Merge pull request 'Release v1.2.1 — stable Agent updates controls' (#53) from release-v1.2.1 into main

.gitea/workflows/ci.yml

@@ -3,7 +3,7 @@
 # Notes for anyone editing this file:
 #
 # Custom runner image
-#   Every job runs inside `gitea.dcglab.co.uk/steve/ci-runner-go`
+#   Every job runs inside `docker.dcglab.co.uk/public/ci-runner-go:latest`
 #   (recipe: https://gitea.dcglab.co.uk/steve/ci/src/branch/main/images/ci-runner-go).
 #   That image already ships:
 #     * Go on PATH at /usr/local/go/bin (so `actions/setup-go` is
@@ -11,8 +11,8 @@
 #       otherwise re-download Go on every job)
 #     * Node.js + npm (used by docs / e2e workflows)
 #     * Docker CLI, Buildx, Compose v2 (used by docker-build steps)
-#   When bumping the Go floor, push a new ci-runner-go image with
-#   the matching Go version and bump the date pin in IMAGE below.
+#   The runner image is intentionally tracked via `latest`: it is shared CI
+#   infrastructure, and updates should propagate to consuming repositories.
 #
 # Self-hosted runner expectations
 #   Each runner host bind-mounts persistent volumes for
@@ -71,10 +71,7 @@ jobs:
     name: Test (${{ matrix.name }})
     runs-on: ubuntu-latest
     container:
-      image: docker.dcglab.co.uk/ci-runner-go:2026-05-15
-      credentials:
-        username: ${{ secrets.ZOT_USERNAME }}
-        password: ${{ secrets.ZOT_PASSWORD }}
+      image: docker.dcglab.co.uk/public/ci-runner-go:latest
     strategy:
       fail-fast: false
       matrix:
@@ -110,10 +107,7 @@ jobs:
     name: Lint
     runs-on: ubuntu-latest
     container:
-      image: docker.dcglab.co.uk/ci-runner-go:2026-05-15
-      credentials:
-        username: ${{ secrets.ZOT_USERNAME }}
-        password: ${{ secrets.ZOT_PASSWORD }}
+      image: docker.dcglab.co.uk/public/ci-runner-go:latest
     steps:
       - uses: actions/checkout@v4
       - uses: golangci/golangci-lint-action@v7
@@ -130,10 +124,7 @@ jobs:
     name: Build (${{ matrix.goos }}/${{ matrix.goarch }})
     runs-on: ubuntu-latest
     container:
-      image: docker.dcglab.co.uk/ci-runner-go:2026-05-15
-      credentials:
-        username: ${{ secrets.ZOT_USERNAME }}
-        password: ${{ secrets.ZOT_PASSWORD }}
+      image: docker.dcglab.co.uk/public/ci-runner-go:latest
     strategy:
       fail-fast: false
       matrix:

.gitea/workflows/e2e.yml

@@ -31,7 +31,7 @@ jobs:
   e2e:
     name: Playwright vs docker-compose
     runs-on: ubuntu-latest
-    container: gitea.dcglab.co.uk/steve/ci-runner-go:2026-05-08
+    container: docker.dcglab.co.uk/public/ci-runner-go:latest
     timeout-minutes: 15
     steps:
       - uses: actions/checkout@v4

.gitea/workflows/release.yml

@@ -12,7 +12,9 @@
 #     plus install.sh / install.ps1 / the systemd unit baked in under
 #     /opt/restic-manager/dist (the read-only fallback path the server
 #     handlers use when <DataDir>/... is empty).
-#   * Pushes to zot OCI registry (docker.dcglab.co.uk).
+#   * Pushes to the public namespace in the zot OCI registry
+#     (docker.dcglab.co.uk/public/restic-manager). Pulls are anonymous;
+#     pushes remain authenticated.
 #
 # Tag fan-out
 #   * tag push: :vX.Y.Z, :X.Y, :X
@@ -29,7 +31,7 @@ on:
 
 env:
   REGISTRY: docker.dcglab.co.uk
-  IMAGE_NAME: restic-manager
+  IMAGE_NAME: public/restic-manager
 
 # Force bash as the default shell — see ci.yml header.
 defaults:
@@ -41,10 +43,7 @@ jobs:
     name: Build + push image
     runs-on: ubuntu-latest
     container:
-      image: docker.dcglab.co.uk/ci-runner-go:2026-05-15
-      credentials:
-        username: ${{ secrets.ZOT_USERNAME }}
-        password: ${{ secrets.ZOT_PASSWORD }}
+      image: docker.dcglab.co.uk/public/ci-runner-go:latest
     steps:
       - uses: actions/checkout@v4
 

CHANGELOG.md

@@ -6,6 +6,49 @@ and the project follows [Semantic Versioning](https://semver.org/).
 
 ## [Unreleased]
 
+## [1.2.1] - 2026-08-22
+
+### Fixed
+
+- Made the Agent updates host-selection screen stable and responsive: removed
+  the redundant typed-count confirmation, corrected name/version/eligibility
+  filtering, and stopped replacing idle or terminal controls through polling.
+  Progress polling now runs only while a rollout is actively running. ([#50])
+
+## [1.2.0] - 2026-08-22
+
+### Added
+
+- Fleet-wide agent updates are now discoverable from Settings and the
+  dashboard, with arbitrary host subsets, name/tag/version/eligibility
+  filtering, explicit exclusion reasons, server-side membership validation,
+  and a canary-first pause after the first verified reconnect. Halted rolls
+  expose retry and resume actions. ([#43])
+- Running jobs with no recent persisted activity are detected server-side and
+  surfaced through deduplicated alerts. Kind-aware thresholds protect normal
+  long-running work, terminal jobs self-resolve, and Prometheus exports stuck
+  job count and oldest inactivity age. ([#41])
+- Snapshot projections now retain host-level refresh timestamps—including
+  authoritative empty reports—expose a derived stale flag, refresh after
+  backup, forget, and prune, and support an explicit operator reconciliation
+  endpoint. ([#40])
+
+### Fixed
+
+- Raised the bounded WebSocket read limit on both peers so ordinary large
+  snapshot reports and restic events no longer disconnect otherwise healthy
+  agents. Regression coverage exercises payloads beyond the library's former
+  32 KiB default in both directions. ([#44])
+- Fleet-update timeout verification now performs a final authoritative read of
+  the agent version delivered by the reconnect `hello`, avoiding a poll/deadline
+  race and reporting the last observed version when verification fails. ([#43])
+
+### Changed
+
+- CI runner images and release images use the anonymous-pull `public`
+  namespace at `docker.dcglab.co.uk`; registry authentication remains required
+  only for publishing.
+
 ## [1.1.1] - 2026-08-22
 
