Wire the provider health monitor into the manager, with gauges

--provider-check-interval (1m; 0 = off, else at least 10s) and
--provider-check-timeout (15s) are validated before the providers config
loads. Checkers are collected from the unwrapped providers in config
order, so checks never count as provider requests. Three scrape-time
gauges per {provider,target} read the monitor's snapshot: up, check
duration and check timestamp. The e2e suite waits for the kubernetes
ProviderHealth to turn Ready and asserts its gauge in the scrape.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

internal/metrics/metrics.go

@@ -117,3 +117,51 @@ func (c *constCollector) Collect(ch chan<- prometheus.Metric) {
 		ch <- prometheus.MustNewConstMetric(c.desc, prometheus.GaugeValue, float64(value), label)
 	}
 }
+
+// ProviderTarget is one provider health-check target as the gauges see it.
+type ProviderTarget struct {
+	Provider, Target string
+	Up               bool
+	Duration         time.Duration
+	CheckedAt        time.Time
+}
+
+// RegisterProviderTargets registers the provider-check gauges. Like the
+// other scrape-time collectors they are read from the source at every
+// scrape, so a target that goes away stops being reported rather than
+// leaving a stale series behind.
+func RegisterProviderTargets(reg prometheus.Registerer, read func() []ProviderTarget) error {
+	labels := []string{"provider", "target"}
+	return reg.Register(&providerTargetCollector{
+		up: prometheus.NewDesc("proxy_operator_provider_target_up",
+			"Whether the target passed its latest provider health check (1) or not (0).", labels, nil),
+		duration: prometheus.NewDesc("proxy_operator_provider_check_duration_seconds",
+			"Duration of the target's latest provider health check.", labels, nil),
+		checked: prometheus.NewDesc("proxy_operator_provider_check_timestamp_seconds",
+			"Unix time of the target's latest provider health check.", labels, nil),
+		read: read,
+	})
+}
+
+type providerTargetCollector struct {
+	up, duration, checked *prometheus.Desc
+	read                  func() []ProviderTarget
+}
+
+func (c *providerTargetCollector) Describe(ch chan<- *prometheus.Desc) {
+	ch <- c.up
+	ch <- c.duration
+	ch <- c.checked
+}
+
+func (c *providerTargetCollector) Collect(ch chan<- prometheus.Metric) {
+	for _, t := range c.read() {
+		up := 0.0
+		if t.Up {
+			up = 1
+		}
+		ch <- prometheus.MustNewConstMetric(c.up, prometheus.GaugeValue, up, t.Provider, t.Target)
+		ch <- prometheus.MustNewConstMetric(c.duration, prometheus.GaugeValue, t.Duration.Seconds(), t.Provider, t.Target)
+		ch <- prometheus.MustNewConstMetric(c.checked, prometheus.GaugeValue, float64(t.CheckedAt.UnixNano())/1e9, t.Provider, t.Target)
+	}
+}

internal/metrics/metrics_test.go

@@ -111,3 +111,45 @@ func TestRegister_freshRegistryPerTest(t *testing.T) {
 	register(t, m1, nil, 0)
 	register(t, m2, nil, 0)
 }
+
+func TestRegisterProviderTargets_scrape(t *testing.T) {
+	t.Parallel()
+	checked := time.Unix(1757602800, 0)
+	targets := []ProviderTarget{
+		{Provider: "kvm-home", Target: "nas1", Up: true, Duration: 84 * time.Millisecond, CheckedAt: checked},
+		{Provider: "kvm-home", Target: "nas2", Up: false, Duration: 15 * time.Second, CheckedAt: checked},
+	}
+	reg := prometheus.NewRegistry()
+	if err := RegisterProviderTargets(reg, func() []ProviderTarget { return targets }); err != nil {
+		t.Fatalf("RegisterProviderTargets: %v", err)
+	}
+
+	expected := `
+# HELP proxy_operator_provider_check_duration_seconds Duration of the target's latest provider health check.
+# TYPE proxy_operator_provider_check_duration_seconds gauge
+proxy_operator_provider_check_duration_seconds{provider="kvm-home",target="nas1"} 0.084
+proxy_operator_provider_check_duration_seconds{provider="kvm-home",target="nas2"} 15
+# HELP proxy_operator_provider_check_timestamp_seconds Unix time of the target's latest provider health check.
+# TYPE proxy_operator_provider_check_timestamp_seconds gauge
+proxy_operator_provider_check_timestamp_seconds{provider="kvm-home",target="nas1"} 1.7576028e+09
+proxy_operator_provider_check_timestamp_seconds{provider="kvm-home",target="nas2"} 1.7576028e+09
+# HELP proxy_operator_provider_target_up Whether the target passed its latest provider health check (1) or not (0).
+# TYPE proxy_operator_provider_target_up gauge
+proxy_operator_provider_target_up{provider="kvm-home",target="nas1"} 1
+proxy_operator_provider_target_up{provider="kvm-home",target="nas2"} 0
+`
+	if err := testutil.GatherAndCompare(reg, strings.NewReader(expected)); err != nil {
+		t.Error(err)
+	}
+}
+
+func TestRegisterProviderTargets_emptySnapshotHasNoSeries(t *testing.T) {
+	t.Parallel()
+	reg := prometheus.NewRegistry()
+	if err := RegisterProviderTargets(reg, func() []ProviderTarget { return nil }); err != nil {
+		t.Fatalf("RegisterProviderTargets: %v", err)
+	}
+	if n, err := testutil.GatherAndCount(reg); err != nil || n != 0 {
+		t.Errorf("series before the first check = %d (%v), want 0", n, err)
+	}
+}