Skip to content

armgit / Pulse

Public Imported from GitHub
0

Real-time monitoring dashboard for Proxmox VE, PBS, Docker, Kubernetes, TrueNAS and vSphere. Self-hosted, with alerts for failed backups, full disks, unhealthy containers and offline nodes.

96.5 KB
Raw Blame
package monitoring

import (
	"context"
	"crypto/sha256"
	"encoding/json"
	"errors"
	"fmt"
	"math"
	"net/http"
	"net/http/httptest"
	"strconv"
	"strings"
	"sync/atomic"
	"testing"
	"time"

	"github.com/gorilla/websocket"
	"github.com/rcourtman/pulse-go-rewrite/internal/alerts"
	"github.com/rcourtman/pulse-go-rewrite/internal/config"
	"github.com/rcourtman/pulse-go-rewrite/internal/mock"
	"github.com/rcourtman/pulse-go-rewrite/internal/models"
	"github.com/rcourtman/pulse-go-rewrite/internal/storagehealth"
	"github.com/rcourtman/pulse-go-rewrite/internal/truenas"
	"github.com/rcourtman/pulse-go-rewrite/internal/unifiedresources"
	"github.com/rcourtman/pulse-go-rewrite/internal/vmware"
	"github.com/rcourtman/pulse-go-rewrite/pkg/metrics"
	"github.com/rcourtman/pulse-go-rewrite/pkg/pbs"
	"github.com/rcourtman/pulse-go-rewrite/pkg/pmg"
)

type testPollProvider struct {
	providerType      InstanceType
	instances         []string
	describeInstances []PollProviderInstanceInfo
	connectionStatus  map[string]bool
	connectionKey     string
	interval          time.Duration
	buildPollTask     func(instanceName string) (PollTask, error)
}

func (p testPollProvider) Type() InstanceType { return p.providerType }

func (p testPollProvider) ListInstances(_ *Monitor) []string {
	out := make([]string, len(p.instances))
	copy(out, p.instances)
	return out
}

func (p testPollProvider) DescribeInstances(_ *Monitor) []PollProviderInstanceInfo {
	out := make([]PollProviderInstanceInfo, len(p.describeInstances))
	for i := range p.describeInstances {
		out[i] = PollProviderInstanceInfo{
			Name:        p.describeInstances[i].Name,
			DisplayName: p.describeInstances[i].DisplayName,
			Connection:  p.describeInstances[i].Connection,
			Metadata:    cloneProviderMetadata(p.describeInstances[i].Metadata),
		}
	}
	return out
}

func (p testPollProvider) ConnectionStatuses(_ *Monitor) map[string]bool {
	if len(p.connectionStatus) == 0 {
		return nil
	}
	out := make(map[string]bool, len(p.connectionStatus))
	for key, healthy := range p.connectionStatus {
		out[key] = healthy
	}
	return out
}

func (p testPollProvider) ConnectionHealthKey(_ *Monitor, instanceName string) string {
	if strings.TrimSpace(p.connectionKey) != "" {
		return strings.TrimSpace(p.connectionKey)
	}
	return ""
}

func (p testPollProvider) BaseInterval(_ *Monitor) time.Duration { return p.interval }

func (p testPollProvider) BuildPollTask(_ *Monitor, instanceName string) (PollTask, error) {
	if p.buildPollTask == nil {
		return PollTask{
			InstanceName: instanceName,
			InstanceType: string(p.providerType),
		}, nil
	}
	return p.buildPollTask(instanceName)
}

type testSupplementalPollProvider struct {
	testPollProvider
	source           unifiedresources.DataSource
	ownedSources     []unifiedresources.DataSource
	recordsByOrg     map[string][]unifiedresources.IngestRecord
	changesByOrg     map[string][]unifiedresources.ResourceChange
	lastRequestedOrg string
}

func (p *testSupplementalPollProvider) SupplementalSource() unifiedresources.DataSource {
	return p.source
}

func (p *testSupplementalPollProvider) SupplementalRecords(_ *Monitor, orgID string) []unifiedresources.IngestRecord {
	p.lastRequestedOrg = orgID
	records := p.recordsByOrg[orgID]
	out := make([]unifiedresources.IngestRecord, len(records))
	copy(out, records)
	return out
}

func (p *testSupplementalPollProvider) SupplementalChanges(_ *Monitor, orgID string) []unifiedresources.ResourceChange {
	p.lastRequestedOrg = orgID
	return cloneTestResourceChanges(p.changesByOrg[orgID])
}

func (p *testSupplementalPollProvider) SnapshotOwnedSources(_ *Monitor) []unifiedresources.DataSource {
	out := make([]unifiedresources.DataSource, len(p.ownedSources))
	copy(out, p.ownedSources)
	return out
}

type testMonitorSupplementalProvider struct {
	recordsByOrg     map[string][]unifiedresources.IngestRecord
	changesByOrg     map[string][]unifiedresources.ResourceChange
	ownedSources     []unifiedresources.DataSource
	lastRequestedOrg string
}

func ptrFloat64(value float64) *float64 {
	return &value
}

func ptrInt64(value int64) *int64 {
	return &value
}

func cloneTestResourceChanges(in []unifiedresources.ResourceChange) []unifiedresources.ResourceChange {
	if in == nil {
		return nil
	}
	out := make([]unifiedresources.ResourceChange, len(in))
	for i := range in {
		out[i] = in[i]
		if in[i].OccurredAt != nil {
			occurredAt := in[i].OccurredAt.UTC()
			out[i].OccurredAt = &occurredAt
		}
		if in[i].RelatedResources != nil {
			out[i].RelatedResources = append([]string(nil), in[i].RelatedResources...)
		}
		if in[i].Metadata != nil {
			out[i].Metadata = make(map[string]any, len(in[i].Metadata))
			for key, value := range in[i].Metadata {
				out[i].Metadata[key] = value
			}
		}
	}
	return out
}

func (p *testMonitorSupplementalProvider) SupplementalRecords(_ *Monitor, orgID string) []unifiedresources.IngestRecord {
	p.lastRequestedOrg = orgID
	records := p.recordsByOrg[orgID]
	out := make([]unifiedresources.IngestRecord, len(records))
	copy(out, records)
	return out
}

func (p *testMonitorSupplementalProvider) SupplementalChanges(_ *Monitor, orgID string) []unifiedresources.ResourceChange {
	p.lastRequestedOrg = orgID
	return cloneTestResourceChanges(p.changesByOrg[orgID])
}

func (p *testMonitorSupplementalProvider) SnapshotOwnedSources() []unifiedresources.DataSource {
	out := make([]unifiedresources.DataSource, len(p.ownedSources))
	copy(out, p.ownedSources)
	return out
}

type testSupplementalResourceStore struct {
	snapshotCalls   int
	lastSnapshot    models.StateSnapshot
	recordsBySource map[unifiedresources.DataSource][]unifiedresources.IngestRecord
	recordedChanges []unifiedresources.ResourceChange
}

func (s *testSupplementalResourceStore) ShouldSkipAPIPolling(string) bool { return false }

func (s *testSupplementalResourceStore) GetPollingRecommendations() map[string]float64 { return nil }

func (s *testSupplementalResourceStore) GetAll() []unifiedresources.Resource { return nil }

func (s *testSupplementalResourceStore) PopulateFromSnapshot(snapshot models.StateSnapshot) {
	s.lastSnapshot = snapshot
	s.snapshotCalls++
}

func (s *testSupplementalResourceStore) PopulateSupplementalRecords(source unifiedresources.DataSource, records []unifiedresources.IngestRecord) {
	if s.recordsBySource == nil {
		s.recordsBySource = make(map[unifiedresources.DataSource][]unifiedresources.IngestRecord)
	}
	cloned := make([]unifiedresources.IngestRecord, len(records))
	copy(cloned, records)
	s.recordsBySource[source] = append(s.recordsBySource[source], cloned...)
}

func (s *testSupplementalResourceStore) RecordChange(change unifiedresources.ResourceChange) error {
	s.recordedChanges = append(s.recordedChanges, cloneTestResourceChanges([]unifiedresources.ResourceChange{change})...)
	return nil
}

type testAtomicResourceStore struct {
	snapshotCalls    int
	atomicCalls      int
	lastSnapshot     models.StateSnapshot
	lastRecordsBySrc map[unifiedresources.DataSource][]unifiedresources.IngestRecord
}

func (s *testAtomicResourceStore) ShouldSkipAPIPolling(string) bool { return false }

func (s *testAtomicResourceStore) GetPollingRecommendations() map[string]float64 { return nil }

func (s *testAtomicResourceStore) GetAll() []unifiedresources.Resource { return nil }

func (s *testAtomicResourceStore) PopulateFromSnapshot(snapshot models.StateSnapshot) {
	s.lastSnapshot = snapshot
	s.snapshotCalls++
}

func (s *testAtomicResourceStore) PopulateSnapshotAndSupplemental(snapshot models.StateSnapshot, recordsBySource map[unifiedresources.DataSource][]unifiedresources.IngestRecord) {
	s.lastSnapshot = snapshot
	s.atomicCalls++
	s.lastRecordsBySrc = make(map[unifiedresources.DataSource][]unifiedresources.IngestRecord, len(recordsBySource))
	for source, records := range recordsBySource {
		cloned := make([]unifiedresources.IngestRecord, len(records))
		copy(cloned, records)
		s.lastRecordsBySrc[source] = cloned
	}
}

func TestBuildScheduledTasksUsesConfiguredIntervals(t *testing.T) {
	now := time.Now()
	cfg := &config.Config{
		PVEPollingInterval:          2 * time.Minute,
		PBSPollingInterval:          45 * time.Second,
		PMGPollingInterval:          90 * time.Second,
		AdaptivePollingBaseInterval: 10 * time.Second,
	}

	monitor := &Monitor{
		config:     cfg,
		pveClients: map[string]PVEClientInterface{"pve-1": nil},
		pbsClients: map[string]*pbs.Client{"pbs-1": nil},
		pmgClients: map[string]*pmg.Client{"pmg-1": nil},
	}

	tasks := monitor.buildScheduledTasks(now)
	if len(tasks) != 3 {
		t.Fatalf("expected 3 tasks, got %d", len(tasks))
	}

	got := map[InstanceType]time.Duration{}
	for _, task := range tasks {
		if !task.NextRun.Equal(now) {
			t.Fatalf("expected NextRun to equal provided time, got %v", task.NextRun)
		}
		got[task.InstanceType] = task.Interval
	}

	if got[InstanceTypePVE] != cfg.PVEPollingInterval {
		t.Fatalf("expected PVE interval %v, got %v", cfg.PVEPollingInterval, got[InstanceTypePVE])
	}
	if got[InstanceTypePBS] != cfg.PBSPollingInterval {
		t.Fatalf("expected PBS interval %v, got %v", cfg.PBSPollingInterval, got[InstanceTypePBS])
	}
	if got[InstanceTypePMG] != cfg.PMGPollingInterval {
		t.Fatalf("expected PMG interval %v, got %v", cfg.PMGPollingInterval, got[InstanceTypePMG])
	}
}

func TestRescheduleTaskUsesInstanceIntervalWhenSchedulerDisabled(t *testing.T) {
	cfg := &config.Config{
		PVEPollingInterval:          75 * time.Second,
		AdaptivePollingBaseInterval: 10 * time.Second,
	}

	monitor := &Monitor{
		config:    cfg,
		taskQueue: NewTaskQueue(),
	}

	task := ScheduledTask{
		InstanceName: "pve-1",
		InstanceType: InstanceTypePVE,
		Interval:     0,
		NextRun:      time.Now(),
	}

	monitor.rescheduleTask(task)

	monitor.taskQueue.mu.Lock()
	entry, ok := monitor.taskQueue.entries[schedulerKey(task.InstanceType, task.InstanceName)]
	monitor.taskQueue.mu.Unlock()
	if !ok {
		t.Fatalf("expected task to be rescheduled in queue")
	}

	if entry.task.Interval != cfg.PVEPollingInterval {
		t.Fatalf("expected interval %v, got %v", cfg.PVEPollingInterval, entry.task.Interval)
	}

	remaining := time.Until(entry.task.NextRun)
	if remaining < cfg.PVEPollingInterval-2*time.Second || remaining > cfg.PVEPollingInterval+time.Second {
		t.Fatalf("expected NextRun about %v from now, got %v", cfg.PVEPollingInterval, remaining)
	}
}

func TestUpdateResourceStoreSyncsUnifiedIncidentAlerts(t *testing.T) {
	alertManager := alerts.NewManagerWithDataDir(t.TempDir())
	defer alertManager.Stop()

	store := unifiedresources.NewMonitorAdapter(nil)
	monitor := &Monitor{
		state:         models.NewState(),
		resourceStore: store,
		alertManager:  alertManager,
		orgID:         "default",
		supplementalProviders: map[unifiedresources.DataSource]MonitorSupplementalRecordsProvider{
			unifiedresources.SourceTrueNAS: &testMonitorSupplementalProvider{
				recordsByOrg: map[string][]unifiedresources.IngestRecord{
					"default": {{
						SourceID: "pool:tank",
						Resource: unifiedresources.Resource{
							ID:         "storage:tank",
							Type:       unifiedresources.ResourceTypeStorage,
							Name:       "tank",
							ParentName: "truenas-main",
							Sources:    []unifiedresources.DataSource{unifiedresources.SourceTrueNAS},
							Storage: &unifiedresources.StorageMeta{
								Platform:   "truenas",
								Topology:   "pool",
								Protection: "zfs",
								IsZFS:      true,
							},
							Incidents: []unifiedresources.ResourceIncident{{
								Provider: "truenas",
								NativeID: "alert-1",
								Code:     "truenas_volume_status",
								Severity: storagehealth.RiskWarning,
								Summary:  "Pool tank is DEGRADED",
							}},
						},
					}},
				},
			},
		},
	}

	monitor.updateResourceStore(models.StateSnapshot{}, monitor.mockModeFence.begin())

