Extended ZFS states

README.md

@@ -69,7 +69,8 @@ unusable subscription.
 The ZFS collector discovers all pools on every node and exports their recursive
 topology with `cluster`, `node`, `pool`, `component`, `path`, and `leaf` labels:
 
-- `pve_node_zfs_state` (`1` for `ONLINE`, otherwise `0`)
+- `pve_node_zfs_state` (`0=UNKNOWN`, `1=ONLINE`, `2=DEGRADED`, `3=FAULTED`,
+  `4=OFFLINE`, `5=REMOVED`, `6=UNAVAIL`)
 - `pve_node_zfs_read_errors`
 - `pve_node_zfs_write_errors`
 - `pve_node_zfs_checksum_errors`

metrics/pve_node_zfs_collector.go

@@ -26,17 +26,26 @@ type zfsMetricComponent struct {
 	Path string
 }
 
+var zfsStateValues = map[string]float64{
+	"ONLINE":   1,
+	"DEGRADED": 2,
+	"FAULTED":  3,
+	"OFFLINE":  4,
+	"REMOVED":  5,
+	"UNAVAIL":  6,
+}
+
 // NewPveNodeZfsCollector creates a ZFS metrics collector.
 func NewPveNodeZfsCollector(apiClient *proxmox.PveApiClient, registry *TTLRegistry) *PveNodeZfsCollector {
 	c := PveNodeZfsCollector{apiClient: apiClient, registry: registry}
-	labelNames := []string{"cluster", "node", "pool", "component", "path", "leaf"}
+	componentLabelNames := []string{"cluster", "node", "pool", "component", "path", "leaf"}
 
 	c.state = NewTTLGaugeVec(
 		prometheus.GaugeOpts{
 			Name: "pve_node_zfs_state",
-			Help: "ZFS pool component state (1 = ONLINE, 0 = any other state).",
+			Help: "ZFS pool component state (0 = UNKNOWN, 1 = ONLINE, 2 = DEGRADED, 3 = FAULTED, 4 = OFFLINE, 5 = REMOVED, 6 = UNAVAIL).",
 		},
-		labelNames,
+		componentLabelNames,
 		5*time.Minute,
 	)
 	c.registry.Register(c.state)
@@ -46,7 +55,7 @@ func NewPveNodeZfsCollector(apiClient *proxmox.PveApiClient, registry *TTLRegist
 			Name: "pve_node_zfs_read_errors",
 			Help: "ZFS pool component read error count.",
 		},
-		labelNames,
+		componentLabelNames,
 		5*time.Minute,
 	)
 	c.registry.Register(c.readErrors)
@@ -56,7 +65,7 @@ func NewPveNodeZfsCollector(apiClient *proxmox.PveApiClient, registry *TTLRegist
 			Name: "pve_node_zfs_write_errors",
 			Help: "ZFS pool component write error count.",
 		},
-		labelNames,
+		componentLabelNames,
 		5*time.Minute,
 	)
 	c.registry.Register(c.writeErrors)
@@ -66,7 +75,7 @@ func NewPveNodeZfsCollector(apiClient *proxmox.PveApiClient, registry *TTLRegist
 			Name: "pve_node_zfs_checksum_errors",
 			Help: "ZFS pool component checksum error count.",
 		},
-		labelNames,
+		componentLabelNames,
 		5*time.Minute,
 	)
 	c.registry.Register(c.checksumErrors)
@@ -146,8 +155,8 @@ func appendZfsComponent(components *[]zfsMetricComponent, component proxmox.PveZ
 }
 
 func zfsStateNumeric(state string) float64 {
-	if strings.EqualFold(state, "ONLINE") {
-		return 1
+	if value, ok := zfsStateValues[strings.ToUpper(state)]; ok {
+		return value
 	}
 	return 0
 }

metrics/pve_node_zfs_collector_test.go

@@ -55,10 +55,24 @@ func TestFlattenZfsComponents(t *testing.T) {
 }
 
 func TestZfsStateNumeric(t *testing.T) {
-	if got := zfsStateNumeric("ONLINE"); got != 1 {
-		t.Fatalf("zfsStateNumeric(ONLINE) = %v, want 1", got)
+	tests := []struct {
+		state string
+		want  float64
+	}{
+		{state: "UNKNOWN", want: 0},
+		{state: "ONLINE", want: 1},
+		{state: "DEGRADED", want: 2},
+		{state: "FAULTED", want: 3},
+		{state: "OFFLINE", want: 4},
+		{state: "REMOVED", want: 5},
+		{state: "UNAVAIL", want: 6},
 	}
-	if got := zfsStateNumeric("DEGRADED"); got != 0 {
-		t.Fatalf("zfsStateNumeric(DEGRADED) = %v, want 0", got)
+
+	for _, tt := range tests {
+		t.Run(tt.state, func(t *testing.T) {
+			if got := zfsStateNumeric(tt.state); got != tt.want {
+				t.Fatalf("zfsStateNumeric(%q) = %v, want %v", tt.state, got, tt.want)
+			}
+		})
 	}
 }