Files
mesh-sdk/go/compose/frombus/held_rows_test.go
T
jschoubben 7631a7d5de Hold a refused manifest in the bus reader instead of ending the whole read (ADR 0578 §3)
Issue 1585, S2: one refused manifest silenced disk-load for 45 hours. frombus decodes leniently, takes the
indexer's verdict, stands a refused module in as its stub and names it in Input.Refused; DataItemsHeld names it.

Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
2026-11-01 01:57:36 +01:00

299 lines
16 KiB
Go

package frombus
import (
"context"
"encoding/json"
"errors"
"fmt"
"strings"
"testing"
"git.novox.be/novox/mesh-sdk/go/catalogue"
"git.novox.be/novox/mesh-sdk/go/compose"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
// aNode is a bus with n1 the hub, each of manifests placed on it and held at the indexer, in the order given.
func aNode(t *testing.T, manifests ...string) (*bus, map[string]string) {
t.Helper()
p := delivery.Placement{Node: "n1", Decided: delivery.NodeDecided{Hub: true, Address: "10.77.0.1"}}
b := &bus{records: map[string]delivery.NodeRecord{"n1": {Node: "n1", TunnelKey: publicKey(t), SealingKey: "sk", Capabilities: []string{"systemd"}}},
modules: map[string]delivery.Module{}, manifests: map[string][]byte{}, builds: map[string]*delivery.BuilderBuild{}}
fps := map[string]string{}
for i, raw := range manifests {
module := strings.SplitN(strings.TrimPrefix(raw, `{"module":"`), `"`, 2)[0]
p.Modules = append(p.Modules, delivery.PlacedModule{Module: module})
b.modules[module] = delivery.Module{Manifest: delivery.ManifestDigest([]byte(raw))}
b.manifests[delivery.ManifestDigest([]byte(raw))] = []byte(raw)
fps[module] = "src2:" + strings.Repeat(string(rune('1'+i)), 64)
}
b.placements = map[string]delivery.Placement{"n1": p}
return b, fps
}
// Issue 774, 2026-10-30 19:30: a node refused whole after holding a module says what it held and why, since the last
// refusal can be of the held module's stub (node-tools held, then "its build produced 0 bundle(s)", its own why unsaid).
func aNodeRefusedWholeSaysWhatItHeld(t *testing.T) {
refused := `{"module":"power","version":"1","items":[{"collection":"contribution:node-bar:block","attributes":{"shows":"state"},"template":"files/a"}]}`
needs := `{"module":"lamp","version":"1","requires":["nothing-provides-this"]}`
b, fps := aNode(t, refused, needs)
refuse(b, "power", `"shows" is none of them`)
_, _, held, err := Compose(context.Background(), b, Mesh{}, Scope{Node: "n1"}, fps)
if err == nil || !strings.Contains(err.Error(), "nothing provides") || !strings.Contains(err.Error(), `held before: power: `) ||
!strings.Contains(err.Error(), `"shows" is none of them`) {
t.Errorf("the node refused whole with %v (held %v), want lamp's refusal and power held before, with its why", err, held)
}
}
// refuse writes the indexer's refusal of module's manifest into its modules value (novox/hq ADR 0578 §1).
func refuse(b *bus, module string, problems ...string) {
m := b.modules[module]
m.Verdict, m.Problems = delivery.VerdictRefused, problems
b.modules[module] = m
}
// ADR 0578 §3: the bus reader never ends a whole read on a refused manifest. Input, Plans and DataItems compose the rest
// and hold the refused module, named with the indexer's words in Refused; a field the decoder does not know is set aside
// and named, and its module composed.
func aRefusedManifestNeverEndsTheRead(t *testing.T) {
ok := `{"module":"lamp","version":"1","resources":[{"id":"d","type":"directory","path":"/srv/lamp","mode":"0700"}],` +
`"data":{"own":[{"id":"d","path":"${dir:d}","class":"valuable"}]}}`
bad := `{"module":"power","version":"1","resources":[{"id":"d","type":"directory","path":"/srv/power","mode":"0700"}],` +
`"data":{"own":[{"id":"d","path":"${dir:d}","class":"valuable"}]}}`
later := `{"module":"later","version":"1","a-field-from-later":1}`
b, _ := aNode(t, ok, bad, later)
refuse(b, "power", "removes a stream it may not")
ctx := context.Background()
in, err := Input(ctx, b, Mesh{}, Scope{Node: "n1"})
if err != nil {
t.Fatalf("one refused manifest ended the read: %v", err)
}
if why := in.Refused["power"]; !strings.Contains(why, "refused the manifest of power: removes a stream it may not") ||
len(in.Refused) != 1 {
t.Fatalf("refused: %v", in.Refused)
}
if aside := in.SetAside["later"]; !strings.Contains(aside, "a-field-from-later") {
t.Fatalf("set aside: %v", in.SetAside)
}
if !in.Catalogue["power"].Stubbed || in.Catalogue["lamp"].Stubbed || in.Catalogue["later"].Module != "later" {
t.Fatalf("the catalogue: %+v", in.Catalogue)
}
if _, err := Plans(ctx, b, Mesh{}, Scope{Node: "n1"}); err != nil {
t.Fatalf("plans: %v", err)
}
items, _, held, err := DataItemsHeld(ctx, b, Mesh{}, Scope{Node: "n1"})
if err != nil || len(items["lamp"]) != 1 || items["power"] != nil || !strings.Contains(held["power"], "removes a stream") {
t.Fatalf("data items %v held %v: %v", items, held, err)
}
}
// Issue 774: the culprit rule takes a module whose stub composes or fails otherwise; the runtime only where its stub
// composes, since its stub always fails otherwise.
