Files
orchestra/internal/federation/federation_test.go
T
kami 0d67af9976 Admit a first lease when a quota limit has no receipt history
Burn-in run 2 ingested its task and then sat queued forever. Every herdr in
the live config declares quota_limit_5h and quota_limit_weekly, the event log
holds zero QuotaReported events, and QuotaSince reported an empty window as
unknown. QuotaAvailability fails closed on unknown, so no harness could ever
be leased, and the only producer of a receipt is a completed lease.

The event log is Orchestra's whole accounting source, so a window holding no
receipts is observable zero consumption. QuotaSince now reports known for an
empty window and for a harness that has never reported. A receipt that
declares its own consumption unknown still fails closed.

The refusal also lied about its cause. federatedAvailability collapsed a base
gate refusal into the federation health string, so router health said "stale
heartbeat or unhealthy local backend" while the heartbeat was one second old.
Availability gates now name themselves through an optional
ReasonedAvailability contract: quota refusals say whether usage is unknown or
the window is exhausted and by how much, and worker refusals distinguish an
unregistered worker, a never-probed backend, a stale heartbeat, a stale
health check, and an unreachable backend.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 00:28:57 +04:00

308 lines
10 KiB
Go

package federation
import (
"os"
"path/filepath"
"testing"
"time"
)
func TestCursorIsMonotonicAndAuthenticationIsRequired(t *testing.T) {
r := &Registry{}
if err := r.Register(Worker{ID: "workpc", Token: "secret"}, ""); err != nil {
t.Fatal(err)
}
if err := r.Authenticate("workpc", "wrong"); err != ErrUnauthorized {
t.Fatalf("got %v", err)
}
if err := r.Authenticate("workpc", "secret"); err != nil {
t.Fatal(err)
}
if err := r.Ack("workpc", 7); err != nil {
t.Fatal(err)
}
if err := r.Ack("workpc", 6); err == nil {
t.Fatal("backwards cursor accepted")
}
if got, _ := r.Cursor("workpc"); got != 7 {
t.Fatalf("cursor = %d", got)
}
}
func TestSupportedProjectsPersistAndGateAvailability(t *testing.T) {
path := filepath.Join(t.TempDir(), "state.json")
r := &Registry{StatePath: path}
if err := r.Register(Worker{ID: "w", Token: "t", SupportedProjects: []string{"test-e2e"}}, ""); err != nil {
t.Fatal(err)
}
if !r.Supports("w", "test-e2e") || r.Supports("w", "correx") {
t.Fatalf("unexpected project support")
}
restarted := &Registry{StatePath: path}
if err := restarted.Load(); err != nil {
t.Fatal(err)
}
if err := restarted.Register(Worker{ID: "w", Token: "t", SupportedProjects: []string{"test-e2e"}}, ""); err != nil {
t.Fatal(err)
}
if !restarted.Supports("w", "test-e2e") || restarted.Supports("w", "correx") {
t.Fatalf("project support did not survive restart")
}
}
func TestAvailableRequiresFreshReachableLocalHerdrHealth(t *testing.T) {
r := &Registry{}
if err := r.Register(Worker{ID: "w", Token: "t"}, ""); err != nil {
t.Fatal(err)
}
if r.Available("w") {
t.Fatal("registration without local herdr probe admitted a worker")
}
if err := r.Heartbeat("w", WorkerHealth{HerdrStatus: "reachable", CheckedAt: time.Now().UTC()}); err != nil {
t.Fatal(err)
}
if !r.Available("w") {
t.Fatal("fresh reachable local herdr was not admitted")
}
if err := r.Heartbeat("w", WorkerHealth{HerdrStatus: "unreachable", CheckedAt: time.Now().UTC()}); err != nil {
t.Fatal(err)
}
if r.Available("w") {
t.Fatal("unreachable local herdr was admitted")
}
}
func TestPendingApprovalSurvivesRegistryRestart(t *testing.T) {
path := filepath.Join(t.TempDir(), "federation-state.json")
r := &Registry{StatePath: path}
if err := r.Load(); err != nil {
t.Fatal(err)
}
if err := r.Register(Worker{ID: "w", Token: "t"}, ""); err != nil {
t.Fatal(err)
}
capture, err := r.PutCapture("w", Capture{TaskID: "task", PaneID: "pane", Text: "Allow command?"})
if err != nil {
t.Fatal(err)
}
queued, err := r.Queue("w", Command{TaskID: "task", Kind: "grant_approval", PaneID: "pane", CaptureRevision: capture.Revision})
if err != nil {
t.Fatal(err)
}
info, err := os.Stat(path)
if err != nil {
t.Fatal(err)
}
if info.Mode().Perm() != 0600 {
t.Fatalf("federation state permissions = %o, want 0600", info.Mode().Perm())
}
restarted := &Registry{StatePath: path}
if err := restarted.Load(); err != nil {
t.Fatal(err)
}
if err := restarted.Register(Worker{ID: "w", Token: "intruder"}, ""); err != ErrUnauthorized {
t.Fatalf("recovered worker identity was hijackable: %v", err)
}
// A restart does not mark the worker online; it must prove its retained
// identity by registering again before recovered controls become available.