 ### Fixed
@@ -140,7 +183,8 @@ with a web UI, JSON API, and self-updating agent fleet.
   go vet, golangci-lint).
 - Threat model published (`docs/threat-model.md`).
 
-[Unreleased]: https://gitea.dcglab.co.uk/steve/restic-manager/compare/v1.1.1...HEAD
+[Unreleased]: https://gitea.dcglab.co.uk/steve/restic-manager/compare/v1.2.0...HEAD
+[1.2.0]: https://gitea.dcglab.co.uk/steve/restic-manager/compare/v1.1.1...v1.2.0
 [1.1.1]: https://gitea.dcglab.co.uk/steve/restic-manager/compare/v1.1.0...v1.1.1
 [1.1.0]: https://gitea.dcglab.co.uk/steve/restic-manager/releases/tag/v1.1.0
 [1.0.0]: https://gitea.dcglab.co.uk/steve/restic-manager/releases/tag/v1.0.0
@@ -148,3 +192,7 @@ with a web UI, JSON API, and self-updating agent fleet.
 [#37]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/37
 [#36]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/36
 [#34]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/34
+[#40]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/40
+[#41]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/41
+[#43]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/43
+[#44]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/44

internal/alert/engine.go

@@ -61,9 +61,20 @@ type Engine struct {
 	stuckThresholds   map[string]time.Duration
 
 	closeOnce sync.Once
+	notifyWG  sync.WaitGroup
 	done      chan struct{}
 }
 
+func (e *Engine) dispatchNotification(ctx context.Context, payload notification.Payload) {
+	e.notifyWG.Add(1)
+	go func() {
+		defer e.notifyWG.Done()
+		e.hub.Dispatch(ctx, payload)
+	}()
+}
+
+func (e *Engine) waitNotifications() { e.notifyWG.Wait() }
+
 // NewEngine builds the engine. agentOfflineFloor + tickPeriod default
 // to 15min and 60s respectively when zero.
 func NewEngine(st *store.Store, hub *notification.Hub) *Engine {

internal/alert/rules.go

@@ -60,7 +60,7 @@ func (e *Engine) raiseAndNotify(ctx context.Context, hostID, kind, dedupKey, sev
 	if err == nil {
 		hostName = host.Name
 	}
-	go e.hub.Dispatch(ctx, notification.Payload{
+	e.dispatchNotification(ctx, notification.Payload{
 		Event:    notification.EventRaised,
 		AlertID:  id,
 		Severity: severity,
@@ -85,7 +85,7 @@ func (e *Engine) Acknowledge(ctx context.Context, alertID, userID string, when t
 		return nil //nolint:nilerr
 	}
 	p := alertPayload(ctx, e.store, notification.EventAcknowledged, a)
-	go e.hub.Dispatch(context.WithoutCancel(ctx), p)
+	e.dispatchNotification(context.WithoutCancel(ctx), p)
 	return nil
 }
 
@@ -99,7 +99,7 @@ func (e *Engine) Resolve(ctx context.Context, alertID string, when time.Time) er
 		return nil
 	}
 	p := alertPayload(ctx, e.store, notification.EventResolved, a)
-	go e.hub.Dispatch(context.WithoutCancel(ctx), p)
+	e.dispatchNotification(context.WithoutCancel(ctx), p)
 	return nil
 }
 
@@ -164,7 +164,7 @@ func (e *Engine) resolveAndNotify(ctx context.Context, hostID, kind, dedupKey st
 		if a.Kind != kind || a.DedupKey != dedupKey {
 			continue
 		}
-		go e.hub.Dispatch(ctx, notification.Payload{
+		e.dispatchNotification(ctx, notification.Payload{
 			Event:    notification.EventResolved,
 			AlertID:  a.ID,
 			Severity: a.Severity,

internal/alert/rules_test.go

@@ -25,6 +25,7 @@ func setupEngine(t *testing.T) (*Engine, *store.Store, string) {
 	aead, _ := crypto.NewAEAD(key)
 	hub := notification.NewHub(st, aead, "https://rm.example")
 	eng := NewEngine(st, hub)
+	t.Cleanup(eng.waitNotifications)
 	hostID := ulid.Make().String()
 	if err := st.CreateHost(context.Background(), store.Host{
 		ID: hostID, Name: "alfa-01", OS: "linux", Arch: "amd64",

internal/server/fleetupdate/worker.go

@@ -71,6 +71,16 @@ func NewWorker(st *store.Store, hub Hub, disp Dispatcher, alerts AlertRaiser) *W
 // worker goroutine. Returns the new fleet_update_id on success.
 // store.ErrFleetUpdateRunning bubbles up unchanged.
 func (w *Worker) Start(ctx context.Context, userID, targetVersion string, hostIDs []string) (string, error) {
+	return w.start(ctx, userID, targetVersion, hostIDs, false)
+}
+
+// StartCanary starts the same sequential rollout but pauses after the first
+// verified host so an operator can inspect it before starting the remainder.
+func (w *Worker) StartCanary(ctx context.Context, userID, targetVersion string, hostIDs []string) (string, error) {
+	return w.start(ctx, userID, targetVersion, hostIDs, true)
+}
+
+func (w *Worker) start(ctx context.Context, userID, targetVersion string, hostIDs []string, pauseAfterFirst bool) (string, error) {
 	if userID == "" || targetVersion == "" {
 		return "", errors.New("fleetupdate: userID and targetVersion required")
 	}
@@ -85,6 +95,7 @@ func (w *Worker) Start(ctx context.Context, userID, targetVersion string, hostID
 		StartedByUserID: userID,
 		TargetVersion:   targetVersion,
 		Status:          "running",
+		PauseAfterFirst: pauseAfterFirst,
 	}, hostIDs); err != nil {
 		return "", err
 	}
@@ -137,6 +148,13 @@ func (w *Worker) run(ctx context.Context, fuID, userID, targetVersion string) {
 