func theRuntimeIsBlamedOnlyWhereItsStubComposes(t *testing.T) {
err := errors.New("lamp refuses")
for _, row := range []struct {
module string
again error
blamed bool
}{
{"lamp", nil, true},
{"lamp", errors.New("other"), true},
{"lamp", errors.New("lamp refuses"), false},
{catalogue.RuntimeModule, nil, true},
{catalogue.RuntimeModule, errors.New("node-tools is assigned to n1 and its build produced 0 bundle(s)"), false},
{catalogue.RuntimeModule, fmt.Errorf("compose: %w", &catalogue.NoRuntimeToRunError{Held: true}), true},
{catalogue.RuntimeModule, &catalogue.NoRuntimeToRunError{}, false},
} {
if got := blames(row.module, row.again, err); got != row.blamed {
t.Errorf("%s with %v: blamed %v, want %v", row.module, row.again, got, row.blamed)
}
}
}
// Issue 774, 2026-10-30 19:30: the trunk's node-tools moved to a build not yet done, and ace, g14 and shanks composed
// it from the indexer's manifest, which names no bundle: "its build produced 0 bundle(s)", every module refused. A
// module that runs is now held until its build lands, the runtime composed as what runs of it; and a module wanted at
// another build than the trunk's is composed from the build its want names.
func aRunningModuleWaitsForItsBuild(t *testing.T) {
dg := func(c string) string { return "sha256:" + strings.Repeat(c, 64) }
tools := `{"module":"node-tools","version":"1","own-secrets":{"broker":"/var/lib/mesh/node-tools/broker"},` +
`"build":{"artifacts":[{"name":"runtime","kind":"bundle","language":"go","system":"arch","from":"cmd/node-tools"}]}}`
lamp := `{"module":"lamp","version":"1","resources":[{"id":"service","type":"service","unit":"lamp.service","state":"running","boot":"enabled"}]}`
store := `{"module":"artifacts","version":"1","provides":[{"name":"artifact-store","scope":"mesh"}],"serves":{"artifact-store":{"port":5100}},` +
`"receives":{"artifact-store":"/var/lib/artifacts/received.json"}}`
old, next := "src2:"+strings.Repeat("a", 64), "src2:"+strings.Repeat("b", 64)
running := []delivery.Composed{{Node: "n1", Module: catalogue.RuntimeModule, FP: old, Resources: []json.RawMessage{
json.RawMessage(`{"id":"node-tools.runtime","module":"node-tools","type":"process","name":"node-tools","env":{"MESH_TOOL_MODULES":""}}`),
json.RawMessage(`{"id":"node-tools.bundle-runtime","module":"node-tools","type":"archive","source":"artifact-store://node-tools/runtime/blobs/` +
dg("1") + `","digest":"` + dg("1") + `","path":"` + catalogue.BundlePath("node-tools", "runtime") + `"}`)}}}
for _, row := range []struct {
name string
want string // the build node-tools' want names
done bool // the trunk's build is done
held string // words node-tools' hold says; empty: composed
digest string // the runtime bundle's digest composed
}{
{name: "the trunk's build not done yet", want: next, held: "no done build yet"},
{name: "wanted at the build that runs, the trunk's not done", want: old, digest: dg("1")},
{name: "the trunk's build done", want: next, done: true, digest: dg("2")},
} {
b, fps := aNode(t, store, tools, lamp)
b.held = true // every grant held, so a contribution's credential stands in for nothing
b.own = []compose.OwnSecretRow{{Node: "n1", Module: "node-tools", Name: "broker", Origin: "made", SealedTo: "sk", Sealed: `"box"`}}
mod := b.modules["node-tools"]
mod.FP2 = next
b.modules["node-tools"] = mod
b.builds["node-tools "+old] = &delivery.BuilderBuild{State: delivery.BuildDone, Manifest: mod.Manifest,
Artifacts: []delivery.BuiltArtifact{{Name: "runtime", Kind: "bundle", Reference: "artifact-store://node-tools/runtime/blobs/" + dg("1"), Digest: dg("1")}}}
b.builds["node-tools "+next] = &delivery.BuilderBuild{State: "building", Manifest: mod.Manifest}
if row.done {
b.builds["node-tools "+next] = &delivery.BuilderBuild{State: delivery.BuildDone, Manifest: mod.Manifest,
Artifacts: []delivery.BuiltArtifact{{Name: "runtime", Kind: "bundle", Reference: "artifact-store://node-tools/runtime/blobs/" + dg("2"), Digest: dg("2")}}}
}
fps["node-tools"] = row.want
out, _, held, err := Compose(context.Background(), b, Mesh{}, Scope{Node: "n1", Running: running}, fps)
