The bus on NATS: both transports behind seams, and the rollout switch #87

Merged
jschoubben merged 40 commits from feat/nats-genesis into main 2026-09-27 17:36:41 +00:00
8 changed files with 248 additions and 2 deletions
Showing only changes of commit 8e2824201a - Show all commits
+17
View File
@@ -702,6 +702,23 @@ func raiseTheBus(ctx context.Context, inv *inventory.Inventory, address string)
}
defer js.Close()
// **Its own user, before anything else.** The controller's account is created by the installer at
// a bootstrap password, before there is a controller to mint one — so nothing recorded a hash for
// it, and the first composition would leave the writer out of the file it was writing. Recorded
// only if absent: a credential the mesh minted since is the one that counts.
// **Its own user, before anything else it does here.** The controller's account is created by the
// installer at a bootstrap password, before there is a controller to mint one — so nothing
// recorded a hash for it, and the first composition would leave the writer out of the file it was
// writing: a bus nothing can connect to, produced by the thing connected to it. Recorded only if
// absent, so a restart cannot put the bootstrap credential back over a rotated one.
if user, password, _ := broker.CredentialIn(address); user != "" && password != "" {
if err := inv.SeedBusUser(ctx, inventory.BusUser{
Username: user, Kind: inventory.BusController,
}, password); err != nil {
return fmt.Errorf("cannot record the credential this control plane is using: %w", err)
}
}
nodes, err := inv.Nodes(ctx)
if err != nil {
return err
+126
View File
@@ -0,0 +1,126 @@
package broker
import (
"encoding/json"
"os"
"path/filepath"
"regexp"
"sort"
"strings"
"testing"
"golang.org/x/crypto/bcrypt"
)
// **The first user list the installer carries must be the one the controller would compose.**
//
// At genesis there is no mesh to write the bus's user list, so the installer carries one: the
// controller's own account, at a bootstrap password, the way the store is reached at
// `postgres:bootstrap` (novox/hq design 25 §4, task 1.7). It is written by hand in a template and
// derived in code here, which is two statements of one fact — so this compares them.
//
// Getting it wrong is the worst kind of silent: a controller whose carried permissions are narrower
// than the ones it derives comes up, connects, and is refused on the first thing it tries, with an
// authorisation error that names a subject and not the template that forgot it. And a mesh cannot be
// raised twice to find out.
func TestTheInstallersFirstUserListIsWhatTheControllerWouldCompose(t *testing.T) {
accounts := theCarriedAccounts(t)
want, err := PermissionsFor(Principal{Kind: KindController, PasswordHash: "x"})
if err != nil {
t.Fatal(err)
}
carriedPub := subjectsIn(accounts, "publish")
carriedSub := subjectsIn(accounts, "subscribe")
if diff := missing(want.Publish, carriedPub); len(diff) > 0 {
t.Errorf("the installer's user list does not let the controller publish %v — it would come up "+
"and be refused on the first thing it tried", diff)
}
if diff := missing(want.Subscribe, carriedSub); len(diff) > 0 {
t.Errorf("the installer's user list does not let the controller subscribe %v", diff)
}
// And nothing wider than what it derives, or genesis quietly grants a privilege the composition
// takes away again on the first push.
if diff := missing(carriedPub, want.Publish); len(diff) > 0 {
t.Errorf("the installer's user list lets the controller publish %v, which it does not derive", diff)
}
if diff := missing(carriedSub, want.Subscribe); len(diff) > 0 {
t.Errorf("the installer's user list lets the controller subscribe %v, which it does not derive", diff)
}
// The credential is the bootstrap one and the hash really is of it, because a hash of something
// else is a controller that cannot log in to the bus it was just given.
hash := regexp.MustCompile(`\$2[aby]?\$[0-9]+\$[A-Za-z0-9./]{53}`).FindString(accounts)
if hash == "" {
t.Fatal("the installer's user list carries no password hash")
}
if err := bcrypt.CompareHashAndPassword([]byte(hash), []byte("bootstrap")); err != nil {
t.Fatalf("the carried hash does not verify the bootstrap credential the template also carries: %v", err)
}
}
// theCarriedAccounts is the accounts file the installer's template writes at genesis.
func theCarriedAccounts(t *testing.T) string {
t.Helper()
path := filepath.Join("..", "..", "..", "mesh-host", "examples", "foundation-first-node-nats.lock")
raw, err := os.ReadFile(path)
if err != nil {
t.Skipf("the host's checkout is not beside this one: %v", err)
}
// The template is JSON with line comments, which is how every one of them is written.