 		next := pending[0]
 		w.processHost(ctx, fuID, userID, next)
+		if fu.PauseAfterFirst && next.Position == 0 {
+			updated, _, gerr := w.store.GetFleetUpdate(ctx, fuID)
+			if gerr == nil && updated.Status == "running" {
+				_ = w.store.HaltFleetUpdate(ctx, fuID, "canary succeeded; review the host, then resume remaining agents", time.Now().UTC())
+			}
+			return
+		}
 	}
 }
 
@@ -195,7 +213,19 @@ func (w *Worker) processHost(ctx context.Context, fuID, userID string, slot stor
 			return
 		}
 	}
-	reason := fmt.Sprintf("timeout waiting for %s to reach %s", hostID, w.targetVersion)
+	// One authoritative final read closes the poll/deadline race. AgentVersion
+	// is written directly from the reconnect hello handshake.
+	lastVersion := "unknown"
+	if h, err := w.store.GetHost(ctx, hostID); err == nil && h != nil {
+		if h.AgentVersion == w.targetVersion {
+			_ = w.store.SetFleetUpdateHostStatus(ctx, fuID, hostID, "succeeded", "", jobID)
+			return
+		}
+		if h.AgentVersion != "" {
+			lastVersion = h.AgentVersion
+		}
+	}
+	reason := fmt.Sprintf("timeout waiting for %s reconnect hello at %s (last reported %s)", hostID, w.targetVersion, lastVersion)
 	_ = w.store.SetFleetUpdateHostStatus(ctx, fuID, hostID, "failed", reason, jobID)
 	w.halt(ctx, fuID, reason)
 }

internal/server/fleetupdate/worker_test.go

@@ -160,6 +160,35 @@ func TestWorkerTwoHostsBothSucceed(t *testing.T) {
 	}
 }
 
+func TestWorkerCanaryPausesAfterFirstVerifiedHost(t *testing.T) {
+	st := openStore(t)
+	uid := mustCreateAdmin(t, st)
+	h1 := mustCreateHost(t, st, "canary", "v0")
+	h2 := mustCreateHost(t, st, "remainder", "v0")
+	hub := &fakeHub{online: map[string]bool{h1: true, h2: true}}
+	disp := &fakeDispatcher{st: st, target: "v2", delayMS: 20}
+	alerts := &recAlert{}
+	w := NewWorker(st, hub, disp, alerts)
+	w.pollPeriod = 10 * time.Millisecond
+	w.hostTimeout = time.Second
+
+	fuID, err := w.StartCanary(context.Background(), uid, "v2", []string{h1, h2})
+	if err != nil {
+		t.Fatalf("start canary: %v", err)
+	}
+	fu := waitForStatus(t, st, fuID, "halted", 2*time.Second)
+	if fu.HaltedReason != "canary succeeded; review the host, then resume remaining agents" {
+		t.Fatalf("halt reason: %q", fu.HaltedReason)
+	}
+	_, hosts, _ := st.GetFleetUpdate(context.Background(), fuID)
+	if hosts[0].Status != "succeeded" || hosts[1].Status != "pending" {
+		t.Fatalf("canary statuses: %+v", hosts)
+	}
+	if len(alerts.reasons) != 0 {
+		t.Fatalf("successful canary pause must not alert: %v", alerts.reasons)
+	}
+}
+
 func TestWorkerSecondHostTimesOutHalts(t *testing.T) {
 	st := openStore(t)
 	uid := mustCreateAdmin(t, st)

internal/server/http/fleet_update.go

@@ -32,6 +32,13 @@ import (
 type fleetUpdateStartReq struct {
 	TargetVersion string   `json:"target_version,omitempty"`
 	HostIDs       []string `json:"host_ids,omitempty"`
+	CanaryFirst   bool     `json:"canary_first,omitempty"`
+}
+
+type fleetUpdateCandidate struct {
+	Host     store.Host
+	Eligible bool
+	Reason   string
 }
 
 // fleetUpdateHostView is one row in the JSON response for GET
@@ -56,6 +63,7 @@ type fleetUpdateView struct {
 	CurrentHostID   string                `json:"current_host_id,omitempty"`
 	HaltedReason    string                `json:"halted_reason,omitempty"`
 	CompletedAt     *string               `json:"completed_at,omitempty"`
+	CanaryFirst     bool                  `json:"canary_first"`
 	Hosts           []fleetUpdateHostView `json:"hosts"`
 }
 
@@ -65,6 +73,7 @@ type fleetUpdateView struct {
 type fleetUpdatePage struct {
 	// Idle-state fields.
 	OutOfDateHosts []store.Host // online hosts whose version != target
+	Candidates     []fleetUpdateCandidate
 	TargetVersion  string
 
