Make snapshot projections self-diagnosing

internal/agent/runner/runner.go

@@ -77,6 +77,12 @@ func (r *Runner) resticEnv() restic.Env {
 	}
 }
 
+// RefreshSnapshots reconciles the server's cached projection without running
+// a mutating repository job.
+func (r *Runner) RefreshSnapshots(ctx context.Context) error {
+	return r.reportSnapshots(ctx, r.resticEnv())
+}
+
 // sendStarted ships a job.started envelope.
 func (r *Runner) sendStarted(jobID string, kind api.JobKind, startedAt time.Time) {
 	env, _ := api.Marshal(api.MsgJobStarted, jobID, api.JobStartedPayload{
@@ -226,14 +232,9 @@ func (r *Runner) RunBackup(ctx context.Context, jobID string, paths, excludes, t
 		}
 	}
 
-	r.sendFinished(ctx, jobID, finishedAt, err, statsBlob)
-
 	// On a successful backup, refresh the server's snapshot projection.
-	// We do this *after* job.finished so the UI sees the job land first;
-	// the snapshot list is a follow-up that the host detail page polls
-	// or the dashboard sees on its next refresh. A failure here is
-	// logged but doesn't fail the job — the next successful backup will
-	// catch the projection up.
+	// Do this before job.finished so a failure in terminal reporting cannot
+	// prevent the independently useful projection refresh.
 	if err == nil {
 		if rerr := r.reportSnapshots(ctx, env); rerr != nil {
 			slog.Warn("runner: snapshots.report failed", "job_id", jobID, "err", rerr)
@@ -242,6 +243,7 @@ func (r *Runner) RunBackup(ctx context.Context, jobID string, paths, excludes, t
 			slog.Warn("runner: stats.report after backup failed", "job_id", jobID, "err", rerr)
 		}
 	}
+	r.sendFinished(ctx, jobID, finishedAt, err, statsBlob)
 
 	if err != nil {
 		return fmt.Errorf("runner backup: %w", err)
@@ -282,8 +284,6 @@ func (r *Runner) RunForget(ctx context.Context, jobID string, groups []restic.Fo
 	var seq atomic.Int64
 	err := env.RunForget(ctx, groups, dryRun, r.streamHandler(jobID, &seq))
 	finishedAt := time.Now().UTC()
-	r.sendFinished(ctx, jobID, finishedAt, err, nil)
-
 	// Refresh the server's snapshot projection — forget rewrites the
 	// index so the host's snapshot list almost certainly shrunk.
 	if err == nil {
@@ -292,6 +292,7 @@ func (r *Runner) RunForget(ctx context.Context, jobID string, groups []restic.Fo
 				"job_id", jobID, "err", rerr)
 		}
 	}
+	r.sendFinished(ctx, jobID, finishedAt, err, nil)
 
 	if err != nil {
 		return fmt.Errorf("runner forget: %w", err)
@@ -318,6 +319,9 @@ func (r *Runner) RunPrune(ctx context.Context, jobID string) error {
 		if rerr := r.reportStats(ctx, env, api.RepoStatsPayload{LastPruneAt: &pruneAt}); rerr != nil {
 			slog.Warn("runner: stats.report after prune failed", "job_id", jobID, "err", rerr)
 		}
+		if rerr := r.reportSnapshots(ctx, env); rerr != nil {
+			slog.Warn("runner: snapshots.report after prune failed", "job_id", jobID, "err", rerr)
+		}
 	}
 
 	r.sendFinished(ctx, jobID, finishedAt, err, nil)

internal/agent/runner/runner_test.go

@@ -116,7 +116,8 @@ func envelopeOrder(envs []api.Envelope) []api.MessageType {
 // TestRunPruneShipsExpectedEnvelopes drives RunPrune with a fake
 // binary that prints "prune" on stdout (for the log.stream envelope)
 // and emits valid stats JSON so reportStats can populate size fields.
-// Expected sequence: job.started → log.stream → repo.stats → job.finished.
+// Expected sequence: job.started → log.stream → repo.stats → snapshots.report
+// → job.finished.
 func TestRunPruneShipsExpectedEnvelopes(t *testing.T) {
 	t.Parallel()
 
@@ -126,6 +127,7 @@ func TestRunPruneShipsExpectedEnvelopes(t *testing.T) {
 case "$1" in
   prune)  echo "prune" ;;
   stats)  echo '`+statsJSON+`' ;;
+	  snapshots) echo "[]" ;;
   *)      echo "unknown: $*" ;;
 esac
 `)
@@ -138,7 +140,7 @@ esac
 
 	order := envelopeOrder(tx.envs)
 	// Confirm landmark envelope types appear in the required order.
-	wantTypes := []api.MessageType{api.MsgJobStarted, api.MsgLogStream, api.MsgRepoStats, api.MsgJobFinished}
+	wantTypes := []api.MessageType{api.MsgJobStarted, api.MsgLogStream, api.MsgRepoStats, api.MsgSnapshotsRpt, api.MsgJobFinished}
 	positions := map[api.MessageType]int{}
 	for i, mt := range order {
 		if _, seen := positions[mt]; !seen {
@@ -379,6 +381,15 @@ func TestRunInitShipsStartedAndFinished(t *testing.T) {
 	_ = firstEnvOfType(t, tx.envs, api.MsgJobFinished)
 }
 
+func firstIndexOfType(envs []api.Envelope, typ api.MessageType) int {
+	for i, env := range envs {
+		if env.Type == typ {
+			return i
+		}
+	}
+	return -1
+}
+
 // TestRunForgetShipsStartedAndFinished confirms the refactored
 // RunForget still produces job.started and job.finished envelopes.
 func TestRunForgetShipsStartedAndFinished(t *testing.T) {
@@ -402,5 +413,9 @@ esac
 		t.Fatalf("RunForget: %v", err)
 	}
 	_ = firstEnvOfType(t, tx.envs, api.MsgJobStarted)
-	_ = firstEnvOfType(t, tx.envs, api.MsgJobFinished)
+	finished := firstIndexOfType(tx.envs, api.MsgJobFinished)
+	refreshed := firstIndexOfType(tx.envs, api.MsgSnapshotsRpt)
+	if refreshed < 0 || finished < 0 || refreshed >= finished {
+		t.Fatalf("snapshot refresh must precede terminal report: %v", envelopeOrder(tx.envs))
+	}
 }