var lines []string
for _, l := range strings.Split(string(raw), "\n") {
if !strings.HasPrefix(strings.TrimSpace(l), "//") {
lines = append(lines, l)
}
}
var bundle struct {
Resources []map[string]any `json:"resources"`
}
if err := json.Unmarshal([]byte(strings.Join(lines, "\n")), &bundle); err != nil {
t.Fatalf("the template is not readable: %v", err)
}
for _, r := range bundle.Resources {
if r["id"] == "bus-accounts" {
content, _ := r["content"].(string)
if content == "" {
t.Fatal("the template's accounts file is empty, so the bus would refuse every connection")
}
return content
}
}
t.Fatal("the template carries no accounts file, so a mesh raised from it has a bus nobody may use")
return ""
}
// subjectsIn reads one allow-list out of a composed accounts file.
func subjectsIn(accounts, which string) []string {
found := regexp.MustCompile(which + `: \{ allow: \[([^\]]*)\]`).FindStringSubmatch(accounts)
if len(found) != 2 {
return nil
}
var out []string
for _, part := range strings.Split(found[1], ",") {
if s := strings.Trim(strings.TrimSpace(part), `"`); s != "" {
out = append(out, s)
}
}
sort.Strings(out)
return out
}
// missing is what is in want and not in got.
func missing(want, got []string) []string {
have := map[string]bool{}
for _, g := range got {
have[g] = true
}
var out []string
for _, w := range want {
if !have[w] {
out = append(out, w)
}
}
return out
}
-1
View File
@@ -167,7 +167,6 @@ func PermissionsFor(p Principal) (Permissions, error) {
// like the catalogue does — which is why no holder needs to publish into anybody's inbox.
for _, seat := range meshSeatsTheControllerUses {
pub = append(pub, "mesh.seat."+seat+".accept.>")
sub = append(sub, "mesh.seat."+seat+".event.>")
}
// The two events it reacts to, and its ack subject on the stream they arrive from
+20
View File
@@ -35,6 +35,26 @@ func OnNATS() (address string, on bool, err error) {
return address, true, nil
}
// CredentialIn reads the user and password out of a bus address, and the address without them.
//
// The controller's own credential arrives in its address, the way the old bus's does. Split out so the
// controller can record a hash of what it is actually using: its user is created by the installer at a
// bootstrap password, before the controller exists to mint one, and a composition that left itself out
// would produce a bus the writer cannot connect to.
func CredentialIn(address string) (user, password, bare string) {
at := strings.LastIndex(address, "@")
if at < 0 {
return "", "", address
}
scheme := ""
rest := address[:at]
if i := strings.Index(rest, "://"); i >= 0 {
scheme, rest = rest[:i+3], rest[i+3:]
}
user, password, _ = strings.Cut(rest, ":")
return user, password, scheme + address[at+1:]
}
// MustBeOneBus refuses a configuration that names both buses for the mesh's own traffic.
//
// **Both clients ship and that is the point; both being live is not.** The rollout moves every node
+20
View File
@@ -38,3 +38,23 @@ func TestOneBusOrNeitherIsAllowed(t *testing.T) {
}
}
}
// The controller's own credential arrives in its address, and has to be readable out of it — its user
// is created by the installer at a bootstrap password, before the controller exists to mint one.
func TestACredentialIsReadOutOfABusAddress(t *testing.T) {
for _, c := range []struct{ in, user, password, bare string }{
{"nats://controller:secret@127.0.0.1:4222", "controller", "secret", "nats://127.0.0.1:4222"},
{"controller:secret@127.0.0.1:4222", "controller", "secret", "127.0.0.1:4222"},
{"nats://127.0.0.1:4222", "", "", "nats://127.0.0.1:4222"},
{"127.0.0.1:4222", "", "", "127.0.0.1:4222"},
// A password containing an at-sign: split on the last one, or the address becomes part of the
// credential and the connection goes somewhere nobody named.