	active := alertManager.GetActiveAlerts()
	if len(active) != 1 {
		t.Fatalf("expected 1 active alert, got %d", len(active))
	}
	if active[0].Type != "zfs-pool-state" {
		t.Fatalf("alert type = %q, want zfs-pool-state", active[0].Type)
	}

	snapshot := monitor.state.GetSnapshot()
	if len(snapshot.ActiveAlerts) != 1 {
		t.Fatalf("expected state snapshot to contain 1 active alert, got %d", len(snapshot.ActiveAlerts))
	}
}

func TestUpdateResourceStoreSyncsCanonicalStorageMetrics(t *testing.T) {
	t.Run("supplemental truenas storage", func(t *testing.T) {
		cfg := metrics.DefaultConfig(t.TempDir())
		store, err := metrics.NewStore(cfg)
		if err != nil {
			t.Fatalf("metrics.NewStore() error = %v", err)
		}
		defer func() { _ = store.Close() }()

		resourceStore := unifiedresources.NewMonitorAdapter(nil)
		used := int64(620)
		total := int64(1000)
		monitor := &Monitor{
			resourceStore:  resourceStore,
			metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
			metricsStore:   store,
			orgID:          "default",
			supplementalProviders: map[unifiedresources.DataSource]MonitorSupplementalRecordsProvider{
				unifiedresources.SourceTrueNAS: &testMonitorSupplementalProvider{
					recordsByOrg: map[string][]unifiedresources.IngestRecord{
						"default": {{
							SourceID: "pool:tank",
							Resource: unifiedresources.Resource{
								ID:        "storage:tank",
								Type:      unifiedresources.ResourceTypeStorage,
								Name:      "tank",
								Status:    unifiedresources.StatusOnline,
								LastSeen:  time.Now().UTC(),
								UpdatedAt: time.Now().UTC(),
								Sources:   []unifiedresources.DataSource{unifiedresources.SourceTrueNAS},
								Metrics: &unifiedresources.ResourceMetrics{
									Disk: &unifiedresources.MetricValue{Used: &used, Total: &total, Percent: 62},
								},
								Storage: &unifiedresources.StorageMeta{
									Platform:   "truenas",
									Topology:   "pool",
									Protection: "zfs",
								},
							},
						}},
					},
				},
			},
		}

		monitor.updateResourceStore(models.StateSnapshot{}, monitor.mockModeFence.begin())

		memory := monitor.GetStorageMetrics("pool:tank", time.Hour)
		if got := len(memory["usage"]); got == 0 {
			t.Fatalf("expected in-memory usage history for pool:tank")
		}

		storeBacked := monitor.GetStorageMetricsForChart("pool:tank", 7*24*time.Hour)
		if got := len(storeBacked["usage"]); got == 0 {
			t.Fatalf("expected persisted usage history for pool:tank")
		}
	})

	t.Run("agent unraid storage", func(t *testing.T) {
		cfg := metrics.DefaultConfig(t.TempDir())
		store, err := metrics.NewStore(cfg)
		if err != nil {
			t.Fatalf("metrics.NewStore() error = %v", err)
		}
		defer func() { _ = store.Close() }()

		resourceStore := unifiedresources.NewMonitorAdapter(nil)
		snapshot := models.StateSnapshot{
			Hosts: []models.Host{
				{
					ID:        "host-tower",
					Hostname:  "tower",
					Status:    "online",
					LastSeen:  time.Now().UTC(),
					MachineID: "machine-tower",
					Disks: []models.Disk{
						{Mountpoint: "/mnt/user", Total: 1000, Used: 400, Free: 600, Usage: 40},
					},
					Unraid: &models.HostUnraidStorage{
						ArrayStarted: true,
						ArrayState:   "STARTED",
						NumProtected: 1,
					},
				},
			},
		}
		monitor := &Monitor{
			resourceStore:  resourceStore,
			metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
			metricsStore:   store,
		}

		monitor.updateResourceStore(snapshot, monitor.mockModeFence.begin())

		var storageResourceID string
		for _, resource := range resourceStore.GetAll() {
			if resource.Type == unifiedresources.ResourceTypeStorage && resource.Storage != nil && resource.Storage.Platform == "unraid" {
				storageResourceID = resource.ID
				break
			}
		}
		if storageResourceID == "" {
			t.Fatal("expected unraid storage resource in unified store")
		}

		target := resourceStore.MetricsTargetForResource(storageResourceID)
		if target == nil || target.ResourceType != "storage" || target.ResourceID != "host-tower/storage:unraid-array" {
			t.Fatalf("unexpected unraid storage metrics target %+v", target)
		}

		memory := monitor.GetStorageMetrics(target.ResourceID, time.Hour)
		if got := len(memory["usage"]); got == 0 {
			t.Fatalf("expected in-memory usage history for %s", target.ResourceID)
		}

		storeBacked := monitor.GetStorageMetricsForChart(target.ResourceID, 7*24*time.Hour)
		if got := len(storeBacked["usage"]); got == 0 {
			t.Fatalf("expected persisted usage history for %s", target.ResourceID)
		}
	})
}

func TestBuildBroadcastFrontendStateIncludesUnifiedIncidentAlerts(t *testing.T) {
	alertManager := alerts.NewManagerWithDataDir(t.TempDir())
	defer alertManager.Stop()

	store := unifiedresources.NewMonitorAdapter(nil)
	monitor := &Monitor{
		state:         models.NewState(),
		resourceStore: store,
		alertManager:  alertManager,
		orgID:         "default",
		supplementalProviders: map[unifiedresources.DataSource]MonitorSupplementalRecordsProvider{
			unifiedresources.SourceTrueNAS: &testMonitorSupplementalProvider{
				recordsByOrg: map[string][]unifiedresources.IngestRecord{
					"default": {{
						SourceID: "system:truenas-main",
						Resource: unifiedresources.Resource{
							ID:      "agent:truenas-main",
							Type:    unifiedresources.ResourceTypeAgent,
							Name:    "truenas-main",
							Sources: []unifiedresources.DataSource{unifiedresources.SourceTrueNAS},
							TrueNAS: &unifiedresources.TrueNASData{Hostname: "truenas-main"},
							Incidents: []unifiedresources.ResourceIncident{{
								Provider: "truenas",
								NativeID: "alert-2",
								Code:     "truenas_volume_status",
								Severity: storagehealth.RiskCritical,
								Summary:  "Pool tank is FAULTED",
							}},
						},
					}},
				},
			},
		},
	}

	frontend := monitor.buildBroadcastFrontendStateFromSnapshot(models.StateSnapshot{}, monitor.mockModeFence.begin())

	if len(frontend.ActiveAlerts) != 1 {
		t.Fatalf("expected 1 active alert in frontend state, got %d", len(frontend.ActiveAlerts))
	}
	if frontend.ActiveAlerts[0].Type != "resource-incident" {
		t.Fatalf("alert type = %q, want resource-incident", frontend.ActiveAlerts[0].Type)
	}
}

// #2400: exercise the actual TLS/RPC -> GetApps -> APIFetcher -> provider
// poll -> canonical resources -> History path, not only numeric unmarshalling.
// Three short, explicitly due polls are a source regression, not a native
// sixty-second/Traefik soak. Only the first two of thirteen apps' CPU shapes
// come from the reporter; names and all remaining inventory are synthetic.
func TestTrueNASFractionalAppStatsPollAndHistory(t *testing.T) {
	previous := truenas.IsFeatureEnabled()
	truenas.SetFeatureEnabled(true)
	t.Cleanup(func() { truenas.SetFeatureEnabled(previous) })
	const appCount = 13
	const hostMemory = int64(8 * 1024 * 1024 * 1024)
	cpus := [][2]string{{"0.0286", "0.0"}, {"0.0017857142857142859", "0.0"}, {"17", "0"}}
	var poll atomic.Int32
	var sessions, logins, statsSubscriptions, realtimeSubscriptions, unsubscribes atomic.Int32
	upgrader := websocket.Upgrader{}
	server := httptest.NewTLSServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
		if r.URL.Path != "/api/current" || !websocket.IsWebSocketUpgrade(r) {
			t.Error("modern polling unexpectedly requested REST")
			http.NotFound(w, r)
			return
		}
		conn, err := upgrader.Upgrade(w, r, nil)
		if err != nil {
			t.Error(err)
			return
		}
		defer conn.Close()
		sessions.Add(1)
		for {
			var request struct {
				ID     int64             `json:"id"`
				Method string            `json:"method"`
				Params []json.RawMessage `json:"params"`
			}
			if err := conn.ReadJSON(&request); err != nil {
				return
			}
			currentPoll := int(poll.Load())
			var result any = []any{}
			var collection string
			var fields any
			switch request.Method {
			case "auth.login_ex":
				logins.Add(1)
				result = map[string]any{"response_type": "SUCCESS"}
			case "system.info":
				result = map[string]any{"hostname": "fractional-nas", "version": "TrueNAS-SCALE-26.0.0-BETA.3", "system_serial": "FIXTURE", "physical_cores": 4, "physmem": hostMemory}
			case "app.query":
				apps := make([]any, appCount)
				for i := range apps {
					name := fmt.Sprintf("fixture-app-%d", i)
					apps[i] = map[string]any{"id": name, "name": name, "state": "RUNNING"}
				}
				result = apps
			case "core.subscribe":
				if len(request.Params) != 1 || json.Unmarshal(request.Params[0], &collection) != nil {
					t.Error("invalid subscription arguments")
					return
				}
				switch collection {
				case `reporting.realtime:{"interval":2}`:
					realtimeSubscriptions.Add(1)
					result = "realtime-sub"
					fields = map[string]any{"cpu": map[string]any{"usage": 10}, "memory": map[string]any{"physical_memory_total": hostMemory, "physical_memory_available": hostMemory / 2}}
				case `app.stats:{"interval":2}`:
					statsSubscriptions.Add(1)
					result = "app-sub"
					apps := make([]any, appCount)
					for i := range apps {
						cpu := strconv.Itoa(i)
						if i < 2 {
							cpu = cpus[currentPoll][i]
						}
						apps[i] = map[string]any{
							"app_name": fmt.Sprintf("fixture-app-%d", i), "cpu_usage": json.Number(cpu),
							"memory": 1072459776 + currentPoll*1024 + i,
							"blkio":  map[string]any{"read": 29720576 + currentPoll*4096, "write": currentPoll * 2048},
							"networks": []any{
								map[string]any{"interface_name": "eth0", "rx_bytes": 3671 + currentPoll, "tx_bytes": 284 + currentPoll},
								map[string]any{"interface_name": "eth1", "rx_bytes": 2, "tx_bytes": 1},
							},
						}
					}
					fields = apps
				default:
					t.Errorf("unexpected subscription %s", collection)
					return
				}
			case "core.unsubscribe":
				var id string
				if len(request.Params) != 1 || json.Unmarshal(request.Params[0], &id) != nil || (id != "app-sub" && id != "realtime-sub") {
					t.Error("subscription cleanup used an unexpected ID")
					return
				}
				unsubscribes.Add(1)
				result = nil
			case "reporting.get_data", "pool.query", "boot.get_state", "pool.dataset.query", "disk.query", "alert.list", "service.query", "vm.query", "sharing.smb.query", "sharing.nfs.query", "zfs.resource.snapshot.query", "replication.query":
			default:
				t.Errorf("unexpected or mutating RPC method %s", request.Method)
				return
			}
			if conn.WriteJSON(map[string]any{"jsonrpc": "2.0", "id": request.ID, "result": result}) != nil {
				return
			}
			if collection != "" {
				if conn.WriteJSON(map[string]any{"jsonrpc": "2.0", "method": "collection_update", "params": map[string]any{"collection": collection, "fields": fields}}) != nil {
					return
				}
			}
		}
	}))
	t.Cleanup(server.Close)
	client, err := truenas.NewClient(truenas.ClientConfig{
		Host: server.URL, APIKey: "fixture-key", Username: "fixture-user", Timeout: 2 * time.Second,
		// The fixture certificate is self-signed; authenticate its exact leaf
		// with the client's normal pin verifier rather than the system CA set.
		InsecureSkipVerify: true, Fingerprint: fmt.Sprintf("%x", sha256.Sum256(server.Certificate().Raw)),
	})
	if err != nil {
		t.Fatal(err)
	}
	t.Cleanup(client.Close)
	instance := config.TrueNASInstance{ID: "fractional-connection", Host: server.URL, Enabled: true, PollIntervalSecs: 60}
	provider := truenas.NewLiveProviderForConnection(&truenas.APIFetcher{Client: client}, instance.ID)
	poller := NewTrueNASPoller(nil, 0, nil)
	poller.providersByOrg["default"] = map[string]*truenas.Provider{instance.ID: provider}
	poller.configsByOrg["default"] = map[string]config.TrueNASInstance{instance.ID: instance}
	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	metricsStore, err := metrics.NewStore(metrics.DefaultConfig(t.TempDir()))
	if err != nil {
		t.Fatal(err)
	}
	t.Cleanup(func() { _ = metricsStore.Close() })
	monitor := &Monitor{resourceStore: resourceStore, metricsHistory: NewMetricsHistory(1024, 24*time.Hour), metricsStore: metricsStore}
	var lastSuccess, lastStats time.Time
	var firstIDs = make(map[string]string)
	for currentPoll := range cpus {
		poll.Store(int32(currentPoll))
		poller.ensureConnectionRuntimeStatusLocked("default", instance.ID).nextPollAt = time.Now().Add(-time.Second)
		poller.pollAll(context.Background())
		summary := poller.ConnectionSummaries("default", []config.TrueNASInstance{instance})[instance.ID]
		if summary.Poll == nil || summary.Poll.LastSuccessAt == nil || summary.Poll.LastError != nil || !summary.Poll.LastSuccessAt.After(lastSuccess) {
			t.Fatalf("poll %d did not complete with fresh success: %+v", currentPoll, summary)
		}
		lastSuccess = *summary.Poll.LastSuccessAt
		if summary.Transport == nil || summary.Transport.AuthMechanism != "api-key-plain" || !summary.Transport.Connected || summary.Transport.Reconnects != 0 {
			t.Fatalf("fractional stats discarded/re-authenticated the session: %+v", summary.Transport)
		}
		snapshot := provider.Snapshot()
		if snapshot == nil || len(snapshot.Apps) != appCount || snapshot.System.CPUPercent != 10 {
			t.Fatalf("poll %d lost inventory or host telemetry", currentPoll)
		}
		if snapshot.Apps[0].Stats == nil || !snapshot.Apps[0].Stats.CollectedAt.After(lastStats) {
			t.Fatalf("poll %d did not collect fresh app stats", currentPoll)
		}
		lastStats = snapshot.Apps[0].Stats.CollectedAt
		resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
			unifiedresources.SourceTrueNAS: poller.GetCurrentRecordsForOrg("default"),
		})
		monitor.syncUnifiedAppContainerMetrics(resourceStore)
		metricsStore.Flush()
		appResources := 0
		for _, resource := range resourceStore.GetAll() {
			if resource.Type != unifiedresources.ResourceTypeAppContainer {
				continue
			}
			appResources++
			var i int
			if _, err := fmt.Sscanf(resource.Name, "fixture-app-%d", &i); err != nil || i < 0 || i >= appCount {
				t.Fatalf("unexpected app identity %s", resource.Name)
			}
			cpu := float64(i)
			if i < 2 {
				cpu, _ = strconv.ParseFloat(cpus[currentPoll][i], 64)
			}
			memory := int64(1072459776 + currentPoll*1024 + i)
			want := map[string]float64{"cpu": cpu, "memory": float64(memory) / float64(hostMemory) * 100, "netin": float64(3673 + currentPoll), "netout": float64(285 + currentPoll)}
			if resource.Metrics == nil || resource.Metrics.CPU == nil || resource.Metrics.CPU.Percent != cpu ||
				resource.Metrics.Memory == nil || resource.Metrics.Memory.Used == nil || *resource.Metrics.Memory.Used != memory ||
				resource.Metrics.NetIn == nil || resource.Metrics.NetIn.Value != want["netin"] || resource.Metrics.NetOut == nil || resource.Metrics.NetOut.Value != want["netout"] ||
				resource.TrueNAS == nil || resource.TrueNAS.App == nil || resource.TrueNAS.App.Stats == nil || !resource.TrueNAS.App.Stats.CollectedAt.Equal(lastStats) {
				t.Fatalf("poll %d app %d lost precision, sibling readings or observation time: %+v", currentPoll, i, resource)
			}
			if currentPoll == 0 {
				firstIDs[resource.Name] = resource.ID
			} else if resource.ID != firstIDs[resource.Name] {
				t.Fatalf("poll %d changed canonical app identity", currentPoll)
			}
			target := resourceStore.MetricsTargetForResource(resource.ID)
			if target == nil || target.ResourceType != "app-container" || target.ResourceID != "system:fractional-connection/app:"+resource.Name {
				t.Fatalf("unexpected connection-scoped History target: %+v", target)
			}
			history := monitor.GetGuestMetrics("docker:"+target.ResourceID, time.Hour)
			// The first poll also crosses the real persistent/chart read path.
			// Later short polls may share a persisted time bucket; do not treat
			// its aggregation as three native sixty-second observations.
			if currentPoll == 0 {
				chart := monitor.GetGuestMetricsForChart("docker:"+target.ResourceID, "dockerContainer", target.ResourceID, 7*24*time.Hour)
				for metric, value := range want {
					points := chart[metric]
					if len(points) != 1 || math.Abs(points[0].Value-value) > 1e-12 {
						t.Fatalf("app %d %s persisted/chart History lost the live sample: %+v, want %v", i, metric, points, value)
					}
				}
			}
			for metric, value := range want {
				points := history[metric]
				if len(points) != currentPoll+1 || math.Abs(points[len(points)-1].Value-value) > 1e-12 || !points[len(points)-1].Timestamp.Equal(resource.LastSeen) {
					t.Fatalf("poll %d app %d %s History lost value/freshness: %+v, want %v", currentPoll, i, metric, points, value)
				}
			}
		}
		if appResources != appCount || len(poller.GetCurrentRecordsForOrg("other-tenant")) != 0 || sessions.Load() != 1 || logins.Load() != 1 ||
			statsSubscriptions.Load() != int32(currentPoll+1) || realtimeSubscriptions.Load() != int32(currentPoll+1) || unsubscribes.Load() != int32(2*(currentPoll+1)) {
			t.Fatalf("poll %d crossed tenant scope, lost apps or churned session/subscriptions: apps=%d sessions=%d logins=%d stats=%d realtime=%d unsubscribe=%d", currentPoll, appResources, sessions.Load(), logins.Load(), statsSubscriptions.Load(), realtimeSubscriptions.Load(), unsubscribes.Load())
		}
		// Rendering/History without a new observation must not renew samples.
		monitor.syncUnifiedAppContainerMetrics(resourceStore)
		history := monitor.GetGuestMetrics("docker:system:fractional-connection/app:fixture-app-0", time.Hour)
		for _, metric := range []string{"cpu", "memory", "netin", "netout"} {
			if len(history[metric]) != currentPoll+1 {
				t.Fatalf("rereading the same poll renewed %s History", metric)
			}
		}
	}
	client.Close()
	if client.TransportStatus().Connected {
		t.Fatal("closing the client left its authenticated session active")
	}
}

