add support to show overall cluster maintenance mode (#260)

Co-authored-by: Stefano Torresi <[email protected]>
ClusterLabs · Aug 2, 2024 · e26395b · e26395b
1 parent d0d3705
commit e26395b
Show file tree

Hide file tree

Showing 4 changed files with 17 additions and 1 deletion.
diff --git a/collector/pacemaker/crmmon/data.go b/collector/pacemaker/crmmon/data.go
@@ -17,7 +17,8 @@ type Root struct {
 			Blocked  int `xml:"blocked,attr"`
 		} `xml:"resources_configured"`
 		ClusterOptions struct {
-			StonithEnabled bool `xml:"stonith-enabled,attr"`
+			StonithEnabled  bool `xml:"stonith-enabled,attr"`
+			MaintenanceMode bool `xml:"maintenance-mode,attr"`
 		} `xml:"cluster_options"`
 	} `xml:"summary"`
 	Nodes          []Node `xml:"nodes>node"`

diff --git a/collector/pacemaker/pacemaker.go b/collector/pacemaker/pacemaker.go
@@ -33,6 +33,7 @@ func NewCollector(crmMonPath string, cibAdminPath string, timestamps bool, logge
 	c.SetDescriptor("node_attributes", "Metadata attributes of each node; value is always 1", []string{"node", "name", "value"})
 	c.SetDescriptor("resources", "The status of each resource in the cluster; 1 means the resource is in that status, 0 otherwise", []string{"node", "resource", "role", "managed", "status", "agent", "group", "clone"})
 	c.SetDescriptor("stonith_enabled", "Whether or not stonith is enabled", nil)
+	c.SetDescriptor("maintenance_mode_enabled", "Whether or not cluster wide maintenance-mode is enabled", nil)
 	c.SetDescriptor("fail_count", "The Fail count number per node and resource id", []string{"node", "resource"})
 	c.SetDescriptor("migration_threshold", "The migration_threshold number per node and resource id", []string{"node", "resource"})
 	c.SetDescriptor("config_last_change", "The timestamp of the last change of the cluster configuration", nil)
@@ -61,6 +62,7 @@ func (c *pacemakerCollector) CollectWithError(ch chan<- prometheus.Metric) error
 	}
 
 	c.recordStonithStatus(crmMon, ch)
+	c.recordMaintenanceModeStatus(crmMon, ch)
 	c.recordNodes(crmMon, ch)
 	c.recordNodeAttributes(crmMon, ch)
 	c.recordResources(crmMon, ch)
@@ -85,6 +87,15 @@ func (c *pacemakerCollector) Collect(ch chan<- prometheus.Metric) {
 	}
 }
 
+func (c *pacemakerCollector) recordMaintenanceModeStatus(crmMon crmmon.Root, ch chan<- prometheus.Metric) {
+	var maintenanceModeEnabled float64
+	if crmMon.Summary.ClusterOptions.MaintenanceMode {
+		maintenanceModeEnabled = 1
+	}
+
+	ch <- c.MakeGaugeMetric("maintenance_mode_enabled", maintenanceModeEnabled)
+}
+
 func (c *pacemakerCollector) recordStonithStatus(crmMon crmmon.Root, ch chan<- prometheus.Metric) {
 	var stonithEnabled float64
 	if crmMon.Summary.ClusterOptions.StonithEnabled {

diff --git a/go.sum b/go.sum
@@ -122,6 +122,7 @@ golang.org/x/text v0.16.0/go.mod h1:GhwF1Be+LQoKShO3cGOHzqOgRrGaYc9AvblQOmPVHnI=
 golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20191119224855-298f0cb1881e/go.mod h1:b+2E5dAYhXwXZwtnZ6UAqBI28+e2cm9otk0dWdXHAEo=
 golang.org/x/tools v0.1.1/go.mod h1:o0xws9oXOQQZyjljx8fwUC0k7L1pTE6eaCbjGeHmOkk=
+golang.org/x/tools v0.6.0 h1:BOw41kyTf3PuCW1pVQf8+Cyg8pMlkYB1oo9iJ6D/lKM=
 golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191011141410-1b5146add898/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20200804184101-5ec99f83aff1/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=

diff --git a/test/pacemaker.metrics b/test/pacemaker.metrics
@@ -161,3 +161,6 @@ ha_cluster_pacemaker_resources{agent="stonith:external/sbd",clone="",group="",ma
 # HELP ha_cluster_pacemaker_stonith_enabled Whether or not stonith is enabled
 # TYPE ha_cluster_pacemaker_stonith_enabled gauge
 ha_cluster_pacemaker_stonith_enabled 1
+# HELP ha_cluster_pacemaker_maintenance_mode_enabled Whether or not cluster wide maintenance-mode is enabled
+# TYPE ha_cluster_pacemaker_maintenance_mode_enabled gauge
+ha_cluster_pacemaker_maintenance_mode_enabled 0