Merge pull request #96 from aruniiird/BZ-2014167

Fixing the queries for alerts 'CephMgrIsAbsent' and 'CephMgrIsMissingReplicas'
ceph · Feb 2, 2022 · eeb4045 · eeb4045
2 parents 26c9ec5 + a32a4c3
commit eeb4045
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 4 deletions.
diff --git a/alerts/absent_alerts.libsonnet b/alerts/absent_alerts.libsonnet
@@ -7,7 +7,7 @@
           {
             alert: 'CephMgrIsAbsent',
             expr: |||
-              up{%(cephExporterSelector)s} == 0
+              label_replace((up{%(cephExporterSelector)s} == 0 or absent(up{%(cephExporterSelector)s})), "namespace", "openshift-storage", "", "")
             ||| % $._config,
             'for': $._config.mgrIsAbsentAlertTime,
             labels: {
@@ -23,7 +23,7 @@
           {
             alert: 'CephMgrIsMissingReplicas',
             expr: |||
-              sum(up{%(cephExporterSelector)s}) by (namespace) < %(cephMgrCount)d
+              sum(kube_deployment_spec_replicas{deployment=~"rook-ceph-mgr-.*"}) by (namespace) < %(cephMgrCount)d
             ||| % $._config,
             'for': $._config.mgrMissingReplicasAlertTime,
             labels: {

diff --git a/extras/manifests/prometheus-ceph-rules.yaml b/extras/manifests/prometheus-ceph-rules.yaml
@@ -42,7 +42,7 @@ spec:
         severity_level: critical
         storage_type: ceph
       expr: |
-        up{job="rook-ceph-mgr"} == 0
+        label_replace((up{job="rook-ceph-mgr"} == 0 or absent(up{job="rook-ceph-mgr"})), "namespace", "openshift-storage", "", "")
       for: 5m
       labels:
         severity: critical
@@ -53,7 +53,7 @@ spec:
         severity_level: warning
         storage_type: ceph
       expr: |
-        sum(up{job="rook-ceph-mgr"}) by (namespace) < 1
+        sum(kube_deployment_spec_replicas{deployment=~"rook-ceph-mgr-.*"}) by (namespace) < 1
       for: 5m
       labels:
         severity: warning