func TestSyncUnifiedAppContainerMetricsRecordsTrueNASHistory(t *testing.T) {
	previous := truenas.IsFeatureEnabled()
	truenas.SetFeatureEnabled(true)
	t.Cleanup(func() {
		truenas.SetFeatureEnabled(previous)
	})

	cfg := metrics.DefaultConfig(t.TempDir())
	store, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = store.Close() }()

	fixtures := truenas.DefaultFixtures()
	fixtures.CollectedAt = time.Now().UTC().Truncate(time.Second)
	fixtures.System.CollectedAt = fixtures.CollectedAt

	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	records := truenas.NewProvider(fixtures).Records()
	resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
		unifiedresources.SourceTrueNAS: records,
	})

	monitor := &Monitor{
		resourceStore:  resourceStore,
		metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
		metricsStore:   store,
	}

	monitor.syncUnifiedAppContainerMetrics(resourceStore)

	var appResourceID string
	for _, resource := range resourceStore.GetAll() {
		if resource.Type == unifiedresources.ResourceTypeAppContainer && resource.Name == "Nextcloud" {
			appResourceID = resource.ID
			break
		}
	}
	if appResourceID == "" {
		t.Fatal("expected Nextcloud app-container resource in unified store")
	}

	target := resourceStore.MetricsTargetForResource(appResourceID)
	if target == nil || target.ResourceType != "app-container" || target.ResourceID != "system:truenas-main/app:nextcloud" {
		t.Fatalf("unexpected app-container metrics target %+v", target)
	}

	inMemory := monitor.GetGuestMetrics("docker:system:truenas-main/app:nextcloud", time.Hour)
	if got := len(inMemory["cpu"]); got == 0 {
		t.Fatalf("expected in-memory cpu history for nextcloud")
	}
	if got := len(inMemory["netin"]); got == 0 {
		t.Fatalf("expected in-memory network history for nextcloud")
	}

	storeBacked := monitor.GetGuestMetricsForChart("docker:system:truenas-main/app:nextcloud", "dockerContainer", "system:truenas-main/app:nextcloud", 7*24*time.Hour)
	if got := len(storeBacked["cpu"]); got == 0 {
		t.Fatalf("expected persisted cpu history for nextcloud")
	}
}

func TestSyncUnifiedStorageMetricsUsesPBSObservationTimeAcrossRegistryRebuilds(t *testing.T) {
	observedAt := time.Now().UTC().Add(-time.Minute).Truncate(time.Second)
	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	resourceStore.PopulateFromSnapshot(models.StateSnapshot{
		PBSInstances: []models.PBSInstance{{
			ID:       "pbs-main",
			Name:     "pbs-main",
			Status:   "online",
			LastSeen: observedAt,
			Datastores: []models.PBSDatastore{{
				Name:   "backups",
				Status: "available",
				Total:  1000,
				Used:   400,
				Free:   600,
				Usage:  40,
			}},
		}},
	})

	var targetID string
	for _, resource := range resourceStore.GetAll() {
		if resource.Type != unifiedresources.ResourceTypeStorage || resource.Storage == nil || resource.Storage.Platform != "pbs" {
			continue
		}
		target := resourceStore.MetricsTargetForResource(resource.ID)
		if target == nil || target.ResourceType != "storage" {
			t.Fatalf("unexpected PBS datastore metrics target: %+v", target)
		}
		targetID = target.ResourceID
		break
	}
	if targetID == "" {
		t.Fatal("expected a PBS datastore in the unified resource store")
	}

	cfg := metrics.DefaultConfig(t.TempDir())
	persistentStore, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = persistentStore.Close() }()

	monitor := &Monitor{
		metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
		metricsStore:   persistentStore,
	}

	// PVE node completions and read-side refreshes can rebuild the registry
	// repeatedly without a new PBS poll. They must not invent fresh samples.
	monitor.syncUnifiedStorageMetrics(resourceStore)
	monitor.syncUnifiedStorageMetrics(resourceStore)

	inMemory := monitor.GetStorageMetrics(targetID, time.Hour)["usage"]
	if len(inMemory) != 1 {
		t.Fatalf("in-memory PBS usage points = %d, want one per source observation: %+v", len(inMemory), inMemory)
	}
	if !inMemory[0].Timestamp.Equal(observedAt) {
		t.Fatalf("in-memory PBS usage timestamp = %s, want source observation %s", inMemory[0].Timestamp, observedAt)
	}

	persisted, err := persistentStore.Query("storage", targetID, "usage", observedAt.Add(-time.Second), observedAt.Add(time.Second), 0)
	if err != nil {
		t.Fatalf("query persisted PBS usage: %v", err)
	}
	if len(persisted) != 1 {
		t.Fatalf("persisted PBS usage points = %d, want one per source observation: %+v", len(persisted), persisted)
	}
	if !persisted[0].Timestamp.Equal(observedAt) {
		t.Fatalf("persisted PBS usage timestamp = %s, want source observation %s", persisted[0].Timestamp, observedAt)
	}
}

func TestSyncUnifiedAppContainerMetricsUsesSourceObservationTimeAcrossRegistryRebuilds(t *testing.T) {
	previous := truenas.IsFeatureEnabled()
	truenas.SetFeatureEnabled(true)
	t.Cleanup(func() {
		truenas.SetFeatureEnabled(previous)
	})

	fixtures := truenas.DefaultFixtures()
	fixtures.CollectedAt = time.Now().UTC().Truncate(time.Second)
	fixtures.System.CollectedAt = fixtures.CollectedAt
	observedAt := fixtures.CollectedAt

	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
		unifiedresources.SourceTrueNAS: truenas.NewProvider(fixtures).Records(),
	})

	var targetID string
	for _, resource := range resourceStore.GetAll() {
		if resource.Type != unifiedresources.ResourceTypeAppContainer || resource.Name != "Nextcloud" {
			continue
		}
		target := resourceStore.MetricsTargetForResource(resource.ID)
		if target == nil || target.ResourceType != "app-container" {
			t.Fatalf("unexpected app-container metrics target: %+v", target)
		}
		targetID = target.ResourceID
		break
	}
	if targetID == "" {
		t.Fatal("expected a TrueNAS app-container in the unified resource store")
	}

	cfg := metrics.DefaultConfig(t.TempDir())
	persistentStore, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = persistentStore.Close() }()

	monitor := &Monitor{
		metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
		metricsStore:   persistentStore,
	}

	// Read-side registry rebuilds must not invent a fresh sample for the same
	// source observation (the write amplification in #1966).
	monitor.syncUnifiedAppContainerMetrics(resourceStore)
	monitor.syncUnifiedAppContainerMetrics(resourceStore)

	inMemory := monitor.GetGuestMetrics("docker:"+targetID, time.Hour)["cpu"]
	if len(inMemory) != 1 {
		t.Fatalf("in-memory app-container cpu points = %d, want one per source observation: %+v", len(inMemory), inMemory)
	}
	if !inMemory[0].Timestamp.Equal(observedAt) {
		t.Fatalf("in-memory app-container cpu timestamp = %s, want source observation %s", inMemory[0].Timestamp, observedAt)
	}

	persisted, err := persistentStore.Query("dockerContainer", targetID, "cpu", observedAt.Add(-time.Second), observedAt.Add(time.Second), 0)
	if err != nil {
		t.Fatalf("query persisted app-container cpu: %v", err)
	}
	if len(persisted) != 1 {
		t.Fatalf("persisted app-container cpu points = %d, want one per source observation: %+v", len(persisted), persisted)
	}
	if !persisted[0].Timestamp.Equal(observedAt) {
		t.Fatalf("persisted app-container cpu timestamp = %s, want source observation %s", persisted[0].Timestamp, observedAt)
	}
}

func TestSyncUnifiedVMMetricsUsesSourceObservationTimeAcrossRegistryRebuilds(t *testing.T) {
	previous := vmware.IsFeatureEnabled()
	vmware.SetFeatureEnabled(true)
	t.Cleanup(func() {
		vmware.SetFeatureEnabled(previous)
	})

	observedAt := time.Now().UTC().Truncate(time.Second)
	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
		unifiedresources.SourceVMware: vmware.NewProvider(vmware.InventorySnapshot{
			ConnectionID:   "vc-1",
			ConnectionName: "Lab VC",
			VCenterHost:    "vc.lab.local",
			CollectedAt:    observedAt,
			VMs: []vmware.InventoryVM{{
				VM:            "vm-201",
				Name:          "app-01",
				PowerState:    "POWERED_ON",
				CPUCount:      4,
				MemorySizeMiB: 8192,
				Metrics: &vmware.InventoryMetrics{
					CPUPercent:              ptrFloat64(38.1),
					MemoryPercent:           ptrFloat64(57.5),
					MemoryUsedBytes:         ptrInt64(5033164800),
					MemoryTotalBytes:        ptrInt64(8589934592),
					NetInBytesPerSecond:     ptrFloat64(512),
					NetOutBytesPerSecond:    ptrFloat64(768),
					DiskReadBytesPerSecond:  ptrFloat64(1536),
					DiskWriteBytesPerSecond: ptrFloat64(2048),
				},
			}},
		}).Records(),
	})