 	// Active-state fields. Nil when no fleet update has ever run.
@@ -100,6 +109,11 @@ func (s *Server) handleAPIFleetUpdateStart(w stdhttp.ResponseWriter, r *stdhttp.
 	if target == "" {
 		target = version.Version
 	}
+	if target != version.Version {
+		writeJSONError(w, stdhttp.StatusUnprocessableEntity, "unsupported_target_version",
+			"fleet updates can only target the running server version")
+		return
+	}
 	hostIDs := body.HostIDs
 	if len(hostIDs) == 0 {
 		derived, err := s.deriveOutOfDateOnlineHostIDs(r.Context(), target)
@@ -108,6 +122,19 @@ func (s *Server) handleAPIFleetUpdateStart(w stdhttp.ResponseWriter, r *stdhttp.
 			return
 		}
 		hostIDs = derived
+	} else {
+		validated, reasons, err := s.validateFleetUpdateHostIDs(r.Context(), target, hostIDs)
+		if err != nil {
+			writeJSONError(w, stdhttp.StatusInternalServerError, "internal", err.Error())
+			return
+		}
+		if len(reasons) > 0 {
+			writeJSON(w, stdhttp.StatusUnprocessableEntity, map[string]any{
+				"code": "ineligible_hosts", "message": "one or more selected hosts are ineligible", "reasons": reasons,
+			})
+			return
+		}
+		hostIDs = validated
 	}
 	if len(hostIDs) == 0 {
 		writeJSONError(w, stdhttp.StatusConflict, "no_hosts_eligible",
@@ -115,7 +142,20 @@ func (s *Server) handleAPIFleetUpdateStart(w stdhttp.ResponseWriter, r *stdhttp.
 		return
 	}
 
-	fuID, err := s.deps.FleetWorker.Start(r.Context(), user.ID, target, hostIDs)
+	var fuID string
+	var err error
+	if body.CanaryFirst && len(hostIDs) > 1 {
+		worker, supported := s.deps.FleetWorker.(interface {
+			StartCanary(context.Context, string, string, []string) (string, error)
+		})
+		if !supported {
+			writeJSONError(w, stdhttp.StatusServiceUnavailable, "canary_unavailable", "")
+			return
+		}
+		fuID, err = worker.StartCanary(r.Context(), user.ID, target, hostIDs)
+	} else {
+		fuID, err = s.deps.FleetWorker.Start(r.Context(), user.ID, target, hostIDs)
+	}
 	if err != nil {
 		if errors.Is(err, store.ErrFleetUpdateRunning) {
 			writeJSONError(w, stdhttp.StatusConflict, "fleet_update_in_progress", err.Error())
@@ -129,6 +169,7 @@ func (s *Server) handleAPIFleetUpdateStart(w stdhttp.ResponseWriter, r *stdhttp.
 		"fleet_update_id": fuID,
 		"target_version":  target,
 		"host_count":      len(hostIDs),
+		"canary_first":    body.CanaryFirst,
 	})
 	_ = s.deps.Store.AppendAudit(r.Context(), store.AuditEntry{
 		ID: ulid.Make().String(), UserID: &user.ID, Actor: "user",
@@ -209,6 +250,7 @@ func (s *Server) handleAPIFleetUpdateGet(w stdhttp.ResponseWriter, r *stdhttp.Re
 		Status:          fu.Status,
 		CurrentHostID:   fu.CurrentHostID,
 		HaltedReason:    fu.HaltedReason,
+		CanaryFirst:     fu.PauseAfterFirst,
 		Hosts:           make([]fleetUpdateHostView, 0, len(hosts)),
 	}
 	if fu.CompletedAt != nil {
@@ -286,6 +328,27 @@ func (s *Server) buildFleetUpdatePage(r *stdhttp.Request) (fleetUpdatePage, erro
 	}
 	for _, h := range hosts {
 		page.HostNames[h.ID] = h.Name
+		candidate := fleetUpdateCandidate{Host: h}
+		switch {
+		case h.AgentVersion == "":
+			candidate.Reason = "version unknown"
+		case h.AgentVersion == page.TargetVersion:
+			candidate.Reason = "already current"
+		case s.deps.Hub == nil || !s.deps.Hub.Connected(h.ID):
+			candidate.Reason = "offline"
+		default:
+			updating, uerr := s.deps.Store.RunningUpdateJobForHost(r.Context(), h.ID)
+			if uerr != nil {
+				return page, uerr
+			}
+			if updating != "" {
+				candidate.Reason = "update already running"
+			} else {
+				candidate.Eligible = true
+				page.OutOfDateHosts = append(page.OutOfDateHosts, h)
+			}
+		}
+		page.Candidates = append(page.Candidates, candidate)
 	}
 
 	active, err := s.deps.Store.ActiveFleetUpdate(r.Context())
@@ -328,20 +391,54 @@ func (s *Server) buildFleetUpdatePage(r *stdhttp.Request) (fleetUpdatePage, erro
 		}
 	}
 
-	// Idle list (or "still out of date" reference even when an active
-	// roll is running — cheap to compute, harmless to attach).
-	for _, h := range hosts {
-		if h.Status != "online" {
-			continue
-		}
-		if h.AgentVersion == "" || h.AgentVersion == page.TargetVersion {
-			continue
-		}
-		page.OutOfDateHosts = append(page.OutOfDateHosts, h)
-	}
 	return page, nil
 }
 
+// validateFleetUpdateHostIDs deduplicates an explicit selection while
+// preserving review order and rejects the entire request if membership has
+// changed or any selected host is not dispatchable.
+func (s *Server) validateFleetUpdateHostIDs(ctx context.Context, target string, requested []string) ([]string, map[string]string, error) {
+	hosts, err := s.deps.Store.ListHosts(ctx)
+	if err != nil {
+		return nil, nil, err
+	}
+	byID := make(map[string]store.Host, len(hosts))
+	for _, h := range hosts {
+		byID[h.ID] = h
+	}
+	seen := map[string]bool{}
+	validated := make([]string, 0, len(requested))
+	reasons := map[string]string{}
+	for _, id := range requested {
+		if seen[id] {
+			continue
+		}
+		seen[id] = true
+		h, ok := byID[id]
+		switch {
+		case !ok:
+			reasons[id] = "host not found"
+		case h.AgentVersion == "":
+			reasons[id] = "agent version unknown"
+		case h.AgentVersion == target:
+			reasons[id] = "already at target version"
+		case s.deps.Hub == nil || !s.deps.Hub.Connected(id):
+			reasons[id] = "host offline"
+		default:
+			jobID, jerr := s.deps.Store.RunningUpdateJobForHost(ctx, id)
+			if jerr != nil {
+				return nil, nil, jerr
+			}
+			if jobID != "" {
+				reasons[id] = "update already in progress"
+			} else {
+				validated = append(validated, id)
+			}
+		}
+	}
+	return validated, reasons, nil
+}
+
 // deriveOutOfDateOnlineHostIDs returns the list of host IDs that
 // (a) are online (Hub.Connected) and (b) have an agent_version that's
 // non-empty AND != target. Used by the start endpoint when the caller