{"nats://controller:a@b@127.0.0.1:4222", "controller", "a@b", "nats://127.0.0.1:4222"},
} {
user, password, bare := CredentialIn(c.in)
if user != c.user || password != c.password || bare != c.bare {
t.Errorf("%q read as %q/%q at %q; wanted %q/%q at %q",
c.in, user, password, bare, c.user, c.password, c.bare)
}
}
}
+1 -1
View File
@@ -24,7 +24,7 @@ accounts {
users = [
{ user: "controller", password: "$2a$11$cccccccccccccccccccccc", permissions: {
publish: { allow: ["$JS.ACK.CONTROL.controller.>", "$JS.ACK.EVENTS.controller.>", "$JS.API.>", "_INBOX.enrol.>", "mesh.control.>", "mesh.node.>", "mesh.seat.mesh-build-machine.accept.>"] }
subscribe: { allow: ["$JS.API.>", "_INBOX.controller.>", "mesh.control.>", "mesh.mod.mesh-catalog.event.catching-up", "mesh.mod.mesh-catalog.event.upgraded", "mesh.seat.mesh-build-machine.event.>", "mesh.seat.mesh-build-machine.event.built"] }
subscribe: { allow: ["$JS.API.>", "_INBOX.controller.>", "mesh.control.>", "mesh.mod.mesh-catalog.event.catching-up", "mesh.mod.mesh-catalog.event.upgraded", "mesh.seat.mesh-build-machine.event.built"] }
allow_responses: { max: 1, ttl: "1m" }
} }
{ user: "enrol.one", password: "$2a$11$eeeeeeeeeeeeeeeeeeeeee", permissions: {
+27
View File
@@ -136,3 +136,30 @@ func (i *Inventory) ForgetBusUsersOf(ctx context.Context, node string) error {
_, err := i.store.Pool().Exec(ctx, `delete from bus_user where node = $1`, node)
return err
}
// SeedBusUser records a hash of a credential the mesh did not mint, so a composition contains it.
//
// **Genesis is the reason this exists.** The controller's own user is created before the controller
// runs — by the installer, at a well-known bootstrap password, the way the store's and the old bus's
// are (`postgres:bootstrap`, `guest:guest`). Nothing minted it, so nothing recorded a hash for it, and
// the controller's first composition would leave itself out of the very file it was writing: a bus
// nothing can connect to, produced by the thing connected to it.
//
// Idempotent, and it does not overwrite. A credential the mesh *did* mint is the one that counts, so
// once there is a row this does nothing — otherwise a restart would put the bootstrap password back
// over a rotated one.
func (i *Inventory) SeedBusUser(ctx context.Context, u BusUser, password string) error {
if u.Username == "" || u.Kind == "" || password == "" {
return errors.New("a bus user needs a username, a kind and the credential it is using")
}
hash, err := bcrypt.GenerateFromPassword([]byte(password), bcrypt.DefaultCost)
if err != nil {
return fmt.Errorf("cannot hash a bus password: %w", err)
}
_, err = i.store.Pool().Exec(ctx,
`insert into bus_user (username, kind, node, module, password_hash)
values ($1, $2, $3, $4, $5)
on conflict (username) do nothing`,
u.Username, u.Kind, u.Node, u.Module, string(hash))
return err
}
+37
View File
@@ -121,3 +121,40 @@ func TestForgettingTheUsersOfNoNodeIsRefused(t *testing.T) {
t.Fatal("forgetting the bus users of no node was allowed")
}
}
// The controller's own user is created by the installer, so the mesh has to be able to record a
// credential it did not mint — or the first composition leaves the writer out of the file it writes.
func TestACredentialTheMeshDidNotMintIsRecordedOnceAndNotOverwritten(t *testing.T) {
inv := ForTest(t)
ctx := context.Background()
if err := inv.SeedBusUser(ctx, BusUser{Username: "controller", Kind: BusController},
"bootstrap"); err != nil {
t.Fatal(err)
}
hash, known, err := inv.BusUserHash(ctx, "controller")
if err != nil || !known {
t.Fatalf("the credential was not recorded: %v %v", known, err)
}
if err := bcrypt.CompareHashAndPassword([]byte(hash), []byte("bootstrap")); err != nil {
t.Fatalf("what was recorded does not verify the credential given: %v", err)
}
// Minted since, then seeded again — which is what a restart does. The rotation must stand, or
// every restart would put the bootstrap password back over it.
minted, err := inv.MintBusPassword(ctx, BusUser{Username: "controller", Kind: BusController})
if err != nil {
t.Fatal(err)
}
if err := inv.SeedBusUser(ctx, BusUser{Username: "controller", Kind: BusController},
"bootstrap"); err != nil {
t.Fatal(err)
}
hash, _, err = inv.BusUserHash(ctx, "controller")
if err != nil {
t.Fatal(err)
}
if err := bcrypt.CompareHashAndPassword([]byte(hash), []byte(minted)); err != nil {
t.Fatal("a restart put the bootstrap credential back over a rotated one")
}
}