if err := restarted.Register(Worker{ID: "w", Token: "t"}, ""); err != nil {
t.Fatal(err)
}
gotCapture, ok := restarted.Capture("w", "task")
if !ok || gotCapture.Revision != capture.Revision || gotCapture.Text != capture.Text {
t.Fatalf("capture after restart = %#v, present=%v", gotCapture, ok)
}
commands, err := restarted.Commands("w")
if err != nil || len(commands) != 1 || commands[0].ID != queued.ID {
t.Fatalf("commands after restart = %#v, err=%v", commands, err)
}
if err := restarted.CompleteCommand("w", queued.ID, "acknowledged", ""); err != nil {
t.Fatal(err)
}
again := &Registry{StatePath: path}
if err := again.Load(); err != nil {
t.Fatal(err)
}
if err := again.Register(Worker{ID: "w", Token: "t"}, ""); err != nil {
t.Fatal(err)
}
commands, err = again.Commands("w")
if err != nil || len(commands) != 0 {
t.Fatalf("resolved command recovered as pending: %#v, err=%v", commands, err)
}
}
func TestRegisterRequiresAdmitTokenAndOwnToken(t *testing.T) {
r := &Registry{AdmitToken: "admit-secret"}
if err := r.Register(Worker{ID: "workpc", Token: "secret"}, "wrong"); err != ErrUnauthorized {
t.Fatalf("wrong admit token: got %v", err)
}
if err := r.Register(Worker{ID: "workpc", Token: "secret"}, "admit-secret"); err != nil {
t.Fatal(err)
}
// Re-registering the same id with a different token is a hijack
// attempt (S10), not a legitimate re-registration, and must be refused
// even with a valid admit token.
if err := r.Register(Worker{ID: "workpc", Token: "different"}, "admit-secret"); err != ErrUnauthorized {
t.Fatalf("hijack with different token: got %v", err)
}
// The same worker re-registering with its own token (e.g. after a
// restart) must still succeed.
if err := r.Register(Worker{ID: "workpc", Token: "secret"}, "admit-secret"); err != nil {
t.Fatalf("legitimate re-registration: %v", err)
}
}
func TestOfflineHookRunsOnceOnTransition(t *testing.T) {
called := make(chan Worker, 1)
r := &Registry{TTL: time.Millisecond, OnOffline: func(w Worker) { called <- w }}
if err := r.Register(Worker{ID: "workpc", Token: "secret"}, ""); err != nil {
t.Fatal(err)
}
r.mu.Lock()
w := r.workers["workpc"]
w.LastSeen = time.Now().Add(-time.Second)
r.workers["workpc"] = w
r.mu.Unlock()
r.Snapshot()
select {
case got := <-called:
if got.ID != "workpc" {
t.Fatal(got.ID)
}
case <-time.After(time.Second):
t.Fatal("offline hook not called")
}
r.Snapshot()
select {
case <-called:
t.Fatal("offline hook called twice")
case <-time.After(10 * time.Millisecond):
}
}
func TestHeartbeatProjectsWorkerOwnedHealth(t *testing.T) {
r := &Registry{}
if err := r.Register(Worker{ID: "workpc-opencode", Token: "secret"}, ""); err != nil {
t.Fatal(err)
}
checked := time.Now().UTC().Round(0)
errAt := checked.Add(-time.Minute)
if err := r.Heartbeat("workpc-opencode", WorkerHealth{
HerdrStatus: "unreachable", CheckedAt: checked, ActiveTask: "task-1", ActivePane: "pane-1",
LastError: "local herdr: connection refused", ErrorAt: errAt,
}); err != nil {
t.Fatal(err)
}
workers := r.Snapshot()
if len(workers) != 1 {
t.Fatalf("workers=%#v", workers)
}
h := workers[0].Health
if h.HerdrStatus != "unreachable" || h.ActiveTask != "task-1" || h.ActivePane != "pane-1" || h.LastError == "" || !h.CheckedAt.Equal(checked) || !h.ErrorAt.Equal(errAt) {
t.Fatalf("health=%#v", h)
}
}
func TestCaptureRevisionAndCommandQueue(t *testing.T) {
r := &Registry{}
if err := r.Register(Worker{ID: "w", Token: "t"}, ""); err != nil {
t.Fatal(err)
}