internal/server/http/fleet_update_test.go

@@ -6,7 +6,9 @@ import (
 	"bytes"
 	"context"
 	"encoding/json"
+	"io"
 	stdhttp "net/http"
+	"strings"
 	"sync"
 	"testing"
 	"time"
@@ -19,6 +21,67 @@ import (
 	"gitea.dcglab.co.uk/steve/restic-manager/internal/version"
 )
 
+func TestFleetUpdateIdlePageDoesNotPollOrRequireTypedConfirmation(t *testing.T) {
+	t.Parallel()
+	_, baseURL, st := newTestServerWithUI(t)
+	cookie := loginAsAdmin(t, st)
+	_ = makeHost(t, st, "fleet-filter-host")
+
+	req, _ := stdhttp.NewRequest("GET", baseURL+"/settings/fleet-update", nil)
+	req.AddCookie(cookie)
+	res, err := stdhttp.DefaultClient.Do(req)
+	if err != nil {
+		t.Fatalf("get fleet update page: %v", err)
+	}
+	defer res.Body.Close()
+	if res.StatusCode != stdhttp.StatusOK {
+		t.Fatalf("status: got %d, want 200", res.StatusCode)
+	}
+	raw, _ := io.ReadAll(res.Body)
+	body := string(raw)
+
+	for _, want := range []string{
+		`id="fleet-filter-name"`,
+		`id="fleet-filter-version"`,
+		`id="fleet-filter-state"`,
+		"r.style.display=",
+	} {
+		if !strings.Contains(body, want) {
+			t.Errorf("page missing %q", want)
+		}
+	}
+	if strings.Contains(body, "Type selected count to confirm") || strings.Contains(body, `id="fleet-update-confirm"`) {
+		t.Error("page still renders the typed-count confirmation")
+	}
+	if strings.Contains(body, `hx-trigger="every 3s`) {
+		t.Error("idle selection page must not poll and replace operator input")
+	}
+}
+
+func TestFleetUpdateRunningPagePollsForProgress(t *testing.T) {
+	t.Parallel()
+	_, baseURL, st := newTestServerWithUI(t)
+	cookie, userID := loginAsAdminWithID(t, st)
+	hostID := makeHost(t, st, "fleet-running-host")
+	if err := st.CreateFleetUpdate(context.Background(), store.FleetUpdate{
+		ID: ulid.Make().String(), StartedByUserID: userID, TargetVersion: version.Version,
+	}, []string{hostID}); err != nil {
+		t.Fatalf("create fleet update: %v", err)
+	}
+
+	req, _ := stdhttp.NewRequest("GET", baseURL+"/settings/fleet-update", nil)
+	req.AddCookie(cookie)
+	res, err := stdhttp.DefaultClient.Do(req)
+	if err != nil {
+		t.Fatalf("get fleet update page: %v", err)
+	}
+	defer res.Body.Close()
+	raw, _ := io.ReadAll(res.Body)
+	if body := string(raw); !strings.Contains(body, `hx-trigger="every 3s`) {
+		t.Error("running rollout page must poll for progress")
+	}
+}
+
 // fakeFleetWorker stands in for *fleetupdate.Worker in HTTP tests.
 // It records what was passed to Start/Cancel and lets tests inject
 // canned errors. Satisfies the FleetWorker interface in
@@ -50,6 +113,10 @@ func (f *fakeFleetWorker) Start(_ context.Context, userID, target string, hostID
 	return f.startID, nil
 }
 
+func (f *fakeFleetWorker) StartCanary(ctx context.Context, userID, target string, hostIDs []string) (string, error) {
+	return f.Start(ctx, userID, target, hostIDs)
+}
+
 func (f *fakeFleetWorker) Cancel(_ context.Context, id string) error {
 	f.mu.Lock()
 	defer f.mu.Unlock()
@@ -190,6 +257,67 @@ func TestFleetUpdateStartDerivesHostIDsWhenEmpty(t *testing.T) {
 	}
 }
 
