false, false
}
raw, err := reader{c: c}.Manifest(ctx, m.Manifest)
if err != nil {
return false, false
}
parsed, err := catalogue.ParseManifestForInputs(raw)
if err != nil {
return false, false
}
return len(parsed.DataItems()) > 0, true
}
// reader is the composer's view of the bus as frombus reads it.
type reader struct{ c *Composer }
// VaultAuthority is the vault's authority the node follows (frombus.SignersReader).
func (r reader) VaultAuthority(context.Context) string {
if r.c.env.VaultAuthority == nil {
return ""
}
return r.c.env.VaultAuthority()
}
var _ frombus.SignersReader = reader{}
// certificateWanted says whether a module the placement p puts on this node asks for a certificate for the machine's
// internal name (its manifest's `certificate`), read from the indexer's manifests; a manifest that cannot be read is
// taken as asking none, and the composition says what it misses.
func (c *Composer) certificateWanted(ctx context.Context, r reader, p delivery.Placement) bool {
for _, m := range p.Modules {
c.mu.Lock()
digest := c.modules[m.Module].Manifest
c.mu.Unlock()
if digest == "" {
continue
}
raw, err := r.Manifest(ctx, digest)
if err != nil {
continue
}
if parsed, err := catalogue.ParseManifestForInputs(raw); err == nil && parsed.Certificate != nil {
return true
}
}
return false
}
// certifyIfWanted asks Certify once per serving key when a module placed here asks for a certificate and this node's
// record holds none for its internal name and serving key from an authority it trusts (the one it holds, or the vault's).
func (c *Composer) certifyIfWanted(ctx context.Context, r reader, p delivery.Placement) {
if c.env.Certify == nil || !c.certificateWanted(ctx, r, p) {
return
}
c.mu.Lock()
rec, ok := c.records[c.env.Node]
asked := c.certified
c.mu.Unlock()
if !ok || rec.ServingKey == "" || asked == rec.ServingKey {
return
}
key, err := base64.StdEncoding.DecodeString(rec.ServingKey)
if err != nil || len(key) != ed25519.PublicKeySize {
return
}
trusted := []string{rec.Authority, r.VaultAuthority(ctx)}
if rec.Certificate != "" && delivery.CheckNodeCertificate(rec.Certificate, trusted, overlay.InternalName(c.env.Node),
ed25519.PublicKey(key), time.Now()) == nil {
return
}
if err := c.env.Certify(ctx); err != nil {
c.sayOnce(fmt.Sprintf("a module here asks for a certificate, and the vault's sign gave none: %v", err))
return
}
c.mu.Lock()
c.certified = rec.ServingKey
c.mu.Unlock()
}
func (r reader) Placements(context.Context) (map[string]delivery.Placement, error) {
r.c.mu.Lock()
defer r.c.mu.Unlock()
+36 -7
View File
@@ -3,6 +3,7 @@ package selfcompose
import (
"context"
"crypto/sha256"
"encoding/base64"
"encoding/hex"
"encoding/json"
"fmt"
@@ -315,9 +316,9 @@ func TestTheNodeComposesItsWantsFromTheBus(t *testing.T) {
}
// Issue 1126, the applying path (review of the seed): one module's own refusal holds that module alone. On novox,
// de-spiegel's unset route names (its manifest at the indexer's commit e25e1beb, captured) refused the whole node and
// marked all its modules not composed. Now lemurs is handed; de-spiegel is held, and proxy with it, which receives its
// routes: what was handed of de-spiegel stays handed, so the node keeps running it, and the agent is told those two
// alone, de-spiegel in its config form's words.
// marked all its modules not composed. Now lemurs and proxy are handed; de-spiegel is held alone (issue 1960: a hold
// never spreads to proxy, which composes without its routes, nor to its smtp provider): what was handed of de-spiegel
// stays handed, so the node keeps running it, and the agent is told it alone, in its config form's words.
spiegel, err := os.ReadFile("testdata/de-spiegel-e25e1beb.module.json")
if err != nil {
t.Fatal(err)
@@ -353,15 +354,17 @@ func TestTheNodeComposesItsWantsFromTheBus(t *testing.T) {
if u, ok := handed3[lemursKey3]; !ok || u.Deleted {
t.Errorf("lemurs, which composes, is not handed beside de-spiegel's refusal: %v (told %q)", handed3, *told)
}
if u, ok := handed3[proxyKey]; ok {
t.Errorf("proxy is handed without held de-spiegel's routes: %+v", u)
// Issue 1960: a hold never spreads. proxy is handed, composed without held de-spiegel's routes, which do not compose.