	var targetID string
	for _, resource := range resourceStore.GetAll() {
		if resource.Type != unifiedresources.ResourceTypeVM || resource.Name != "app-01" {
			continue
		}
		target := resourceStore.MetricsTargetForResource(resource.ID)
		if target == nil || target.ResourceType != "vm" {
			t.Fatalf("unexpected VM metrics target: %+v", target)
		}
		targetID = target.ResourceID
		break
	}
	if targetID == "" {
		t.Fatal("expected a VMware VM in the unified resource store")
	}

	cfg := metrics.DefaultConfig(t.TempDir())
	persistentStore, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = persistentStore.Close() }()

	monitor := &Monitor{
		metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
		metricsStore:   persistentStore,
	}

	monitor.syncUnifiedVMMetrics(resourceStore)
	monitor.syncUnifiedVMMetrics(resourceStore)

	inMemory := monitor.GetGuestMetrics(targetID, time.Hour)["cpu"]
	if len(inMemory) != 1 {
		t.Fatalf("in-memory VM cpu points = %d, want one per source observation: %+v", len(inMemory), inMemory)
	}
	if !inMemory[0].Timestamp.Equal(observedAt) {
		t.Fatalf("in-memory VM cpu timestamp = %s, want source observation %s", inMemory[0].Timestamp, observedAt)
	}

	persisted, err := persistentStore.Query("vm", targetID, "cpu", observedAt.Add(-time.Second), observedAt.Add(time.Second), 0)
	if err != nil {
		t.Fatalf("query persisted VM cpu: %v", err)
	}
	if len(persisted) != 1 {
		t.Fatalf("persisted VM cpu points = %d, want one per source observation: %+v", len(persisted), persisted)
	}
	if !persisted[0].Timestamp.Equal(observedAt) {
		t.Fatalf("persisted VM cpu timestamp = %s, want source observation %s", persisted[0].Timestamp, observedAt)
	}
}

func TestSyncUnifiedAppContainerMetricsSkipsMockOwnedTrueNASHistoryWhenMockEnabled(t *testing.T) {
	previousFeature := truenas.IsFeatureEnabled()
	truenas.SetFeatureEnabled(true)
	t.Cleanup(func() {
		truenas.SetFeatureEnabled(previousFeature)
	})

	previousMock := mock.IsMockEnabled()
	mustSetMockEnabled(t, true)
	t.Cleanup(func() {
		mustSetMockEnabled(t, previousMock)
	})

	cfg := metrics.DefaultConfig(t.TempDir())
	store, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = store.Close() }()

	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
		unifiedresources.SourceTrueNAS: truenas.NewProvider(truenas.DefaultFixtures()).Records(),
	})

	monitor := &Monitor{
		resourceStore:  resourceStore,
		metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
		metricsStore:   store,
	}

	monitor.syncUnifiedAppContainerMetrics(resourceStore)

	if got := len(monitor.GetGuestMetrics("docker:system:truenas-main/app:nextcloud", time.Hour)["cpu"]); got != 0 {
		t.Fatalf("expected mock-owned TrueNAS app history to be skipped, got %d cpu points", got)
	}
}

func TestSyncUnifiedAgentMetricsRecordsTrueNASHostHistory(t *testing.T) {
	previous := truenas.IsFeatureEnabled()
	truenas.SetFeatureEnabled(true)
	t.Cleanup(func() {
		truenas.SetFeatureEnabled(previous)
	})

	cfg := metrics.DefaultConfig(t.TempDir())
	store, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = store.Close() }()

	fixtures := truenas.DefaultFixtures()
	fixtures.CollectedAt = time.Now().UTC().Truncate(time.Second)
	fixtures.System.CollectedAt = fixtures.CollectedAt

	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	records := truenas.NewProvider(fixtures).Records()
	resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
		unifiedresources.SourceTrueNAS: records,
	})

	monitor := &Monitor{
		resourceStore:  resourceStore,
		metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
		metricsStore:   store,
	}

	monitor.syncUnifiedAgentMetrics(resourceStore)

	var systemResourceID string
	for _, resource := range resourceStore.GetAll() {
		if resource.Type == unifiedresources.ResourceTypeAgent && resource.Name == "truenas-main" {
			systemResourceID = resource.ID
			break
		}
	}
	if systemResourceID == "" {
		t.Fatal("expected TrueNAS system resource in unified store")
	}

	target := resourceStore.MetricsTargetForResource(systemResourceID)
	if target == nil || target.ResourceType != "agent" || target.ResourceID != "truenas-main" {
		t.Fatalf("unexpected agent metrics target %+v", target)
	}

	inMemory := monitor.GetGuestMetrics("agent:truenas-main", time.Hour)
	if got := len(inMemory["cpu"]); got == 0 {
		t.Fatalf("expected in-memory cpu history for truenas-main")
	}
	if got := len(inMemory["netin"]); got == 0 {
		t.Fatalf("expected in-memory network history for truenas-main")
	}

	storeBacked := monitor.GetGuestMetricsForChart("agent:truenas-main", "agent", "truenas-main", 7*24*time.Hour)
	if got := len(storeBacked["cpu"]); got == 0 {
		t.Fatalf("expected persisted cpu history for truenas-main")
	}
}

func TestSyncUnifiedAgentMetricsSkipsMockOwnedProviderHistoryWhenMockEnabled(t *testing.T) {
	previousTrueNAS := truenas.IsFeatureEnabled()
	truenas.SetFeatureEnabled(true)
	t.Cleanup(func() {
		truenas.SetFeatureEnabled(previousTrueNAS)
	})

	previousMock := mock.IsMockEnabled()
	mustSetMockEnabled(t, true)
	t.Cleanup(func() {
		mustSetMockEnabled(t, previousMock)
	})

	cfg := metrics.DefaultConfig(t.TempDir())
	store, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = store.Close() }()

	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
		unifiedresources.SourceTrueNAS: truenas.NewProvider(truenas.DefaultFixtures()).Records(),
	})

	monitor := &Monitor{
		resourceStore:  resourceStore,
		metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
		metricsStore:   store,
	}

	monitor.syncUnifiedAgentMetrics(resourceStore)

	if got := len(monitor.GetGuestMetrics("agent:truenas-main", time.Hour)["cpu"]); got != 0 {
		t.Fatalf("expected mock-owned TrueNAS host history to be skipped, got %d cpu points", got)
	}
}

func TestSyncUnifiedAgentMetricsRecordsVMwareHostHistory(t *testing.T) {
	previous := vmware.IsFeatureEnabled()
	vmware.SetFeatureEnabled(true)
	t.Cleanup(func() {
		vmware.SetFeatureEnabled(previous)
	})

	cfg := metrics.DefaultConfig(t.TempDir())
	store, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = store.Close() }()

	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	records := vmware.NewProvider(vmware.InventorySnapshot{
		ConnectionID:   "vc-1",
		ConnectionName: "Lab VC",
		VCenterHost:    "vc.lab.local",
		CollectedAt:    time.Now().UTC().Truncate(time.Second),
		Hosts: []vmware.InventoryHost{{
			Host:            "host-101",
			Name:            "esxi-01.lab.local",
			ConnectionState: "CONNECTED",
			PowerState:      "POWERED_ON",
			HostUUID:        "uuid-host-1",
			Metrics: &vmware.InventoryMetrics{
				CPUPercent:              ptrFloat64(21.4),
				MemoryPercent:           ptrFloat64(63.2),
				MemoryUsedBytes:         ptrInt64(27144105984),
				MemoryTotalBytes:        ptrInt64(42949672960),
				NetInBytesPerSecond:     ptrFloat64(1024),
				NetOutBytesPerSecond:    ptrFloat64(2048),
				DiskReadBytesPerSecond:  ptrFloat64(4096),
				DiskWriteBytesPerSecond: ptrFloat64(8192),
			},
		}},
	}).Records()
	resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
		unifiedresources.SourceVMware: records,
	})

	monitor := &Monitor{
		resourceStore:  resourceStore,
		metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
		metricsStore:   store,
	}

	monitor.syncUnifiedAgentMetrics(resourceStore)

	var systemResourceID string
	for _, resource := range resourceStore.GetAll() {
		if resource.Type == unifiedresources.ResourceTypeAgent && resource.Name == "esxi-01.lab.local" {
			systemResourceID = resource.ID
			break
		}
	}
	if systemResourceID == "" {
		t.Fatal("expected VMware host resource in unified store")
	}

	target := resourceStore.MetricsTargetForResource(systemResourceID)
	if target == nil || target.ResourceType != "agent" || target.ResourceID != "vc-1:host:host-101" {
		t.Fatalf("unexpected agent metrics target %+v", target)
	}

	inMemory := monitor.GetGuestMetrics("agent:vc-1:host:host-101", time.Hour)
	if got := len(inMemory["cpu"]); got == 0 {
		t.Fatalf("expected in-memory cpu history for VMware host")
	}
	if got := len(inMemory["netin"]); got == 0 {
		t.Fatalf("expected in-memory network history for VMware host")
	}

	storeBacked := monitor.GetGuestMetricsForChart("agent:vc-1:host:host-101", "agent", "vc-1:host:host-101", 7*24*time.Hour)
	if got := len(storeBacked["cpu"]); got == 0 {
		t.Fatalf("expected persisted cpu history for VMware host")
	}
}

func TestSyncUnifiedVMMetricsRecordsVMwareVMHistory(t *testing.T) {
	previous := vmware.IsFeatureEnabled()
	vmware.SetFeatureEnabled(true)
	t.Cleanup(func() {
		vmware.SetFeatureEnabled(previous)
	})

	cfg := metrics.DefaultConfig(t.TempDir())
	store, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = store.Close() }()

	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	records := vmware.NewProvider(vmware.InventorySnapshot{
		ConnectionID:   "vc-1",
		ConnectionName: "Lab VC",
		VCenterHost:    "vc.lab.local",
		CollectedAt:    time.Now().UTC().Truncate(time.Second),
		VMs: []vmware.InventoryVM{{
			VM:            "vm-201",
			Name:          "app-01",
			PowerState:    "POWERED_ON",
			CPUCount:      4,
			MemorySizeMiB: 8192,
			Metrics: &vmware.InventoryMetrics{
				CPUPercent:              ptrFloat64(38.1),
				MemoryPercent:           ptrFloat64(57.5),
				MemoryUsedBytes:         ptrInt64(5033164800),
				MemoryTotalBytes:        ptrInt64(8589934592),
				NetInBytesPerSecond:     ptrFloat64(512),
				NetOutBytesPerSecond:    ptrFloat64(768),
				DiskReadBytesPerSecond:  ptrFloat64(1536),
				DiskWriteBytesPerSecond: ptrFloat64(2048),
			},
		}},
	}).Records()
	resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
		unifiedresources.SourceVMware: records,
	})

	monitor := &Monitor{
		resourceStore:  resourceStore,
		metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
		metricsStore:   store,
	}

	monitor.syncUnifiedVMMetrics(resourceStore)

	var vmResourceID string
	for _, resource := range resourceStore.GetAll() {
		if resource.Type == unifiedresources.ResourceTypeVM && resource.Name == "app-01" {
			vmResourceID = resource.ID
			break
		}
	}
	if vmResourceID == "" {
		t.Fatal("expected VMware VM resource in unified store")
	}

	target := resourceStore.MetricsTargetForResource(vmResourceID)
	if target == nil || target.ResourceType != "vm" || target.ResourceID != "vc-1:vm:vm-201" {
		t.Fatalf("unexpected vm metrics target %+v", target)
	}

	inMemory := monitor.GetGuestMetrics("vc-1:vm:vm-201", time.Hour)
	if got := len(inMemory["cpu"]); got == 0 {
		t.Fatalf("expected in-memory cpu history for VMware VM")
	}
	if got := len(inMemory["netin"]); got == 0 {
		t.Fatalf("expected in-memory network history for VMware VM")
	}

	storeBacked := monitor.GetGuestMetricsForChart("vc-1:vm:vm-201", "vm", "vc-1:vm:vm-201", 7*24*time.Hour)
	if got := len(storeBacked["cpu"]); got == 0 {
		t.Fatalf("expected persisted cpu history for VMware VM")
	}
}

func TestSyncUnifiedPhysicalDiskMetricsRecordsTrueNASDiskHistory(t *testing.T) {
	previous := truenas.IsFeatureEnabled()
	truenas.SetFeatureEnabled(true)
	t.Cleanup(func() {
		truenas.SetFeatureEnabled(previous)
	})

	cfg := metrics.DefaultConfig(t.TempDir())
	store, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = store.Close() }()