+func TestFleetUpdateStartDeduplicatesExplicitSelection(t *testing.T) {
+	t.Parallel()
+	srv, ts, st := rawTestServer(t)
+	worker := &fakeFleetWorker{startID: ulid.Make().String()}
+	srv.deps.FleetWorker = worker
+	cookie := loginAsAdmin(t, st)
+	hostID := helloOnlineHost(t, srv, st, "duplicate-host", "v0")
+	raw, _ := json.Marshal(map[string]any{"host_ids": []string{hostID, hostID}})
+	req, _ := stdhttp.NewRequest("POST", ts.URL+"/api/fleet/update", bytes.NewReader(raw))
+	req.AddCookie(cookie)
+	req.Header.Set("Content-Type", "application/json")
+	res, err := stdhttp.DefaultClient.Do(req)
+	if err != nil {
+		t.Fatalf("do: %v", err)
+	}
+	defer res.Body.Close()
+	if res.StatusCode != stdhttp.StatusAccepted {
+		t.Fatalf("status: %d", res.StatusCode)
+	}
+	if got := worker.startCalls[0].HostIDs; len(got) != 1 || got[0] != hostID {
+		t.Fatalf("deduplicated ids: %v", got)
+	}
+}
+
+func TestFleetUpdateStartRejectsUnknownAndIneligibleHosts(t *testing.T) {
+	t.Parallel()
+	srv, ts, st := rawTestServer(t)
+	worker := &fakeFleetWorker{startID: ulid.Make().String()}
+	srv.deps.FleetWorker = worker
+	cookie := loginAsAdmin(t, st)
+	offline := makeHost(t, st, "offline-host")
+	if err := st.MarkHostHello(context.Background(), offline, "v0", "0.17", api.CurrentProtocolVersion, time.Now().UTC()); err != nil {
+		t.Fatalf("mark offline host: %v", err)
+	}
+	raw, _ := json.Marshal(map[string]any{"host_ids": []string{offline, "does-not-exist"}})
+	req, _ := stdhttp.NewRequest("POST", ts.URL+"/api/fleet/update", bytes.NewReader(raw))
+	req.AddCookie(cookie)
+	req.Header.Set("Content-Type", "application/json")
+	res, err := stdhttp.DefaultClient.Do(req)
+	if err != nil {
+		t.Fatalf("do: %v", err)
+	}
+	defer res.Body.Close()
+	if res.StatusCode != stdhttp.StatusUnprocessableEntity {
+		t.Fatalf("status: got %d, want 422", res.StatusCode)
+	}
+	var body struct {
+		Code    string            `json:"code"`
+		Reasons map[string]string `json:"reasons"`
+	}
+	if err := json.NewDecoder(res.Body).Decode(&body); err != nil {
+		t.Fatalf("decode: %v", err)
+	}
+	if body.Code != "ineligible_hosts" || len(body.Reasons) != 2 {
+		t.Fatalf("structured reasons: %+v", body)
+	}
+	if len(worker.startCalls) != 0 {
+		t.Fatal("worker must not start with invalid membership")
+	}
+}
+
 func TestFleetUpdateCancelHappyPath(t *testing.T) {
 	t.Parallel()
 	srv, ts, st := rawTestServer(t)

internal/store/fleet_updates.go

@@ -42,9 +42,9 @@ func (st *Store) CreateFleetUpdate(ctx context.Context, fu FleetUpdate, hostIDs
 	}
 
 	if _, err := tx.ExecContext(ctx,
-		`INSERT INTO fleet_updates (id, started_at, started_by_user_id, target_version, status)
-		 VALUES (?, ?, ?, ?, ?)`,
-		fu.ID, fu.StartedAt.UTC().Format(time.RFC3339Nano), fu.StartedByUserID, fu.TargetVersion, fu.Status,
+		`INSERT INTO fleet_updates (id, started_at, started_by_user_id, target_version, status, pause_after_first)
+		 VALUES (?, ?, ?, ?, ?, ?)`,
+		fu.ID, fu.StartedAt.UTC().Format(time.RFC3339Nano), fu.StartedByUserID, fu.TargetVersion, fu.Status, fu.PauseAfterFirst,
 	); err != nil {
 		return fmt.Errorf("store: insert fleet_updates: %w", err)
 	}
@@ -67,12 +67,13 @@ func (st *Store) ActiveFleetUpdate(ctx context.Context) (*FleetUpdate, error) {
 	var current sql.NullString
 	var halted sql.NullString
 	var completedAt sql.NullString
+	var pauseAfterFirst int
 	err := st.db.QueryRowContext(ctx,
 		`SELECT id, started_at, started_by_user_id, target_version, status,
-		   current_host_id, halted_reason, completed_at
+		   current_host_id, halted_reason, completed_at, pause_after_first
 		 FROM fleet_updates WHERE status = 'running' LIMIT 1`).
 		Scan(&fu.ID, &startedAt, &fu.StartedByUserID, &fu.TargetVersion, &fu.Status,
-			&current, &halted, &completedAt)
+			&current, &halted, &completedAt, &pauseAfterFirst)
 	if errors.Is(err, sql.ErrNoRows) {
 		return nil, nil
 	}
@@ -82,6 +83,7 @@ func (st *Store) ActiveFleetUpdate(ctx context.Context) (*FleetUpdate, error) {
 	fu.StartedAt, _ = time.Parse(time.RFC3339Nano, startedAt)
 	fu.CurrentHostID = current.String
 	fu.HaltedReason = halted.String
+	fu.PauseAfterFirst = pauseAfterFirst != 0
 	if completedAt.Valid {
 		t, _ := time.Parse(time.RFC3339Nano, completedAt.String)
 		fu.CompletedAt = &t
@@ -97,12 +99,13 @@ func (st *Store) GetFleetUpdate(ctx context.Context, id string) (*FleetUpdate, [
 	var current sql.NullString
 	var halted sql.NullString
 	var completedAt sql.NullString
+	var pauseAfterFirst int
 	err := st.db.QueryRowContext(ctx,
 		`SELECT id, started_at, started_by_user_id, target_version, status,
-		   current_host_id, halted_reason, completed_at
+		   current_host_id, halted_reason, completed_at, pause_after_first
 		 FROM fleet_updates WHERE id = ?`, id).
 		Scan(&fu.ID, &startedAt, &fu.StartedByUserID, &fu.TargetVersion, &fu.Status,
-			&current, &halted, &completedAt)
+			&current, &halted, &completedAt, &pauseAfterFirst)
 	if errors.Is(err, sql.ErrNoRows) {
 		return nil, nil, ErrNotFound
 	}
@@ -112,6 +115,7 @@ func (st *Store) GetFleetUpdate(ctx context.Context, id string) (*FleetUpdate, [
 	fu.StartedAt, _ = time.Parse(time.RFC3339Nano, startedAt)
 	fu.CurrentHostID = current.String
 	fu.HaltedReason = halted.String
+	fu.PauseAfterFirst = pauseAfterFirst != 0
 	if completedAt.Valid {
 		t, _ := time.Parse(time.RFC3339Nano, completedAt.String)
 		fu.CompletedAt = &t

internal/store/migrations/0026_fleet_update_canary.sql

@@ -0,0 +1 @@
+ALTER TABLE fleet_updates ADD COLUMN pause_after_first INTEGER NOT NULL DEFAULT 0;

internal/store/types.go

@@ -231,6 +231,7 @@ type FleetUpdate struct {
 	CurrentHostID   string
 	HaltedReason    string
 	CompletedAt     *time.Time
+	PauseAfterFirst bool
 }
 