if err != nil {
t.Errorf("%s: the node is refused: %v", row.name, err)
continue
}
if lampKey, _ := delivery.ComposedKey("n1", "lamp", fps["lamp"]); out[lampKey].Module != "lamp" {
t.Errorf("%s: lamp is not composed (held %v)", row.name, held)
}
if why := held["node-tools"]; (row.held == "") != (why == "") || !strings.Contains(why, row.held) {
t.Errorf("%s: node-tools held %q, want %q", row.name, why, row.held)
}
if row.digest != "" {
key, _ := delivery.ComposedKey("n1", "node-tools", row.want)
if c := out[key]; !strings.Contains(fmt.Sprint(rawStrings(c.Resources)), row.digest) {
t.Errorf("%s: node-tools is composed as %v, want its bundle %s", row.name, rawStrings(c.Resources), row.digest)
}
}
}
}
func rawStrings(rs []json.RawMessage) []string {
var out []string
for _, r := range rs {
out = append(out, string(r))
}
return out
}
// Issue 774: a refusal of a held module's stand-in leads with why the module was held, its first reason, as mesh-self-check's
// on ace should have; a refusal naming no held module is as it was.
func aHeldModulesFirstReasonLeads(t *testing.T) {
held := map[string]string{"mesh-self-check": "mesh-self-check on ace has no bus credential: nothing was issued for ace.mesh-self-check"}
for _, row := range []struct {
err, want string
}{
{"mesh-self-check runs as its own account and declares no own secret broker",
"mesh-self-check is held: mesh-self-check on ace has no bus credential: nothing was issued for ace.mesh-self-check; " +
"its stand-in is then refused: mesh-self-check runs as its own account and declares no own secret broker"},
{"ace is the hub and has no endpoint", "ace is the hub and has no endpoint"},
} {
err := errors.New(row.err)
got := firstReason(err, held)
if got.Error() != row.want || !errors.Is(got, err) {
t.Errorf("refused %q, want %q", got, row.want)
}
}
}
// Issue 774, 2026-10-30 21:04: mesh-conditions and mesh-healers on novox and mesh-self-check on ace were placed before their
// bus accounts were issued. Held for "needs a secret called broker and none was made", each one's bare stub was then
// refused as "runs as its own account and declares no own secret broker", and that refusal failed every module of the
// node (79 on novox, 74 on ace). A held module's stand-in is never refused by a rule about its own runtime or its own
// secrets: the node composes every other module, and the held one keeps the words it was first held for.
func aHeldModuleOfItsOwnAccountNeverRefusesTheNode(t *testing.T) {
dg := "sha256:" + strings.Repeat("1", 64)
tools := `{"module":"node-tools","version":"1","own-secrets":{"broker":"/var/lib/mesh/node-tools/broker"},` +
`"build":{"artifacts":[{"name":"runtime","kind":"bundle","language":"go","system":"arch","from":"cmd/node-tools"}]}}`
own := func(name string) string {
return `{"module":"` + name + `","version":"1","runs-as":"` + name + `","secrets-owner":"` + name + `","tools":["status"],` +
`"own-secrets":{"broker":"/var/lib/` + name + `/broker"},"contributes":{"artifact-store":{"name":"` + name + `"}},` +
`"resources":[{"id":"account","type":"user","name":"` + name + `"}],` +
`"build":{"artifacts":[{"name":"tools","kind":"bundle","language":"typescript","entrypoints":["tools/index.js"]}]}}`
}
lamp := `{"module":"lamp","version":"1","resources":[{"id":"service","type":"service","unit":"lamp.service","state":"running","boot":"enabled"}]}`
store := `{"module":"artifacts","version":"1","provides":[{"name":"artifact-store","scope":"mesh"}],"serves":{"artifact-store":{"port":5100}},` +
`"receives":{"artifact-store":"/var/lib/artifacts/received.json"}}`
for _, row := range []struct {
name string
issued []string // the own-account modules whose broker is issued
building bool // node-tools' build is not done, and nothing of it runs
held map[string]string // the modules held, each with words of the reason it was first held for
// ungranted gives the provider a grants file and no consumer's grant is issued yet (a newly placed module's
// usual state): a held module's contribution then stands in for its grant, and is left out (review round 1).