c, err := r.PutCapture("w", Capture{TaskID: "task", PaneID: "pane", Text: "Permission required\n$ ls"})
if err != nil || c.Revision != 1 {
t.Fatalf("capture=%#v err=%v", c, err)
}
again, err := r.PutCapture("w", Capture{TaskID: "task", PaneID: "pane", Text: c.Text})
if err != nil || again.Revision != 1 {
t.Fatalf("same capture=%#v err=%v", again, err)
}
cmd, err := r.Queue("w", Command{TaskID: "task", Kind: "grant_approval", PaneID: "pane", CaptureRevision: 1})
if err != nil {
t.Fatal(err)
}
commands, err := r.Commands("w")
if err != nil || len(commands) != 1 || commands[0].ID != cmd.ID {
t.Fatalf("commands=%#v err=%v", commands, err)
}
if err := r.CompleteCommand("w", cmd.ID, "acknowledged", ""); err != nil {
t.Fatal(err)
}
commands, _ = r.Commands("w")
if len(commands) != 0 {
t.Fatalf("pending=%#v", commands)
}
}
func TestResolvedCommandsArePrunedButPendingOnesSurvive(t *testing.T) {
r := &Registry{}
if err := r.Register(Worker{ID: "w", Token: "t"}, ""); err != nil {
t.Fatal(err)
}
old, err := r.Queue("w", Command{TaskID: "task", Kind: "grant_approval", PaneID: "pane", CaptureRevision: 1})
if err != nil {
t.Fatal(err)
}
if err := r.CompleteCommand("w", old.ID, "acknowledged", ""); err != nil {
t.Fatal(err)
}
pending, err := r.Queue("w", Command{TaskID: "task", Kind: "deny_approval", PaneID: "pane", CaptureRevision: 2})
if err != nil {
t.Fatal(err)
}
// Age both past the retention window; only the resolved one may go.
r.mu.Lock()
for i := range r.commands["w"] {
r.commands["w"][i].CreatedAt = time.Now().UTC().Add(-2 * CommandRetention)
}
r.mu.Unlock()
if _, err := r.Commands("w"); err != nil {
t.Fatal(err)
}
if _, ok := r.Command("w", old.ID); ok {
t.Fatal("resolved command past retention was not pruned")
}
if c, ok := r.Command("w", pending.ID); !ok || c.Status != "pending" {
t.Fatalf("pending command was pruned: %#v ok=%v", c, ok)
}
}
// Each unavailability condition must name itself. One collapsed reason once
// reported a stale heartbeat for a worker whose heartbeat was a second old.
func TestUnavailableNamesTheFailingCondition(t *testing.T) {
r := &Registry{TTL: time.Minute}
if got, want := r.Unavailable("nobody"), "worker unavailable: no worker registered for this harness"; got != want {
t.Fatalf("unregistered reason=%q, want %q", got, want)
}
if err := r.Register(Worker{ID: "w", Token: "t"}, ""); err != nil {
t.Fatal(err)
}
if got, want := r.Unavailable("w"), "worker unavailable: backend health never reported"; got != want {
t.Fatalf("unprobed reason=%q, want %q", got, want)
}
if err := r.Heartbeat("w", WorkerHealth{HerdrStatus: "unreachable", CheckedAt: time.Now().UTC()}); err != nil {
t.Fatal(err)
}
if got, want := r.Unavailable("w"), "worker unavailable: backend unreachable"; got != want {
t.Fatalf("unhealthy-backend reason=%q, want %q", got, want)
}
if err := r.Heartbeat("w", WorkerHealth{HerdrStatus: "reachable", CheckedAt: time.Now().UTC()}); err != nil {
t.Fatal(err)
}
if got := r.Unavailable("w"); got != "" {
t.Fatalf("fresh reachable worker reason=%q, want admitted", got)
}
stale := &Registry{TTL: time.Nanosecond}
if err := stale.Register(Worker{ID: "w", Token: "t"}, ""); err != nil {
t.Fatal(err)
}
if err := stale.Heartbeat("w", WorkerHealth{HerdrStatus: "reachable", CheckedAt: time.Now().UTC()}); err != nil {
t.Fatal(err)
}
if got, want := stale.Unavailable("w"), "worker unavailable: stale heartbeat"; got != want {
t.Fatalf("stale reason=%q, want %q", got, want)
}
}