 // FleetUpdateHost is one host's slot in a fleet update. Position is

web/templates/pages/dashboard.html

@@ -69,7 +69,7 @@
     {{/* ---------- Hosts-behind hero tile (P6-18) ---------- */}}
     {{if gt $page.UpdatesBehind 0}}
       <div class="pt-4">
-        <a href="?updates=behind" class="hero-tile hero-tile--amber" style="display:inline-flex;">
+		<a href="/settings/fleet-update" class="hero-tile hero-tile--amber" style="display:inline-flex;">
           <span class="hero-num">{{$page.UpdatesBehind}}</span>
           <span class="hero-label">{{if eq $page.UpdatesBehind 1}}host behind{{else}}hosts behind{{end}} · review →</span>
         </a>

web/templates/pages/fleet_update.html

@@ -8,18 +8,18 @@
   <div class="crumbs pt-6">
     <a href="/">Dashboard</a><span class="sep">/</span>
     <a href="/settings">Settings</a><span class="sep">/</span>
-    <span class="text-ink-mid">fleet update</span>
+	<span class="text-ink-mid">agent updates</span>
   </div>
 
   {{/* page header */}}
   <div class="flex items-baseline justify-between mt-3.5">
     <div>
       <h1 class="text-[22px] font-medium tracking-[-0.005em]">
-        Fleet update
+		Agent updates
         <span class="text-ink-fade font-normal text-[14px] ml-2 mono">target {{$page.TargetVersion}}</span>
       </h1>
       <p class="text-ink-mute text-[12px] mt-1 max-w-[760px] leading-[1.55]">
-        Rolling, sequential agent self-update. One host at a time, halts on first failure,
+		Rolling, sequential agent self-update. One host at a time, halts on first failure,
         cancellable mid-roll. Only online hosts whose <span class="mono">agent_version</span>
         differs from the server are eligible.
       </p>
@@ -28,5 +28,15 @@
 
   {{template "fleet_update_inner" .}}
 
+  <script>
+    function fleetBoxes(){return Array.from(document.querySelectorAll('.fleet-host:not(:disabled)'))}
+    function fleetReview(){const n=fleetBoxes().filter(x=>x.checked).length;document.getElementById('fleet-selected-count').textContent=n;document.getElementById('fleet-timeout').textContent=n+' × 95s';document.getElementById('fleet-update-start-btn').disabled=n===0}
+    function fleetFilter(){const q=document.getElementById('fleet-filter-name').value.toLowerCase(),v=document.getElementById('fleet-filter-version').value.toLowerCase(),s=document.getElementById('fleet-filter-state').value;document.querySelectorAll('.fleet-candidate').forEach(r=>r.style.display=r.dataset.name.toLowerCase().includes(q)&&r.dataset.version.toLowerCase().includes(v)&&(s==='all'||r.dataset.state===s)?'':'none')}
+    function fleetSelectVisible(on){document.querySelectorAll('.fleet-candidate').forEach(r=>{const x=r.querySelector('.fleet-host:not(:disabled)');if(x&&r.style.display!=='none')x.checked=on});fleetReview()}
+    function fleetSelectOnly(id){fleetBoxes().forEach(x=>x.checked=x.value===id);fleetReview()}
+    async function fleetStart(e){e.preventDefault();const ids=fleetBoxes().filter(x=>x.checked).map(x=>x.value),out=document.getElementById('fleet-start-error');out.textContent='';const res=await fetch('/api/fleet/update',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({host_ids:ids,canary_first:document.getElementById('fleet-canary').checked})});if(res.ok){location.reload();return false}const body=await res.json();out.textContent=body.message||body.code||'Unable to start';return false}
+    if(document.getElementById('fleet-selected-count'))fleetReview()
+  </script>
+
 </div>
 {{end}}

web/templates/pages/settings.html

@@ -40,6 +40,7 @@
       {{if not $page.Form}}<span class="mono text-ink-fade text-[11px] ml-1">{{len $page.Channels}}</span>{{end}}
     </a>
     <a href="/settings/users" class="sub-tab {{if eq $page.ActiveTab "users"}}active{{end}}">Users</a>
+	<a href="/settings/fleet-update" class="sub-tab">Agent updates</a>
     <span class="sub-tab text-ink-fade cursor-default" title="lands later">Authentication</span>
   </div>
 

web/templates/partials/fleet_update_inner.html

@@ -7,11 +7,11 @@
 */}}
 {{define "fleet_update_inner"}}
 {{$page := .Page}}
-<div id="fleet-update-panel" class="mt-5"
+<div id="fleet-update-panel" class="mt-5"{{if and $page.Active (eq $page.Active.Status "running")}}
      hx-get="{{$page.PollURL}}"
      hx-trigger="every 3s [document.visibilityState==='visible']"
      hx-select="#fleet-update-panel"
-     hx-swap="outerHTML">
+     hx-swap="outerHTML"{{end}}>
 