ungranted bool
}{
{name: "none issued", held: map[string]string{"check": "broker", "heal": "broker"}},
{name: "none issued, no grant issued", ungranted: true, held: map[string]string{"check": "broker", "heal": "broker"}},
{name: "one issued", issued: []string{"heal"}, held: map[string]string{"check": "broker"}},
{name: "both issued", issued: []string{"check", "heal"}},
// The runtime held: its stand-in is not refused for its bundle; the modules it would run are held, not the node.
{name: "the runtime's build not done", issued: []string{"check", "heal"}, building: true,
held: map[string]string{"node-tools": "its build produced 0 bundle(s)", "check": "not here to run it from", "heal": "not here to run it from"}},
} {
provider := store
if row.ungranted {
provider = strings.Replace(store, `"receives":`, `"grants":{"artifact-store":"/var/lib/artifacts/grants"},"receives":`, 1)
}
b, fps := aNode(t, provider, tools, own("check"), lamp, own("heal"))
// Every grant held, so the held modules' contributions are composed whole; the row ungranted issues none.
b.held = !row.ungranted
b.own = []compose.OwnSecretRow{{Node: "n1", Module: "node-tools", Name: "broker", Origin: "made", SealedTo: "sk", Sealed: `"box"`}}
for _, m := range row.issued {
b.own = append(b.own, compose.OwnSecretRow{Node: "n1", Module: m, Name: "broker", Origin: "made", SealedTo: "sk", Sealed: `"box"`})
}
for _, m := range []string{"node-tools", "check", "heal"} {
name := "tools"
if m == "node-tools" {
name = "runtime"
}
mod := b.modules[m]
mod.FP2 = fps[m]
b.modules[m] = mod
b.builds[m+" "+fps[m]] = &delivery.BuilderBuild{State: delivery.BuildDone, Manifest: mod.Manifest,
Artifacts: []delivery.BuiltArtifact{{Name: name, Kind: "bundle", Reference: "artifact-store://" + m + "/" + name + "/blobs/" + dg, Digest: dg}}}
if m == "node-tools" && row.building {
b.builds[m+" "+fps[m]] = &delivery.BuilderBuild{State: "building", Manifest: mod.Manifest}
}
}
out, _, held, err := Compose(context.Background(), b, Mesh{}, Scope{Node: "n1"}, fps)
if err != nil {
t.Errorf("%s: the node is refused: %v", row.name, err)
continue
}
// A held module keeps what it contributes (issue 774, 2026-10-31: nextcloud held for its admin secret dropped out
// of postgres's received file, and postgres's helper would have locked its login): only its own secrets are left
// out of its stand-in, never its contributions, so the provider's received file names every consumer.
storeKey, _ := delivery.ComposedKey("n1", "artifacts", fps["artifacts"])
// And never a stand-in for a grant not yet issued: the provider is applied, so a held module whose grant is
// not issued is composed without that contribution, never with "foreseen: made by the next send".
for key, c := range out {
if got, _ := json.Marshal(c.Resources); strings.Contains(string(got), "foreseen") {
t.Errorf("%s: %s applies a stand-in: %s", row.name, key, got)
}
}
for _, m := range []string{"check", "heal"} {
if row.ungranted {
break
}
if got, _ := json.Marshal(out[storeKey].Resources); !strings.Contains(string(got), `\"from\": \"`+m+`\"`) {
t.Errorf("%s: the provider's received file drops %s's contribution: %s", row.name, m, got)
}
}
for _, m := range []string{"node-tools", "check", "lamp", "heal"} {
key, _ := delivery.ComposedKey("n1", m, fps[m])
_, composed := out[key]
why, isHeld := held[m]
want, wantHeld := row.held[m]
if isHeld != wantHeld || composed == wantHeld {
t.Errorf("%s: %s held %v (%q), composed %v; want held %v", row.name, m, isHeld, why, composed, wantHeld)
} else if isHeld && !strings.Contains(why, want) {
t.Errorf("%s: %s is held for %q, want its first reason, %q", row.name, m, why, want)
}
}
}
}