	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	fixtures := truenas.DefaultFixtures()
	fixtures.CollectedAt = time.Now().UTC().Truncate(time.Second)
	fixtures.System.CollectedAt = fixtures.CollectedAt
	records := truenas.NewProvider(fixtures).Records()
	resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
		unifiedresources.SourceTrueNAS: records,
	})

	monitor := &Monitor{
		resourceStore: resourceStore,
		metricsStore:  store,
	}

	monitor.syncUnifiedPhysicalDiskMetrics(resourceStore)

	var diskResourceID string
	for _, resource := range resourceStore.GetAll() {
		if resource.Type != unifiedresources.ResourceTypePhysicalDisk || resource.PhysicalDisk == nil {
			continue
		}
		if strings.TrimSpace(resource.PhysicalDisk.DevPath) == "/dev/sdc" {
			diskResourceID = resource.ID
			break
		}
	}
	if diskResourceID == "" {
		t.Fatal("expected TrueNAS sdc disk resource in unified store")
	}

	target := resourceStore.MetricsTargetForResource(diskResourceID)
	if target == nil || target.ResourceType != "disk" || target.ResourceID != "WD-WX12A3456" {
		t.Fatalf("unexpected physical-disk metrics target %+v", target)
	}

	charts := monitor.GetPhysicalDiskTemperatureCharts(7 * 24 * time.Hour)
	entry, ok := charts[target.ResourceID]
	if !ok {
		t.Fatalf("expected persisted disk chart for %s, got %#v", target.ResourceID, charts)
	}
	if entry.Node != "truenas-main" {
		t.Fatalf("chart node = %q, want truenas-main", entry.Node)
	}
	if len(entry.Temperature) == 0 {
		t.Fatalf("expected temperature history for %s", target.ResourceID)
	}
	last := entry.Temperature[len(entry.Temperature)-1]
	if last.Value != 63 {
		t.Fatalf("expected last disk temperature 63, got %.2f", last.Value)
	}
}

func TestSyncUnifiedStorageAndDiskMetricsSkipMockOwnedTrueNASHistoryWhenMockEnabled(t *testing.T) {
	previousFeature := truenas.IsFeatureEnabled()
	truenas.SetFeatureEnabled(true)
	t.Cleanup(func() {
		truenas.SetFeatureEnabled(previousFeature)
	})

	previousMock := mock.IsMockEnabled()
	mustSetMockEnabled(t, true)
	t.Cleanup(func() {
		mustSetMockEnabled(t, previousMock)
	})

	cfg := metrics.DefaultConfig(t.TempDir())
	store, err := metrics.NewStore(cfg)
	if err != nil {
		t.Fatalf("metrics.NewStore() error = %v", err)
	}
	defer func() { _ = store.Close() }()

	resourceStore := unifiedresources.NewMonitorAdapter(nil)
	resourceStore.PopulateSnapshotAndSupplemental(models.StateSnapshot{}, map[unifiedresources.DataSource][]unifiedresources.IngestRecord{
		unifiedresources.SourceTrueNAS: truenas.NewProvider(truenas.DefaultFixtures()).Records(),
	})

	monitor := &Monitor{
		resourceStore:  resourceStore,
		metricsHistory: NewMetricsHistory(1024, 24*time.Hour),
		metricsStore:   store,
	}

	monitor.syncUnifiedStorageMetrics(resourceStore)
	monitor.syncUnifiedPhysicalDiskMetrics(resourceStore)

	if got := len(monitor.GetStorageMetrics(mock.TrueNASPoolMetricID("truenas-main", "tank"), time.Hour)["usage"]); got != 0 {
		t.Fatalf("expected mock-owned TrueNAS storage history to be skipped, got %d usage points", got)
	}

	points, err := store.Query("disk", "WD-WX12A3456", "smart_temp", time.Now().Add(-time.Hour), time.Now(), 0)
	if err != nil {
		t.Fatalf("store.Query() error = %v", err)
	}
	if len(points) != 0 {
		t.Fatalf("expected mock-owned TrueNAS disk history to be skipped, got %d points", len(points))
	}
}

func TestBuildBroadcastFrontendStatePrefersLiveAlertManagerOverSnapshotAlerts(t *testing.T) {
	alertManager := alerts.NewManagerWithDataDir(t.TempDir())
	defer alertManager.Stop()

	alertManager.SyncUnifiedResourceIncidents([]unifiedresources.Resource{{
		ID:      "agent:truenas-main",
		Type:    unifiedresources.ResourceTypeAgent,
		Name:    "truenas-main",
		Sources: []unifiedresources.DataSource{unifiedresources.SourceTrueNAS},
		TrueNAS: &unifiedresources.TrueNASData{Hostname: "truenas-main"},
		Incidents: []unifiedresources.ResourceIncident{{
			Provider: "truenas",
			NativeID: "alert-live-1",
			Code:     "truenas_volume_status",
			Severity: storagehealth.RiskCritical,
			Summary:  "Pool tank is FAULTED",
		}},
	}})

	monitor := &Monitor{
		state:        models.NewState(),
		alertManager: alertManager,
	}

	staleSnapshot := models.StateSnapshot{
		ActiveAlerts: []models.Alert{{
			ID:           "snapshot-stale-1",
			Type:         "offline",
			ResourceID:   "stale-resource",
			ResourceName: "stale-resource",
			Message:      "stale snapshot alert",
		}},
	}

	frontend := monitor.buildBroadcastFrontendStateFromSnapshot(staleSnapshot, monitor.mockModeFence.begin())

	if len(frontend.ActiveAlerts) != 1 {
		t.Fatalf("expected 1 live active alert in frontend state, got %d", len(frontend.ActiveAlerts))
	}
	if got := frontend.ActiveAlerts[0].ID; got == "snapshot-stale-1" || got == "" {
		t.Fatalf("expected live alert manager id, got %q", got)
	}
	if got := frontend.ActiveAlerts[0].Type; got != "resource-incident" {
		t.Fatalf("alert type = %q, want resource-incident", got)
	}
	if got := frontend.ActiveAlerts[0].Message; got == "stale snapshot alert" {
		t.Fatalf("expected live alert manager payload, got stale snapshot message %q", got)
	}
}

func TestActiveAlertsSnapshotPrefersLiveAlertManagerOverStateSnapshot(t *testing.T) {
	alertManager := alerts.NewManagerWithDataDir(t.TempDir())
	defer alertManager.Stop()

	alertManager.SyncUnifiedResourceIncidents([]unifiedresources.Resource{{
		ID:      "agent:truenas-live",
		Type:    unifiedresources.ResourceTypeAgent,
		Name:    "truenas-live",
		Sources: []unifiedresources.DataSource{unifiedresources.SourceTrueNAS},
		TrueNAS: &unifiedresources.TrueNASData{Hostname: "truenas-live"},
		Incidents: []unifiedresources.ResourceIncident{{
			Provider: "truenas",
			NativeID: "live-active-1",
			Code:     "truenas_volume_status",
			Severity: storagehealth.RiskCritical,
			Summary:  "Pool tank is FAULTED",
		}},
	}})

	state := models.NewState()
	state.UpdateActiveAlerts([]models.Alert{{
		ID:           "snapshot-active-1",
		Type:         "offline",
		ResourceID:   "snapshot-resource",
		ResourceName: "snapshot-resource",
		Message:      "stale snapshot active alert",
	}})

	monitor := &Monitor{
		state:        state,
		alertManager: alertManager,
	}

	active := monitor.ActiveAlertsSnapshot()
	if len(active) != 1 {
		t.Fatalf("expected 1 live active alert, got %d", len(active))
	}
	if got := active[0].ID; got == "snapshot-active-1" || got == "" {
		t.Fatalf("expected live alert manager alert id, got %q", got)
	}
	if got := active[0].Type; got != "resource-incident" {
		t.Fatalf("alert type = %q, want resource-incident", got)
	}
}

func TestRecentlyResolvedSnapshotPrefersLiveAlertManagerOverStateSnapshot(t *testing.T) {
	alertManager := alerts.NewManagerWithDataDir(t.TempDir())
	defer alertManager.Stop()

	resource := unifiedresources.Resource{
		ID:      "agent:truenas-live",
		Type:    unifiedresources.ResourceTypeAgent,
		Name:    "truenas-live",
		Sources: []unifiedresources.DataSource{unifiedresources.SourceTrueNAS},
		TrueNAS: &unifiedresources.TrueNASData{Hostname: "truenas-live"},
		Incidents: []unifiedresources.ResourceIncident{{
			Provider: "truenas",
			NativeID: "live-resolved-1",
			Code:     "truenas_volume_status",
			Severity: storagehealth.RiskCritical,
			Summary:  "Pool tank is FAULTED",
		}},
	}
	alertManager.SyncUnifiedResourceIncidents([]unifiedresources.Resource{resource})
	alertManager.SyncUnifiedResourceIncidents(nil)

	state := models.NewState()
	state.UpdateRecentlyResolved([]models.ResolvedAlert{{
		Alert: models.Alert{
			ID:           "snapshot-resolved-1",
			Type:         "offline",
			ResourceID:   "snapshot-resource",
			ResourceName: "snapshot-resource",
			Message:      "stale snapshot resolved alert",
		},
		ResolvedTime: time.Now().UTC(),
	}})

	monitor := &Monitor{
		state:        state,
		alertManager: alertManager,
	}

	resolved := monitor.RecentlyResolvedSnapshot()
	if len(resolved) == 0 {
		t.Fatal("expected recently resolved alert from live alert manager")
	}
	if got := resolved[0].ID; got == "snapshot-resolved-1" || got == "" {
		t.Fatalf("expected live resolved alert id, got %q", got)
	}
	if got := resolved[0].Type; got != "resource-incident" {
		t.Fatalf("resolved alert type = %q, want resource-incident", got)
	}
}

func TestBuildBroadcastFrontendStatePreservesSnapshotAlertsWithoutAlertManager(t *testing.T) {
	monitor := &Monitor{
		state: models.NewState(),
	}

	snapshot := models.StateSnapshot{
		ActiveAlerts: []models.Alert{{
			ID:           "snapshot-only-1",
			Type:         "offline",
			ResourceID:   "snapshot-resource",
			ResourceName: "snapshot-resource",
			Message:      "snapshot-only alert",
		}},
	}

	frontend := monitor.buildBroadcastFrontendStateFromSnapshot(snapshot, monitor.mockModeFence.begin())
	if len(frontend.ActiveAlerts) != 1 {
		t.Fatalf("expected snapshot alert to be preserved, got %#v", frontend.ActiveAlerts)
	}
	if got := frontend.ActiveAlerts[0].ID; got != "snapshot-only-1" {
		t.Fatalf("alert id = %q, want snapshot-only-1", got)
	}
}

func TestRecordTaskResult_Success(t *testing.T) {
	m := &Monitor{
		pollStatusMap:   make(map[string]*pollStatus),
		failureCounts:   make(map[string]int),
		lastOutcome:     make(map[string]taskOutcome),
		circuitBreakers: make(map[string]*circuitBreaker),
	}

	// Record a success
	m.recordTaskResult(InstanceTypePVE, "test-instance", nil)

	key := schedulerKey(InstanceTypePVE, "test-instance")

	// Verify failure count is reset
	if m.failureCounts[key] != 0 {
		t.Errorf("expected failureCounts[%s] = 0, got %d", key, m.failureCounts[key])
	}

	// Verify last outcome is success
	outcome, ok := m.lastOutcome[key]
	if !ok {
		t.Fatalf("expected lastOutcome[%s] to exist", key)
	}
	if !outcome.success {
		t.Error("expected outcome.success = true")
	}

	// Verify poll status
	status, ok := m.pollStatusMap[key]
	if !ok {
		t.Fatalf("expected pollStatusMap[%s] to exist", key)
	}
	if status.ConsecutiveFailures != 0 {
		t.Errorf("expected ConsecutiveFailures = 0, got %d", status.ConsecutiveFailures)
	}
}

func TestRecordTaskResult_Failure(t *testing.T) {
	m := &Monitor{
		pollStatusMap:   make(map[string]*pollStatus),
		failureCounts:   make(map[string]int),
		lastOutcome:     make(map[string]taskOutcome),
		circuitBreakers: make(map[string]*circuitBreaker),
	}

	testErr := errors.New("connection refused")

	// Record a failure
	m.recordTaskResult(InstanceTypePVE, "test-instance", testErr)

	key := schedulerKey(InstanceTypePVE, "test-instance")

	// Verify failure count is incremented
	if m.failureCounts[key] != 1 {
		t.Errorf("expected failureCounts[%s] = 1, got %d", key, m.failureCounts[key])
	}

	// Verify last outcome is failure
	outcome, ok := m.lastOutcome[key]
	if !ok {
		t.Fatalf("expected lastOutcome[%s] to exist", key)
	}
	if outcome.success {
		t.Error("expected outcome.success = false")
	}
	if outcome.err != testErr {
		t.Errorf("expected outcome.err = %v, got %v", testErr, outcome.err)
	}

	// Verify poll status
	status, ok := m.pollStatusMap[key]
	if !ok {
		t.Fatalf("expected pollStatusMap[%s] to exist", key)
	}
	if status.ConsecutiveFailures != 1 {
		t.Errorf("expected ConsecutiveFailures = 1, got %d", status.ConsecutiveFailures)
	}
	if status.LastErrorMessage != "connection refused" {
		t.Errorf("expected LastErrorMessage = 'connection refused', got %q", status.LastErrorMessage)
	}
}

func TestRecordTaskResult_ConsecutiveFailures(t *testing.T) {
	m := &Monitor{
		pollStatusMap:   make(map[string]*pollStatus),
		failureCounts:   make(map[string]int),
		lastOutcome:     make(map[string]taskOutcome),
		circuitBreakers: make(map[string]*circuitBreaker),
	}

	testErr := errors.New("timeout")

	// Record multiple failures
	m.recordTaskResult(InstanceTypePBS, "pbs-server", testErr)
	m.recordTaskResult(InstanceTypePBS, "pbs-server", testErr)
	m.recordTaskResult(InstanceTypePBS, "pbs-server", testErr)

	key := schedulerKey(InstanceTypePBS, "pbs-server")

	// Verify consecutive failures count
	status := m.pollStatusMap[key]
	if status.ConsecutiveFailures != 3 {
		t.Errorf("expected ConsecutiveFailures = 3, got %d", status.ConsecutiveFailures)
	}

	// FirstFailureAt should be set on first failure and not change
	if status.FirstFailureAt.IsZero() {
		t.Error("expected FirstFailureAt to be set")
	}
}

func TestRecordTaskResult_SuccessResetsFailures(t *testing.T) {
	m := &Monitor{
		pollStatusMap:   make(map[string]*pollStatus),
		failureCounts:   make(map[string]int),
		lastOutcome:     make(map[string]taskOutcome),
		circuitBreakers: make(map[string]*circuitBreaker),
	}

	testErr := errors.New("error")
	key := schedulerKey(InstanceTypePMG, "pmg-server")