if u, ok := handed3[proxyKey]; !ok || u.Deleted {
t.Errorf("proxy is not handed beside held de-spiegel: %v (told %q)", handed3, *told)
}
if u, ok := handed3[spiegelKey]; ok {
t.Errorf("held de-spiegel is handed %+v, want what was handed of it kept", u)
}
if len(*told) != 1 || !strings.HasPrefix((*told)[0], "held de-spiegel: ") || !strings.Contains((*told)[0], "${config:praktijk-name}") ||
!strings.Contains((*told)[0], "held proxy: ") || strings.Contains((*told)[0], "held lemurs") {
t.Errorf("the agent was told %q, want de-spiegel held in its config form's words, and proxy", *told)
!strings.Contains((*told)[0], "without what it contributes to route") || strings.Contains((*told)[0], "held proxy") ||
strings.Contains((*told)[0], "held lemurs") {
t.Errorf("the agent was told %q, want de-spiegel held alone, in its config form's words, without its routes", *told)
}
// A read the bus did not answer, timed out or not, is tried again later; the run's own end and a refusal are not.
@@ -1219,6 +1222,32 @@ func TestTheNodeComposesItsWantsFromTheBus(t *testing.T) {
}
}
b.mu.Unlock()
// L3 s7 (ADR 0550 §2): the node asks the vault for its own certificate only when a module placed here asks for one
// and its record holds none from an authority it trusts; once per serving key.
certBus := newFakeBus()
certAsks := 0
certEnv := certBus.env()
certEnv.Certify = func(context.Context) error { certAsks++; return nil }
certEnv.VaultAuthority = func() string { return "" }
cc := New(certEnv)
withCert := `{"module":"web","version":"1","certificate":{"into":"/etc/web/tls.crt"}}`
cdigest := delivery.ManifestDigest([]byte(withCert))
cc.manifests[digest] = []byte(lemurs)
cc.manifests[cdigest] = []byte(withCert)
cc.modules = map[string]delivery.Module{"lemurs": {Manifest: digest}, "web": {Manifest: cdigest}}
cc.records["n1"] = delivery.NodeRecord{Node: "n1", ServingKey: base64.StdEncoding.EncodeToString(make([]byte, 32))}
onlyLemurs := delivery.Placement{Node: "n1", Modules: []delivery.PlacedModule{{Module: "lemurs"}}}
cc.certifyIfWanted(context.Background(), reader{c: cc}, onlyLemurs)
if certAsks != 0 {
t.Errorf("no module asks for a certificate, and the vault was asked %d times", certAsks)
}
withWeb := delivery.Placement{Node: "n1", Modules: []delivery.PlacedModule{{Module: "lemurs"}, {Module: "web"}}}
cc.certifyIfWanted(context.Background(), reader{c: cc}, withWeb)
cc.certifyIfWanted(context.Background(), reader{c: cc}, withWeb)
if certAsks != 1 {
t.Errorf("a module asks for a certificate: the vault was asked %d times, want once", certAsks)
}
}
// patternBus is a bus whose every watch gets the updates of its bucket matching its pattern, as the bus's own watches
+9
View File
@@ -17,6 +17,15 @@
"consumes": [
"restic.backed-up"
],
"state": [
{
"name": "acts",
"ttl-seconds": 7776000
},
{
"name": "health"
}
],
"config": {
"operator-signers": {
"type": "string",
+12
View File
@@ -68,4 +68,16 @@ func TestTheManifestConsumesTheBackupEvent(t *testing.T) {
if len(m.Consumes) != 1 || m.Consumes[0] != "restic.backed-up" {
t.Fatalf("consumes %v", m.Consumes)
}
// 774 (ADR 0563 §2, §3): the acts of the node seat and the engine's own asks are its declared state acts, the bucket
// mesh-host_acts the grants part creates and each node's engine writes under its own keys.
parsed, err := catalogue.ParseManifest(raw)
if err != nil {
t.Fatal(err)
}
// 774 (the alert table): beside it the state health, one key per node, the node's own health report, kept until
// written again (no TTL: a node that stops writing keeps its last word, which the watchman reads with its time).
if len(parsed.State) != 2 || parsed.State[0].Name != "acts" || parsed.State[0].TTLSeconds != 90*24*60*60 ||
parsed.State[1].Name != "health" || parsed.State[1].TTLSeconds != 0 {
t.Fatalf("the manifest declares the state %+v", parsed.State)
}
}