fix: export domain info metrics for all domain states

Previously, libvirt_domain_info_state and related info metrics were only
exported for running domains. This was because DomainGetInfo() fails for
inactive domains, causing the CollectDomain function to exit early before
emitting any metrics.

This change ensures that domain info and state metrics are always exported
regardless of domain state by:

1. Moving domain info metric emission before DomainGetInfo() call
2. Adding graceful fallback to DomainGetState() when DomainGetInfo() fails
3. Always emitting libvirt_domain_info_state for complete domain visibility

Now users can monitor all domains (running, stopped, paused, crashed, etc.)
and get proper domain state information for inventory and alerting purposes.

Fixes missing metrics for inactive domains while preserving existing
behavior for performance metrics that only apply to running domains.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: ADometrius

pkg/exporter/prometheus-libvirt-exporter.go

@@ -609,34 +609,11 @@ func CollectDomain(ch chan<- prometheus.Metric, l *libvirt.Libvirt, domain domai
 		rState                     uint8
 		rvirCpu                    uint16
 		rmaxmem, rmemory, rcputime uint64
+		domainInfoAvailable        bool = true
 	)
-	type rDomainStatsState struct {
-		rState                     uint8
-		rvirCpu                    uint16
-		rmaxmem, rmemory, rcputime uint64
-		err                        error
-	}
-
-	chDomainStats := make(chan rDomainStatsState, 1)
-	go func() {
-		var data rDomainStatsState
-		data.rState, data.rmaxmem, data.rmemory, data.rvirCpu, data.rcputime, data.err = l.DomainGetInfo(domain.libvirtDomain)
-		chDomainStats <- data
-	}()
 
-	select {
-	case res := <-chDomainStats:
-		if res.err != nil {
-			return res.err, false
-		}
-
-		rState = res.rState
-		rvirCpu = res.rvirCpu
-		rmaxmem = res.rmaxmem
-		rmemory = res.rmemory
-		rcputime = res.rcputime
-	case <-time.After(time.Duration(timeout) * time.Second):
-		return fmt.Errorf("call to DomainGetInfo has timed out"), true
+	promLabels := []string{
+		domain.domainName,
 	}
 
 	openstackInfoLabels := []string{
@@ -657,19 +634,79 @@ func CollectDomain(ch chan<- prometheus.Metric, l *libvirt.Libvirt, domain domai
 		domain.os_type_machine,
 	}
 
-	promLabels := []string{
-		domain.domainName,
-	}
-
+	// Always emit these info metrics for all domains, regardless of state
 	ch <- prometheus.MustNewConstMetric(libvirtDomainInfoDesc, prometheus.GaugeValue, 1.0, infoLabels...)
 	ch <- prometheus.MustNewConstMetric(libvirtDomainOpenstackInfoDesc, prometheus.GaugeValue, 1.0, openstackInfoLabels...)
 
+	// Try to get domain info, but don't fail completely if it doesn't work
+	type rDomainStatsState struct {
+		rState                     uint8
+		rvirCpu                    uint16
+		rmaxmem, rmemory, rcputime uint64
+		err                        error
+	}
+
+	chDomainStats := make(chan rDomainStatsState, 1)
+	go func() {
+		var data rDomainStatsState
+		data.rState, data.rmaxmem, data.rmemory, data.rvirCpu, data.rcputime, data.err = l.DomainGetInfo(domain.libvirtDomain)
+		chDomainStats <- data
+	}()
+
+	select {
+	case res := <-chDomainStats:
+		if res.err != nil {
+			logger.Debug("DomainGetInfo failed, will attempt to get domain state separately", "domain", domain.domainName, "error", res.err)
+			domainInfoAvailable = false
+		} else {
+			rState = res.rState
+			rvirCpu = res.rvirCpu
+			rmaxmem = res.rmaxmem
+			rmemory = res.rmemory
+			rcputime = res.rcputime
+		}
+	case <-time.After(time.Duration(timeout) * time.Second):
+		logger.Debug("DomainGetInfo timed out, will attempt to get domain state separately", "domain", domain.domainName)
+		domainInfoAvailable = false
+	}
+
+	// If DomainGetInfo failed, try to get just the domain state
+	if !domainInfoAvailable {
+		type rDomainState struct {
+			state int32
+			err   error
+		}
+		chDomainState := make(chan rDomainState, 1)
+		go func() {
+			var data rDomainState
+			var reason int32
+			data.state, reason, data.err = l.DomainGetState(domain.libvirtDomain, 0)
+			_ = reason // ignore reason for now
+			chDomainState <- data
+		}()
+
+		select {
+		case res := <-chDomainState:
+			if res.err != nil {
+				logger.Error("failed to get domain state", "domain", domain.domainName, "error", res.err)
+				return res.err, false
+			}
+			rState = uint8(res.state)
+		case <-time.After(time.Duration(timeout) * time.Second):
+			return fmt.Errorf("call to DomainGetState has timed out"), true
+		}
+	}
+
+	// Always emit domain state - this is the key metric for all domains
 	ch <- prometheus.MustNewConstMetric(libvirtDomainState, prometheus.GaugeValue, float64(rState), append(promLabels, domainState[libvirt_schema.DomainState(rState)])...)
 
-	ch <- prometheus.MustNewConstMetric(libvirtDomainInfoMaxMemDesc, prometheus.GaugeValue, float64(rmaxmem)*1024, promLabels...)
-	ch <- prometheus.MustNewConstMetric(libvirtDomainInfoMemoryDesc, prometheus.GaugeValue, float64(rmemory)*1024, promLabels...)
-	ch <- prometheus.MustNewConstMetric(libvirtDomainInfoNrVirtCpuDesc, prometheus.GaugeValue, float64(rvirCpu), promLabels...)
-	ch <- prometheus.MustNewConstMetric(libvirtDomainInfoCpuTimeDesc, prometheus.CounterValue, float64(rcputime)/1e9, promLabels...)
+	// Only emit detailed domain info metrics if DomainGetInfo succeeded
+	if domainInfoAvailable {
+		ch <- prometheus.MustNewConstMetric(libvirtDomainInfoMaxMemDesc, prometheus.GaugeValue, float64(rmaxmem)*1024, promLabels...)
+		ch <- prometheus.MustNewConstMetric(libvirtDomainInfoMemoryDesc, prometheus.GaugeValue, float64(rmemory)*1024, promLabels...)
+		ch <- prometheus.MustNewConstMetric(libvirtDomainInfoNrVirtCpuDesc, prometheus.GaugeValue, float64(rvirCpu), promLabels...)
+		ch <- prometheus.MustNewConstMetric(libvirtDomainInfoCpuTimeDesc, prometheus.CounterValue, float64(rcputime)/1e9, promLabels...)
+	}
 
 	var isActive int32
 	if isActive, err = l.DomainIsActive(domain.libvirtDomain); err != nil {