	// Record some failures first
	m.recordTaskResult(InstanceTypePMG, "pmg-server", testErr)
	m.recordTaskResult(InstanceTypePMG, "pmg-server", testErr)

	if m.pollStatusMap[key].ConsecutiveFailures != 2 {
		t.Fatalf("expected 2 failures before reset")
	}

	// Now record a success
	m.recordTaskResult(InstanceTypePMG, "pmg-server", nil)

	// Verify everything is reset
	if m.failureCounts[key] != 0 {
		t.Errorf("expected failureCounts to be reset to 0, got %d", m.failureCounts[key])
	}
	if m.pollStatusMap[key].ConsecutiveFailures != 0 {
		t.Errorf("expected ConsecutiveFailures to be reset to 0, got %d", m.pollStatusMap[key].ConsecutiveFailures)
	}
	if !m.pollStatusMap[key].FirstFailureAt.IsZero() {
		t.Error("expected FirstFailureAt to be reset to zero")
	}
	// The recorded error must clear too: it means "current outstanding
	// failure", and the connections UI renders it as a live error banner.
	// A stale entry would show a red error on a healthy connection (#1493).
	if m.pollStatusMap[key].LastErrorMessage != "" {
		t.Errorf("expected LastErrorMessage to be cleared, got %q", m.pollStatusMap[key].LastErrorMessage)
	}
	if !m.pollStatusMap[key].LastErrorAt.IsZero() {
		t.Error("expected LastErrorAt to be reset to zero")
	}
	if m.pollStatusMap[key].LastErrorCategory != "" {
		t.Errorf("expected LastErrorCategory to be cleared, got %q", m.pollStatusMap[key].LastErrorCategory)
	}
}

func TestRecordTaskResult_NilMaps(t *testing.T) {
	// Monitor with nil internal maps - should not panic
	m := &Monitor{
		pollStatusMap:   make(map[string]*pollStatus),
		failureCounts:   nil, // nil
		lastOutcome:     nil, // nil
		circuitBreakers: make(map[string]*circuitBreaker),
	}

	// Should not panic
	m.recordTaskResult(InstanceTypePVE, "test", nil)
	m.recordTaskResult(InstanceTypePVE, "test", errors.New("error"))

	// pollStatusMap should still be updated
	key := schedulerKey(InstanceTypePVE, "test")
	if _, ok := m.pollStatusMap[key]; !ok {
		t.Error("expected pollStatusMap to be updated even with nil failureCounts/lastOutcome")
	}
}

func TestDescribeInstancesForScheduler_NoClients(t *testing.T) {
	m := &Monitor{
		pveClients: make(map[string]PVEClientInterface),
		pbsClients: make(map[string]*pbs.Client),
		pmgClients: make(map[string]*pmg.Client),
	}

	descriptors := m.describeInstancesForScheduler()
	if descriptors != nil {
		t.Errorf("expected nil for empty clients, got %v", descriptors)
	}
}

func TestDescribeInstancesForScheduler_PVEOnly(t *testing.T) {
	m := &Monitor{
		pveClients: map[string]PVEClientInterface{"pve-1": nil, "pve-2": nil},
		pbsClients: make(map[string]*pbs.Client),
		pmgClients: make(map[string]*pmg.Client),
	}

	descriptors := m.describeInstancesForScheduler()
	if len(descriptors) != 2 {
		t.Fatalf("expected 2 descriptors, got %d", len(descriptors))
	}

	// Should be sorted
	if descriptors[0].Name != "pve-1" || descriptors[1].Name != "pve-2" {
		t.Errorf("expected sorted order [pve-1, pve-2], got [%s, %s]", descriptors[0].Name, descriptors[1].Name)
	}

	for _, desc := range descriptors {
		if desc.Type != InstanceTypePVE {
			t.Errorf("expected type PVE, got %v", desc.Type)
		}
	}
}

func TestDescribeInstancesForScheduler_PBSOnly(t *testing.T) {
	m := &Monitor{
		pveClients: make(map[string]PVEClientInterface),
		pbsClients: map[string]*pbs.Client{"pbs-backup": nil},
		pmgClients: make(map[string]*pmg.Client),
	}

	descriptors := m.describeInstancesForScheduler()
	if len(descriptors) != 1 {
		t.Fatalf("expected 1 descriptor, got %d", len(descriptors))
	}

	if descriptors[0].Name != "pbs-backup" {
		t.Errorf("expected name 'pbs-backup', got %q", descriptors[0].Name)
	}
	if descriptors[0].Type != InstanceTypePBS {
		t.Errorf("expected type PBS, got %v", descriptors[0].Type)
	}
}

func TestDescribeInstancesForScheduler_PMGOnly(t *testing.T) {
	m := &Monitor{
		pveClients: make(map[string]PVEClientInterface),
		pbsClients: make(map[string]*pbs.Client),
		pmgClients: map[string]*pmg.Client{"pmg-mail": nil},
	}

	descriptors := m.describeInstancesForScheduler()
	if len(descriptors) != 1 {
		t.Fatalf("expected 1 descriptor, got %d", len(descriptors))
	}

	if descriptors[0].Name != "pmg-mail" {
		t.Errorf("expected name 'pmg-mail', got %q", descriptors[0].Name)
	}
	if descriptors[0].Type != InstanceTypePMG {
		t.Errorf("expected type PMG, got %v", descriptors[0].Type)
	}
}

func TestDescribeInstancesForScheduler_AllTypes(t *testing.T) {
	m := &Monitor{
		pveClients: map[string]PVEClientInterface{"pve-1": nil},
		pbsClients: map[string]*pbs.Client{"pbs-1": nil},
		pmgClients: map[string]*pmg.Client{"pmg-1": nil},
	}

	descriptors := m.describeInstancesForScheduler()
	if len(descriptors) != 3 {
		t.Fatalf("expected 3 descriptors, got %d", len(descriptors))
	}

	// Check we have one of each type
	types := make(map[InstanceType]bool)
	for _, desc := range descriptors {
		types[desc.Type] = true
	}
	if !types[InstanceTypePVE] || !types[InstanceTypePBS] || !types[InstanceTypePMG] {
		t.Error("expected one descriptor of each type")
	}
}

func TestDescribeInstancesForScheduler_NilSchedulerAndTracker(t *testing.T) {
	m := &Monitor{
		pveClients:       map[string]PVEClientInterface{"pve-1": nil},
		pbsClients:       make(map[string]*pbs.Client),
		pmgClients:       make(map[string]*pmg.Client),
		scheduler:        nil, // explicitly nil
		stalenessTracker: nil, // explicitly nil
	}

	// Should not panic with nil scheduler and stalenessTracker
	descriptors := m.describeInstancesForScheduler()
	if len(descriptors) != 1 {
		t.Fatalf("expected 1 descriptor, got %d", len(descriptors))
	}

	// LastScheduled and LastSuccess should be zero values
	if !descriptors[0].LastScheduled.IsZero() {
		t.Error("expected LastScheduled to be zero with nil scheduler")
	}
	if !descriptors[0].LastSuccess.IsZero() {
		t.Error("expected LastSuccess to be zero with nil stalenessTracker")
	}
}

func TestRescheduleTask_SuccessfulOutcome(t *testing.T) {
	cfg := &config.Config{
		PVEPollingInterval:          30 * time.Second,
		AdaptivePollingBaseInterval: 10 * time.Second,
	}

	m := &Monitor{
		config:        cfg,
		taskQueue:     NewTaskQueue(),
		lastOutcome:   make(map[string]taskOutcome),
		failureCounts: make(map[string]int),
	}

	task := ScheduledTask{
		InstanceName: "pve-1",
		InstanceType: InstanceTypePVE,
		Interval:     30 * time.Second,
		NextRun:      time.Now(),
	}

	key := schedulerKey(task.InstanceType, task.InstanceName)

	// Record a successful outcome
	m.lastOutcome[key] = taskOutcome{success: true}

	m.rescheduleTask(task)

	// Task should be rescheduled at regular interval (no backoff)
	m.taskQueue.mu.Lock()
	entry, ok := m.taskQueue.entries[key]
	m.taskQueue.mu.Unlock()

	if !ok {
		t.Fatal("expected task to be rescheduled")
	}

	// Should use base interval since scheduler is nil
	if entry.task.Interval != cfg.PVEPollingInterval {
		t.Errorf("expected interval %v, got %v", cfg.PVEPollingInterval, entry.task.Interval)
	}
}

func TestRescheduleTask_TransientFailureWithBackoff(t *testing.T) {
	cfg := &config.Config{
		PVEPollingInterval:          30 * time.Second,
		AdaptivePollingBaseInterval: 10 * time.Second,
	}

	m := &Monitor{
		config:           cfg,
		taskQueue:        NewTaskQueue(),
		lastOutcome:      make(map[string]taskOutcome),
		failureCounts:    make(map[string]int),
		maxRetryAttempts: 5,
		backoffCfg: backoffConfig{
			Initial:    5 * time.Second,
			Multiplier: 2,
			Jitter:     0, // no jitter for predictable testing
			Max:        5 * time.Minute,
		},
	}

	// Add randomFloat method for backoff calculation
	m.rng = nil // will use default random

	task := ScheduledTask{
		InstanceName: "pve-1",
		InstanceType: InstanceTypePVE,
		Interval:     30 * time.Second,
		NextRun:      time.Now(),
	}

	key := schedulerKey(task.InstanceType, task.InstanceName)

	// Record a transient failure (1st attempt, below maxRetryAttempts)
	m.failureCounts[key] = 1
	m.lastOutcome[key] = taskOutcome{
		success:   false,
		transient: true,
		err:       errors.New("connection timeout"),
	}

	m.rescheduleTask(task)

	// Task should be rescheduled with backoff delay
	m.taskQueue.mu.Lock()
	entry, ok := m.taskQueue.entries[key]
	m.taskQueue.mu.Unlock()

	if !ok {
		t.Fatal("expected task to be rescheduled with backoff")
	}

	// With backoff, interval should be modified
	if entry.task.Interval <= 0 {
		t.Errorf("expected positive backoff interval, got %v", entry.task.Interval)
	}
}

func TestRescheduleTask_NonTransientFailureGoesToDeadLetter(t *testing.T) {
	cfg := &config.Config{
		PVEPollingInterval: 30 * time.Second,
	}

	deadLetterQ := NewTaskQueue()

	m := &Monitor{
		config:           cfg,
		taskQueue:        NewTaskQueue(),
		deadLetterQueue:  deadLetterQ,
		lastOutcome:      make(map[string]taskOutcome),
		failureCounts:    make(map[string]int),
		maxRetryAttempts: 5,
	}

	task := ScheduledTask{
		InstanceName: "pve-1",
		InstanceType: InstanceTypePVE,
		Interval:     30 * time.Second,
		NextRun:      time.Now(),
	}

	key := schedulerKey(task.InstanceType, task.InstanceName)

	// Record a non-transient failure (permanent error)
	m.failureCounts[key] = 1
	m.lastOutcome[key] = taskOutcome{
		success:   false,
		transient: false, // non-transient
		err:       errors.New("authentication failed"),
	}

	m.rescheduleTask(task)

	// Task should NOT be in the main queue
	m.taskQueue.mu.Lock()
	_, inMainQueue := m.taskQueue.entries[key]
	m.taskQueue.mu.Unlock()

	if inMainQueue {
		t.Error("expected task to NOT be in main queue after non-transient failure")
	}

	// Task should be in dead letter queue
	deadLetterQ.mu.Lock()
	dlqSize := len(deadLetterQ.entries)
	deadLetterQ.mu.Unlock()

	if dlqSize != 1 {
		t.Errorf("expected 1 task in dead letter queue, got %d", dlqSize)
	}
}

func TestRescheduleTask_ExceededRetryAttemptsGoesToDeadLetter(t *testing.T) {
	cfg := &config.Config{
		PVEPollingInterval: 30 * time.Second,
	}

	deadLetterQ := NewTaskQueue()

	m := &Monitor{
		config:           cfg,
		taskQueue:        NewTaskQueue(),
		deadLetterQueue:  deadLetterQ,
		lastOutcome:      make(map[string]taskOutcome),
		failureCounts:    make(map[string]int),
		maxRetryAttempts: 3,
	}

	task := ScheduledTask{
		InstanceName: "pve-1",
		InstanceType: InstanceTypePVE,
		Interval:     30 * time.Second,
		NextRun:      time.Now(),
	}

	key := schedulerKey(task.InstanceType, task.InstanceName)

	// Exceed max retry attempts (failureCount >= maxRetryAttempts)
	m.failureCounts[key] = 3
	m.lastOutcome[key] = taskOutcome{
		success:   false,
		transient: true, // transient, but exceeded retries
		err:       errors.New("connection timeout"),
	}

	m.rescheduleTask(task)

	// Task should be in dead letter queue
	deadLetterQ.mu.Lock()
	dlqSize := len(deadLetterQ.entries)
	deadLetterQ.mu.Unlock()

	if dlqSize != 1 {
		t.Errorf("expected 1 task in dead letter queue after exceeding retries, got %d", dlqSize)
	}
}

func TestRescheduleTask_NoOutcomeUsesDefaultInterval(t *testing.T) {
	cfg := &config.Config{
		PVEPollingInterval:          45 * time.Second,
		AdaptivePollingBaseInterval: 10 * time.Second,
	}

	m := &Monitor{
		config:        cfg,
		taskQueue:     NewTaskQueue(),
		lastOutcome:   make(map[string]taskOutcome),
		failureCounts: make(map[string]int),
	}

	task := ScheduledTask{
		InstanceName: "pve-1",
		InstanceType: InstanceTypePVE,
		Interval:     0, // no interval set
		NextRun:      time.Now(),
	}

	key := schedulerKey(task.InstanceType, task.InstanceName)

