channelhealth: controller->agent channel health-check (periodic probe + classified operator alert) v0.90.0

New internal/channelhealth Checker: ~60s probe via the PRODUCTION memoized client
(Server.ProbeAgentChannel, GET /storage), classifies failures (spike Q1 map), debounces transient
reasons (N>=2; construction error latches distinctly), seeds first obs, alerts operator+dashboard on
transition. Notifier.NotifyAgentChannelDown/Recovered (English, operator-only), AlertManager dashboard
banner (Hungarian). No agent/hub change. Spike-proven.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EPZ4GJ8L5Jqf8UiPwbn1kt

controller/internal/web/agent_disk_handlers.go

@@ -1,6 +1,7 @@
 package web
 
 import (
+	"context"
 	"encoding/json"
 	"errors"
 	"net/http"
@@ -59,6 +60,22 @@ func (s *Server) agentClient() (*agentapi.Client, error) {
 	return s.agentCli, s.agentCliErr
 }
 
+// ProbeAgentChannel runs one controller→agent channel health probe using the PRODUCTION memoized
+// client (NOT a fresh one — spike SPIKE-controller-agent-channel-health-2026-06-29: it self-heals,
+// reflects exactly what the disk UI sees, and avoids the per-call transport leak the singleton fixed).
+// It returns whether the failure was a CONSTRUCTION error (agentClient() couldn't build — a latching
+// config fault, distinct from a runtime channel failure) and the error (nil = channel up). The probe
+// is GET /storage (cheap, read-only — the same call probeLocalAPI uses at startup). This is the
+// channelhealth.Probe seam.
+func (s *Server) ProbeAgentChannel(ctx context.Context) (constructionErr bool, err error) {
+	client, cerr := s.agentClient()
+	if cerr != nil {
+		return true, cerr
+	}
+	_, serr := client.Storage(ctx)
+	return false, serr
+}
+
 // writeDiskJSON writes the standard {ok,data,error} envelope used by the disk API.
 func writeDiskJSON(w http.ResponseWriter, status int, ok bool, errMsg string, data interface{}) {
 	w.Header().Set("Content-Type", "application/json")

controller/internal/web/alerts.go

@@ -33,6 +33,10 @@ type AlertManager struct {
 	alerts          []Alert
 	logger          *log.Logger
 	hubPushStatusFn func() HubPushStatusData
+	// agentChannelAlert is set/cleared by the channel-health checker (out-of-band from Refresh, which
+	// is health-report-driven). nil = channel up. It is prepended in GetAlerts so a dead controller→
+	// agent link (disk/storage UI broken) is always visible regardless of the health-report cycle.
+	agentChannelAlert *Alert
 }
 
 // NewAlertManager creates a new AlertManager.
@@ -49,6 +53,25 @@ func (am *AlertManager) SetHubPushStatus(fn func() HubPushStatusData) {
 	am.mu.Unlock()
 }
 
+// SetAgentChannelAlert sets (down=true) or clears (down=false) the controller→agent channel-down
+// dashboard banner. Called by the channel-health checker each probe (idempotent). msg is the short
+// Hungarian display line.
+func (am *AlertManager) SetAgentChannelAlert(down bool, msg string) {
+	am.mu.Lock()
+	defer am.mu.Unlock()
+	if !down {
+		am.agentChannelAlert = nil
+		return
+	}
+	am.agentChannelAlert = &Alert{
+		ID:       "agent-channel-down",
+		Level:    "error",
+		Message:  msg,
+		Link:     "/settings",
+		LinkText: "Beállítások",
+	}
+}
+
 // Refresh regenerates alerts from the latest health check report and config state.
 // Called after each health check cycle (every 5 minutes) and on storage state changes.
 func (am *AlertManager) Refresh(report *monitor.HealthReport, cfg *config.Config, backupMgr *backup.Manager, updateAvailable bool, latestVersion string, storagePaths ...[]settings.StoragePath) {
@@ -159,7 +182,7 @@ func (am *AlertManager) GetAlerts(excludeIDs ...string) []Alert {
 	am.mu.RLock()
 	defer am.mu.RUnlock()
 
-	if len(am.alerts) == 0 {
+	if len(am.alerts) == 0 && am.agentChannelAlert == nil {
 		return nil
 	}
 
@@ -169,6 +192,10 @@ func (am *AlertManager) GetAlerts(excludeIDs ...string) []Alert {
 	}
 
 	var result []Alert
+	// Channel-down is prepended (highest priority — the agent link being dead breaks disk/storage UI).
+	if am.agentChannelAlert != nil && !exclude[am.agentChannelAlert.ID] {
+		result = append(result, *am.agentChannelAlert)
+	}
 	for _, a := range am.alerts {
 		if exclude[a.ID] {
 			continue
@@ -239,4 +266,3 @@ func countLevel(alerts []Alert, level string) int {
 	}
 	return n
 }
-