 {{if and $page.Active (eq $page.Active.Status "running")}}
 
@@ -32,6 +32,7 @@
     </div>
     <div class="text-[11.5px] text-ink-mute mt-1">
       target <span class="mono text-ink-mid">{{$page.Active.TargetVersion}}</span>
+	  {{if $page.Active.PauseAfterFirst}} · canary-first{{end}}
       · started <span class="mono text-ink-mid">{{relTime $page.Active.StartedAt}}</span>
       {{if $page.Active.CurrentHostID}}
         · waiting on <span class="mono text-ink-mid">{{index $page.HostNames $page.Active.CurrentHostID}}</span>
@@ -68,6 +69,12 @@
     {{if $page.Active.HaltedReason}}
       <div class="text-[12px] text-bad mt-2">{{$page.Active.HaltedReason}}</div>
     {{end}}
+	{{if eq $page.Active.Status "halted"}}
+	  <div class="mt-3 flex gap-2 text-[12px]">
+		<a class="btn" href="#fleet-update-start-form">Resume remaining as new rollout</a>
+		{{range $page.ActiveRows}}{{if eq .Status "failed"}}<button class="btn" type="button" onclick="fleetSelectOnly('{{.HostID}}');document.getElementById('fleet-update-start-form').scrollIntoView()">Retry {{.HostName}}</button>{{end}}{{end}}
+	  </div>
+	{{end}}
   </div>
 
   {{template "fleet_update_rows" $page}}
@@ -122,11 +129,11 @@
 {{define "fleet_update_idle_panel"}}
 {{$page := .}}
 <div class="panel rounded-[7px] px-5 py-4">
-  {{if eq (len $page.OutOfDateHosts) 0}}
+  {{if eq (len $page.Candidates) 0}}
     <div class="flex items-center gap-3">
       <span class="dot dot-online"></span>
       <div>
-        <div class="text-ink text-[14px] font-medium">All hosts are up to date.</div>
+        <div class="text-ink text-[14px] font-medium">No hosts enrolled.</div>
         <div class="text-ink-mute text-[12px] mt-0.5">
           Every online agent matches server version <span class="mono">{{$page.TargetVersion}}</span>.
         </div>
@@ -137,34 +144,36 @@
       <h2 class="text-[14px] font-medium">{{len $page.OutOfDateHosts}} host{{if ne (len $page.OutOfDateHosts) 1}}s{{end}} out of date</h2>
       <span class="mono text-[11px] text-ink-fade">target {{$page.TargetVersion}}</span>
     </div>
-    <ul class="mt-3 space-y-1 text-[12px]">
-      {{range $page.OutOfDateHosts}}
-        <li class="flex items-center gap-3">
-          <span class="dot dot-online"></span>
-          <span class="mono text-ink">{{.Name}}</span>
-          <span class="mono text-ink-mute">{{if .AgentVersion}}{{.AgentVersion}}{{else}}—{{end}} → {{$page.TargetVersion}}</span>
-        </li>
-      {{end}}
-    </ul>
-
-    <form id="fleet-update-start-form" class="mt-4 flex items-center gap-3"
-          hx-post="/api/fleet/update"
-          hx-headers='{"Content-Type":"application/json"}'
-          hx-vals='{}'
-          hx-swap="none"
-          hx-on::after-request="if(event.detail.successful) location.reload()">
-      <label class="text-[11.5px] text-ink-mute">
-        Type the count
-        <span class="mono text-ink-mid">({{len $page.OutOfDateHosts}})</span>
-        to enable Start:
+    <div class="mt-3 flex flex-wrap gap-2">
+      <input id="fleet-filter-name" class="field text-[12px]" placeholder="Filter name or tag" oninput="fleetFilter()">
+      <input id="fleet-filter-version" class="field text-[12px] mono" placeholder="Version" oninput="fleetFilter()">
+      <select id="fleet-filter-state" class="field text-[12px]" onchange="fleetFilter()">
+        <option value="all">All states</option><option value="eligible">Eligible</option><option value="excluded">Excluded</option>
+      </select>
+      <button type="button" class="btn" onclick="fleetSelectVisible(true)">Select visible eligible</button>
+      <button type="button" class="btn" onclick="fleetSelectVisible(false)">Clear selection</button>
+    </div>
+    <div class="panel mt-3 rounded-[7px] overflow-hidden">
+      {{range $page.Candidates}}
+      <label class="fleet-candidate grid items-center px-3 py-2 hairline text-[12px]"
+             data-name="{{.Host.Name}} {{range .Host.Tags}}{{.}} {{end}}" data-version="{{.Host.AgentVersion}}" data-state="{{if .Eligible}}eligible{{else}}excluded{{end}}"
+             style="grid-template-columns: 28px 1.4fr .8fr .8fr 1.2fr;gap:12px">
+        <input class="fleet-host" type="checkbox" value="{{.Host.ID}}" {{if .Eligible}}checked onchange="fleetReview()"{{else}}disabled{{end}}>
+        <span class="mono">{{.Host.Name}}</span>
+        <span class="mono text-ink-mute">{{if .Host.AgentVersion}}{{.Host.AgentVersion}}{{else}}unknown{{end}}</span>
+        <span class="mono text-ink-mute">{{$page.TargetVersion}}</span>
+        <span class="text-ink-mute">{{if .Eligible}}eligible{{else}}{{.Reason}}{{end}}</span>
       </label>
-      <input type="text" id="fleet-update-confirm" class="field mono text-[12.5px]"
-             style="width: 80px; padding: 5px 8px;"
-             oninput="document.getElementById('fleet-update-start-btn').disabled = (this.value !== '{{len $page.OutOfDateHosts}}');"
-             autocomplete="off" />
-      <button type="submit" id="fleet-update-start-btn" class="btn btn-amber" disabled>
-        Start fleet update
-      </button>
+      {{end}}
+    </div>
+    <div class="mt-4 text-[12px] text-ink-mute">
+      <span id="fleet-selected-count">{{len $page.OutOfDateHosts}}</span> selected · sequential, halts on first failure · worst-case
+      <span id="fleet-timeout" class="mono">{{len $page.OutOfDateHosts}} × 95s</span>
+    </div>
+    <form class="mt-3 flex items-center gap-3" onsubmit="return fleetStart(event)">
+	  <label class="text-[11.5px] text-ink-mute"><input id="fleet-canary" type="checkbox" checked> Pause after first host for canary review</label>
+      <button id="fleet-update-start-btn" class="btn btn-amber">Start agent update</button>
+      <span id="fleet-start-error" class="text-bad text-[12px]"></span>
     </form>
   {{end}}
 </div>