	// No outcome recorded - hasOutcome will be false
	m.rescheduleTask(task)

	m.taskQueue.mu.Lock()
	entry, ok := m.taskQueue.entries[key]
	m.taskQueue.mu.Unlock()

	if !ok {
		t.Fatal("expected task to be rescheduled")
	}

	// Should use config PVE polling interval
	if entry.task.Interval != cfg.PVEPollingInterval {
		t.Errorf("expected interval %v, got %v", cfg.PVEPollingInterval, entry.task.Interval)
	}
}

func TestRescheduleTask_PBSInstance(t *testing.T) {
	cfg := &config.Config{
		PBSPollingInterval:          60 * time.Second,
		AdaptivePollingBaseInterval: 10 * time.Second,
	}

	m := &Monitor{
		config:        cfg,
		taskQueue:     NewTaskQueue(),
		lastOutcome:   make(map[string]taskOutcome),
		failureCounts: make(map[string]int),
	}

	task := ScheduledTask{
		InstanceName: "pbs-1",
		InstanceType: InstanceTypePBS,
		Interval:     0,
		NextRun:      time.Now(),
	}

	key := schedulerKey(task.InstanceType, task.InstanceName)

	m.rescheduleTask(task)

	m.taskQueue.mu.Lock()
	entry, ok := m.taskQueue.entries[key]
	m.taskQueue.mu.Unlock()

	if !ok {
		t.Fatal("expected PBS task to be rescheduled")
	}

	if entry.task.Interval != cfg.PBSPollingInterval {
		t.Errorf("expected PBS interval %v, got %v", cfg.PBSPollingInterval, entry.task.Interval)
	}
}

func TestRescheduleTask_PMGInstance(t *testing.T) {
	cfg := &config.Config{
		PMGPollingInterval:          90 * time.Second,
		AdaptivePollingBaseInterval: 10 * time.Second,
	}

	m := &Monitor{
		config:        cfg,
		taskQueue:     NewTaskQueue(),
		lastOutcome:   make(map[string]taskOutcome),
		failureCounts: make(map[string]int),
	}

	task := ScheduledTask{
		InstanceName: "pmg-1",
		InstanceType: InstanceTypePMG,
		Interval:     0,
		NextRun:      time.Now(),
	}

	key := schedulerKey(task.InstanceType, task.InstanceName)

	m.rescheduleTask(task)

	m.taskQueue.mu.Lock()
	entry, ok := m.taskQueue.entries[key]
	m.taskQueue.mu.Unlock()

	if !ok {
		t.Fatal("expected PMG task to be rescheduled")
	}

	if entry.task.Interval != cfg.PMGPollingInterval {
		t.Errorf("expected PMG interval %v, got %v", cfg.PMGPollingInterval, entry.task.Interval)
	}
}

func TestRescheduleTask_AdaptivePollingMaxIntervalLimit(t *testing.T) {
	cfg := &config.Config{
		PVEPollingInterval:          30 * time.Second,
		AdaptivePollingEnabled:      true,
		AdaptivePollingMaxInterval:  10 * time.Second, // <= 15s enables capping
		AdaptivePollingBaseInterval: 5 * time.Second,
	}

	m := &Monitor{
		config:           cfg,
		taskQueue:        NewTaskQueue(),
		lastOutcome:      make(map[string]taskOutcome),
		failureCounts:    make(map[string]int),
		maxRetryAttempts: 5,
		backoffCfg: backoffConfig{
			Initial:    10 * time.Second, // would normally backoff to 10s+
			Multiplier: 2,
			Jitter:     0,
			Max:        5 * time.Minute,
		},
	}

	task := ScheduledTask{
		InstanceName: "pve-1",
		InstanceType: InstanceTypePVE,
		Interval:     30 * time.Second,
		NextRun:      time.Now(),
	}

	key := schedulerKey(task.InstanceType, task.InstanceName)

	// Simulate transient failure to trigger backoff
	m.failureCounts[key] = 1
	m.lastOutcome[key] = taskOutcome{
		success:   false,
		transient: true,
		err:       errors.New("timeout"),
	}

	m.rescheduleTask(task)

	m.taskQueue.mu.Lock()
	entry, ok := m.taskQueue.entries[key]
	m.taskQueue.mu.Unlock()

	if !ok {
		t.Fatal("expected task to be rescheduled")
	}

	// With AdaptivePollingMaxInterval <= 15s, backoff delay should be capped at 4s
	maxDelay := 4 * time.Second
	if entry.task.Interval > maxDelay {
		t.Errorf("expected backoff interval to be capped at %v, got %v", maxDelay, entry.task.Interval)
	}
}

func TestRescheduleTask_UsesExistingIntervalWhenSet(t *testing.T) {
	cfg := &config.Config{
		PVEPollingInterval:          30 * time.Second,
		AdaptivePollingBaseInterval: 10 * time.Second,
	}

	m := &Monitor{
		config:        cfg,
		taskQueue:     NewTaskQueue(),
		lastOutcome:   make(map[string]taskOutcome),
		failureCounts: make(map[string]int),
	}

	customInterval := 45 * time.Second
	task := ScheduledTask{
		InstanceName: "pve-1",
		InstanceType: InstanceTypePVE,
		Interval:     customInterval, // custom interval already set
		NextRun:      time.Now(),
	}

	key := schedulerKey(task.InstanceType, task.InstanceName)

	m.rescheduleTask(task)

	m.taskQueue.mu.Lock()
	entry, ok := m.taskQueue.entries[key]
	m.taskQueue.mu.Unlock()

	if !ok {
		t.Fatal("expected task to be rescheduled")
	}

	// Should use the existing interval when it's already set
	if entry.task.Interval != customInterval {
		t.Errorf("expected existing interval %v to be preserved, got %v", customInterval, entry.task.Interval)
	}
}

func TestCustomPollProviderIntegration(t *testing.T) {
	const customType InstanceType = "xcp"
	const customName = "xcp-cluster-1"
	customInterval := 42 * time.Second

	var executed atomic.Bool

	monitor := &Monitor{
		taskQueue: NewTaskQueue(),
	}
	if err := monitor.RegisterPollProvider(testPollProvider{
		providerType: customType,
		instances:    []string{customName},
		interval:     customInterval,
		buildPollTask: func(instanceName string) (PollTask, error) {
			return PollTask{
				InstanceName: instanceName,
				InstanceType: string(customType),
				Run: func(context.Context) {
					executed.Store(true)
				},
			}, nil
		},
	}); err != nil {
		t.Fatalf("RegisterPollProvider failed: %v", err)
	}
	monitor.SetExecutor(nil) // restore default executor

	tasks := monitor.buildScheduledTasks(time.Now())
	if len(tasks) != 1 {
		t.Fatalf("expected 1 scheduled task for custom provider, got %d", len(tasks))
	}
	task := tasks[0]
	if task.InstanceType != customType {
		t.Fatalf("expected custom instance type %q, got %q", customType, task.InstanceType)
	}
	if task.InstanceName != customName {
		t.Fatalf("expected custom instance name %q, got %q", customName, task.InstanceName)
	}
	if task.Interval != customInterval {
		t.Fatalf("expected custom interval %v, got %v", customInterval, task.Interval)
	}

	monitor.executeScheduledTask(context.Background(), task)
	if !executed.Load() {
		t.Fatal("expected custom provider poll task callback to execute")
	}
}

func TestAvailabilityPollProviderPublishesSupplementalNetworkEndpoints(t *testing.T) {
	persistence := config.NewConfigPersistence(t.TempDir())
	target := config.NormalizeAvailabilityTarget(config.AvailabilityTarget{
		ID:               "energy-meter",
		Name:             "Energy meter",
		TargetKind:       config.AvailabilityTargetDevice,
		Address:          "192.0.2.44",
		Protocol:         config.AvailabilityProbeICMP,
		Enabled:          true,
		PollIntervalSecs: 30,
		TimeoutMillis:    750,
		FailureThreshold: 2,
	})
	if err := persistence.SaveAvailabilityTargets([]config.AvailabilityTarget{target}); err != nil {
		t.Fatalf("SaveAvailabilityTargets() error = %v", err)
	}

	monitor := &Monitor{
		configPersist: persistence,
		availabilityStatuses: map[string]AvailabilityProbeStatus{
			target.ID: {
				TargetID:            target.ID,
				Name:                target.Name,
				Address:             target.Address,
				Protocol:            string(target.Protocol),
				Enabled:             true,
				Available:           false,
				LastChecked:         time.Date(2026, 5, 6, 12, 0, 0, 0, time.UTC),
				ConsecutiveFailures: 2,
				FailureThreshold:    2,
				LastError:           "icmp probe failed",
			},
		},
	}

	provider := monitor.getPollProvider(InstanceTypeAvailability)
	if provider == nil {
		t.Fatal("expected built-in availability poll provider")
	}
	if got := provider.Type(); got != InstanceTypeAvailability {
		t.Fatalf("provider type = %q, want %q", got, InstanceTypeAvailability)
	}
	if got := provider.ListInstances(monitor); len(got) != 1 || got[0] != target.ID {
		t.Fatalf("instances = %+v, want [%s]", got, target.ID)
	}

	supplemental, ok := provider.(SupplementalRecordsPollProvider)
	if !ok {
		t.Fatalf("availability provider must publish supplemental records")
	}
	if source := supplemental.SupplementalSource(); source != unifiedresources.SourceAvailability {
		t.Fatalf("supplemental source = %q, want %q", source, unifiedresources.SourceAvailability)
	}

	records := supplemental.SupplementalRecords(monitor, "default")
	if len(records) != 1 {
		t.Fatalf("records len = %d, want 1", len(records))
	}
	record := records[0]
	if record.SourceID != target.ID {
		t.Fatalf("record source id = %q, want %q", record.SourceID, target.ID)
	}
	if record.Resource.Type != unifiedresources.ResourceTypeNetworkEndpoint {
		t.Fatalf("resource type = %q, want %q", record.Resource.Type, unifiedresources.ResourceTypeNetworkEndpoint)
	}
	if record.Resource.Availability == nil {
		t.Fatal("expected availability metadata")
	}
	if record.Resource.Availability.TargetID != target.ID {
		t.Fatalf("availability target id = %q, want %q", record.Resource.Availability.TargetID, target.ID)
	}
	if record.Resource.Availability.TargetKind != string(config.AvailabilityTargetDevice) {
		t.Fatalf("availability target kind = %q, want %q", record.Resource.Availability.TargetKind, config.AvailabilityTargetDevice)
	}
	if record.Resource.Status != unifiedresources.StatusOffline {
		t.Fatalf("resource status = %q, want %q", record.Resource.Status, unifiedresources.StatusOffline)
	}
	if len(record.Resource.Incidents) != 1 || record.Resource.Incidents[0].Code != "availability_unreachable" {
		t.Fatalf("expected availability incident, got %+v", record.Resource.Incidents)
	}
}

func TestUpdateResourceStore_IngestsSupplementalRecords(t *testing.T) {
	store := &testSupplementalResourceStore{}
	provider := &testSupplementalPollProvider{
		testPollProvider: testPollProvider{
			providerType: InstanceType("xcp"),
			instances:    []string{"xcp-cluster-1"},
			interval:     30 * time.Second,
		},
		source: unifiedresources.DataSource("xcp"),
		recordsByOrg: map[string][]unifiedresources.IngestRecord{
			"default": {
				{
					SourceID: "xcp-host-1",
					Resource: unifiedresources.Resource{
						Type:     unifiedresources.ResourceTypeAgent,
						Name:     "xcp-host-1",
						Status:   unifiedresources.StatusOnline,
						LastSeen: time.Now().UTC(),
					},
					Identity: unifiedresources.ResourceIdentity{Hostnames: []string{"xcp-host-1"}},
				},
			},
		},
	}

	monitor := &Monitor{
		resourceStore: store,
	}
	if err := monitor.RegisterPollProvider(provider); err != nil {
		t.Fatalf("RegisterPollProvider failed: %v", err)
	}

	monitor.updateResourceStore(models.StateSnapshot{}, monitor.mockModeFence.begin())

	if store.snapshotCalls != 1 {
		t.Fatalf("expected PopulateFromSnapshot to be called once, got %d", store.snapshotCalls)
	}
	if provider.lastRequestedOrg != "default" {
		t.Fatalf("expected default org lookup, got %q", provider.lastRequestedOrg)
	}
	records := store.recordsBySource[unifiedresources.DataSource("xcp")]
	if len(records) != 1 {
		t.Fatalf("expected 1 supplemental record ingested, got %d", len(records))
	}
	if records[0].SourceID != "xcp-host-1" {
		t.Fatalf("expected supplemental source ID xcp-host-1, got %q", records[0].SourceID)
	}
}

func TestUpdateResourceStore_RecordsSupplementalChanges(t *testing.T) {
	store := &testSupplementalResourceStore{}
	occurredAt := time.Date(2026, 3, 30, 18, 30, 0, 0, time.UTC)
	provider := &testSupplementalPollProvider{
		testPollProvider: testPollProvider{
			providerType: InstanceType("vmware"),
			instances:    []string{"vc-1"},
			interval:     30 * time.Second,
		},
		source: unifiedresources.SourceVMware,
		changesByOrg: map[string][]unifiedresources.ResourceChange{
			"default": {
				{
					ID:            "activity-1",
					ResourceID:    "vc-1:vm:vm-201",
					ObservedAt:    occurredAt,
					OccurredAt:    &occurredAt,
					Kind:          unifiedresources.ChangeActivity,
					SourceType:    unifiedresources.SourcePlatformEvent,
					SourceAdapter: unifiedresources.AdapterVMware,
					Confidence:    unifiedresources.ConfidenceHigh,
					Reason:        "Create snapshot (success)",
					Metadata: map[string]any{
						unifiedresources.MetadataActivityType: "vmware_task",
					},
				},
			},
		},
	}

	monitor := &Monitor{
		resourceStore: store,
	}
	if err := monitor.RegisterPollProvider(provider); err != nil {
		t.Fatalf("RegisterPollProvider failed: %v", err)
	}

	monitor.updateResourceStore(models.StateSnapshot{}, monitor.mockModeFence.begin())

	if len(store.recordedChanges) != 1 {
		t.Fatalf("expected 1 supplemental change recorded, got %d", len(store.recordedChanges))
	}
	recorded := store.recordedChanges[0]
	if recorded.Kind != unifiedresources.ChangeActivity {
		t.Fatalf("recorded change kind = %q, want %q", recorded.Kind, unifiedresources.ChangeActivity)
	}
	if recorded.SourceAdapter != unifiedresources.AdapterVMware {
		t.Fatalf("recorded change source adapter = %q, want %q", recorded.SourceAdapter, unifiedresources.AdapterVMware)
	}
	if recorded.OccurredAt == nil || !recorded.OccurredAt.Equal(occurredAt) {
		t.Fatalf("recorded change occurred_at = %v, want %v", recorded.OccurredAt, occurredAt)
	}
	if got := recorded.Metadata[unifiedresources.MetadataActivityType]; got != "vmware_task" {
		t.Fatalf("recorded activity_type = %#v, want vmware_task", got)
	}
}

func TestUpdateResourceStore_SuppressesProviderOwnedSnapshotSources(t *testing.T) {
	store := &testSupplementalResourceStore{}
	provider := &testSupplementalPollProvider{
		testPollProvider: testPollProvider{
			providerType: InstanceType("xcp"),
			instances:    []string{"xcp-cluster-1"},
			interval:     30 * time.Second,
		},
		source:       unifiedresources.SourceProxmox,
		ownedSources: []unifiedresources.DataSource{unifiedresources.SourceProxmox},
		recordsByOrg: map[string][]unifiedresources.IngestRecord{
			"default": {
				{
					SourceID: "xcp-host-1",
					Resource: unifiedresources.Resource{
						Type:     unifiedresources.ResourceTypeAgent,
						Name:     "xcp-host-1",
						Status:   unifiedresources.StatusOnline,
						LastSeen: time.Now().UTC(),
					},
					Identity: unifiedresources.ResourceIdentity{Hostnames: []string{"xcp-host-1"}},
				},
			},
		},
	}

	monitor := &Monitor{
		resourceStore: store,
	}
	if err := monitor.RegisterPollProvider(provider); err != nil {
		t.Fatalf("RegisterPollProvider failed: %v", err)
	}

	monitor.updateResourceStore(models.StateSnapshot{
		Nodes:         []models.Node{{}},
		VMs:           []models.VM{{}},
		Containers:    []models.Container{{}},
		Storage:       []models.Storage{{}},
		PhysicalDisks: []models.PhysicalDisk{{}},
		CephClusters:  []models.CephCluster{{}},
		Hosts:         []models.Host{{}},
	}, monitor.mockModeFence.begin())

	if store.snapshotCalls != 1 {
		t.Fatalf("expected PopulateFromSnapshot to be called once, got %d", store.snapshotCalls)
	}
	if len(store.lastSnapshot.Nodes) != 0 || len(store.lastSnapshot.VMs) != 0 || len(store.lastSnapshot.Containers) != 0 {
		t.Fatalf("expected proxmox compute slices to be suppressed before snapshot ingest")
	}
	if len(store.lastSnapshot.Storage) != 0 || len(store.lastSnapshot.PhysicalDisks) != 0 || len(store.lastSnapshot.CephClusters) != 0 {
		t.Fatalf("expected proxmox storage slices to be suppressed before snapshot ingest")
	}
	if len(store.lastSnapshot.Hosts) != 1 {
		t.Fatalf("expected agent slice to remain in snapshot ingest")
	}

	records := store.recordsBySource[unifiedresources.SourceProxmox]
	if len(records) != 1 {
		t.Fatalf("expected 1 provider-owned supplemental record, got %d", len(records))
	}
}

func TestUpdateResourceStore_IngestsRegisteredSupplementalProvider(t *testing.T) {
	store := &testSupplementalResourceStore{}
	provider := &testMonitorSupplementalProvider{
		recordsByOrg: map[string][]unifiedresources.IngestRecord{
			"default": {
				{
					SourceID: "tn-host-1",
					Resource: unifiedresources.Resource{
						Type:     unifiedresources.ResourceTypeAgent,
						Name:     "tn-host-1",
						Status:   unifiedresources.StatusOnline,
						LastSeen: time.Now().UTC(),
					},
					Identity: unifiedresources.ResourceIdentity{Hostnames: []string{"tn-host-1"}},
				},
			},
		},
	}

	monitor := &Monitor{
		resourceStore: store,
	}
	monitor.SetSupplementalRecordsProvider(unifiedresources.SourceTrueNAS, provider)
	store.snapshotCalls = 0
	store.recordsBySource = nil
	monitor.updateResourceStore(models.StateSnapshot{}, monitor.mockModeFence.begin())

	if store.snapshotCalls != 1 {
		t.Fatalf("expected PopulateFromSnapshot to be called once, got %d", store.snapshotCalls)
	}
	if provider.lastRequestedOrg != "default" {
		t.Fatalf("expected default org lookup, got %q", provider.lastRequestedOrg)
	}
	records := store.recordsBySource[unifiedresources.SourceTrueNAS]
	if len(records) != 1 {
		t.Fatalf("expected 1 supplemental record from direct provider, got %d", len(records))
	}
	if records[0].SourceID != "tn-host-1" {
		t.Fatalf("expected source ID tn-host-1, got %q", records[0].SourceID)
	}
}

func TestUpdateResourceStore_UsesAtomicStoreReplacementWhenAvailable(t *testing.T) {
	store := &testAtomicResourceStore{}
	provider := &testMonitorSupplementalProvider{
		recordsByOrg: map[string][]unifiedresources.IngestRecord{
			"default": {
				{
					SourceID: "tn-host-1",
					Resource: unifiedresources.Resource{
						Type:     unifiedresources.ResourceTypeAgent,
						Name:     "tn-host-1",
						Status:   unifiedresources.StatusOnline,
						LastSeen: time.Now().UTC(),
					},
					Identity: unifiedresources.ResourceIdentity{Hostnames: []string{"tn-host-1"}},
				},
			},
		},
	}

	monitor := &Monitor{
		resourceStore: store,
	}
	monitor.SetSupplementalRecordsProvider(unifiedresources.SourceTrueNAS, provider)
	store.snapshotCalls = 0
	store.atomicCalls = 0
	store.lastSnapshot = models.StateSnapshot{}
	store.lastRecordsBySrc = nil
	snapshot := models.StateSnapshot{
		Hosts: []models.Host{{ID: "host-1", Hostname: "minipc", Status: "online"}},
	}

	monitor.updateResourceStore(snapshot, monitor.mockModeFence.begin())

	if store.atomicCalls != 1 {
		t.Fatalf("expected atomic populate to be called once, got %d", store.atomicCalls)
	}
	if store.snapshotCalls != 0 {
		t.Fatalf("expected legacy PopulateFromSnapshot to be skipped, got %d calls", store.snapshotCalls)
	}
	if len(store.lastSnapshot.Hosts) != 1 || store.lastSnapshot.Hosts[0].Hostname != "minipc" {
		t.Fatalf("expected snapshot to be passed atomically, got %#v", store.lastSnapshot)
	}
	records := store.lastRecordsBySrc[unifiedresources.SourceTrueNAS]
	if len(records) != 1 {
		t.Fatalf("expected 1 atomic supplemental record, got %d", len(records))
	}
	if records[0].SourceID != "tn-host-1" {
		t.Fatalf("expected source ID tn-host-1, got %q", records[0].SourceID)
	}
}

func TestUpdateResourceStore_SuppressesSnapshotForRegisteredSupplementalOwnership(t *testing.T) {
	store := &testSupplementalResourceStore{}
	provider := &testMonitorSupplementalProvider{
		ownedSources: []unifiedresources.DataSource{unifiedresources.SourceProxmox},
		recordsByOrg: map[string][]unifiedresources.IngestRecord{
			"default": {
				{
					SourceID: "tn-host-1",
					Resource: unifiedresources.Resource{
						Type:     unifiedresources.ResourceTypeAgent,
						Name:     "tn-host-1",
						Status:   unifiedresources.StatusOnline,
						LastSeen: time.Now().UTC(),
					},
					Identity: unifiedresources.ResourceIdentity{Hostnames: []string{"tn-host-1"}},
				},
			},
		},
	}

	monitor := &Monitor{
		resourceStore: store,
	}
	monitor.SetSupplementalRecordsProvider(unifiedresources.SourceTrueNAS, provider)
	store.snapshotCalls = 0
	store.recordsBySource = nil
	store.lastSnapshot = models.StateSnapshot{}
	monitor.updateResourceStore(models.StateSnapshot{
		Nodes:      []models.Node{{}},
		VMs:        []models.VM{{}},
		Containers: []models.Container{{}},
		Hosts:      []models.Host{{}},
	}, monitor.mockModeFence.begin())

	if len(store.lastSnapshot.Nodes) != 0 || len(store.lastSnapshot.VMs) != 0 || len(store.lastSnapshot.Containers) != 0 {
		t.Fatalf("expected proxmox slices to be suppressed for direct provider ownership")
	}
	if len(store.lastSnapshot.Hosts) != 1 {
		t.Fatalf("expected non-owned host slice to remain")
	}
	records := store.recordsBySource[unifiedresources.SourceTrueNAS]
	if len(records) != 1 {
		t.Fatalf("expected 1 supplemental record from direct provider, got %d", len(records))
	}
}

func TestSchedulerHealth_UsesProviderInstanceDescriptions(t *testing.T) {
	const customType InstanceType = "xcp"
	monitor := &Monitor{
		config:            &config.Config{},
		instanceInfoCache: make(map[string]*instanceInfo),
	}

	if err := monitor.RegisterPollProvider(testPollProvider{
		providerType: customType,
		instances:    []string{"xcp-a"},
		describeInstances: []PollProviderInstanceInfo{
			{
				Name:        "xcp-a",
				DisplayName: "XCP Cluster A",
				Connection:  "https://xcp-a.example",
			},
		},
		interval: 30 * time.Second,
	}); err != nil {
		t.Fatalf("RegisterPollProvider failed: %v", err)
	}

	resp := monitor.SchedulerHealth()

	key := schedulerKey(customType, "xcp-a")
	for _, inst := range resp.Instances {
		if inst.Key != key {
			continue
		}
		if inst.DisplayName != "XCP Cluster A" {
			t.Fatalf("expected display name %q, got %q", "XCP Cluster A", inst.DisplayName)
		}
		if inst.Connection != "https://xcp-a.example" {
			t.Fatalf("expected connection %q, got %q", "https://xcp-a.example", inst.Connection)
		}
		return
	}

	t.Fatalf("expected instance %q in scheduler health", key)
}

func TestGetConnectionStatuses_CustomProviderStatuses(t *testing.T) {
	const customType InstanceType = "xcp"
	monitor := &Monitor{
		state: models.NewState(),
	}
	if err := monitor.RegisterPollProvider(testPollProvider{
		providerType: customType,
		instances:    []string{"xcp-a"},
		connectionStatus: map[string]bool{
			"xcp-xcp-a": true,
		},
	}); err != nil {
		t.Fatalf("RegisterPollProvider failed: %v", err)
	}

	statuses := monitor.GetConnectionStatuses()
	if connected, ok := statuses["xcp-xcp-a"]; !ok || !connected {
		t.Fatalf("expected xcp-xcp-a to be connected, got exists=%v value=%v", ok, connected)
	}
}

func TestGetConnectionStatuses_CustomProviderFallbackToState(t *testing.T) {
	const customType InstanceType = "xcp"
	monitor := &Monitor{
		state: models.NewState(),
	}
	monitor.state.SetConnectionHealth("xcp-xcp-a", true)

	if err := monitor.RegisterPollProvider(pollProviderAdapter{
		instanceType: customType,
		listInstances: func(*Monitor) []string {
			return []string{"xcp-a"}
		},
		baseInterval: func(*Monitor) time.Duration { return 30 * time.Second },
		buildPollTask: func(*Monitor, string) (PollTask, error) {
			return PollTask{}, nil
		},
	}); err != nil {
		t.Fatalf("RegisterPollProvider failed: %v", err)
	}

	statuses := monitor.GetConnectionStatuses()
	if connected, ok := statuses["xcp-xcp-a"]; !ok || !connected {
		t.Fatalf("expected xcp-xcp-a to be connected via fallback, got exists=%v value=%v", ok, connected)
	}
}

func TestGetConnectionStatuses_BuiltInPMGSupport(t *testing.T) {
	monitor := &Monitor{
		config: &config.Config{
			PMGInstances: []config.PMGInstance{
				{Name: "pmg-1"},
				{Name: "pmg-2"},
			},
		},
		state:      models.NewState(),
		pmgClients: map[string]*pmg.Client{"pmg-1": {}},
	}
	monitor.state.SetConnectionHealth("pmg-pmg-1", true)

	statuses := monitor.GetConnectionStatuses()
	if connected, ok := statuses["pmg-pmg-1"]; !ok || !connected {
		t.Fatalf("expected pmg-pmg-1 connected, got exists=%v value=%v", ok, connected)
	}
	if connected, ok := statuses["pmg-pmg-2"]; !ok || connected {
		t.Fatalf("expected pmg-pmg-2 disconnected, got exists=%v value=%v", ok, connected)
	}
}

func TestSetProviderConnectionHealth_UsesProviderConnectionKey(t *testing.T) {
	const customType InstanceType = "xcp"
	const instanceName = "xcp-a"
	const providerKey = "provider/xcp-a"

	monitor := &Monitor{
		state: models.NewState(),
	}
	if err := monitor.RegisterPollProvider(testPollProvider{
		providerType:  customType,
		instances:     []string{instanceName},
		connectionKey: providerKey,
	}); err != nil {
		t.Fatalf("RegisterPollProvider failed: %v", err)
	}

	monitor.setProviderConnectionHealth(customType, instanceName, true)

	if !monitor.state.ConnectionHealth[providerKey] {
		t.Fatalf("expected provider key %q to be marked healthy", providerKey)
	}
	if _, exists := monitor.state.ConnectionHealth["xcp-"+instanceName]; exists {
		t.Fatalf("did not expect fallback key %q when provider key override is set", "xcp-"+instanceName)
	}
}