PrometheusDown#
Triggers when Prometheus it not running.
Troubleshooting Hints#
Prometheus state#
kubectl get pods -n <monitoring_namespace> | grep prometheus # check pod is
kubectl logs prometheus-<name>-prometheus-0 -n <monitoring_namespace>
Tip
Where <monitoring_namespace> and <name>can be:
Notice the namespace and pod label in the alert
- monitoring_namespace:
caascad-monitoringname:caascad
- monitoring_namespace:
monitoringname:caascad - monitoring_namespace:
monitoring-clientname:client - monitoring_namespace:
monitoring-appname:app
- monitoring_namespace:
monitoringname:caascad - monitoring_namespace:
monitoring-consumptionname:consumption
Check if the pod is up and running or if it is restarting in loop (check the containers restart number) and see its state. Usually a
state: Terminated
reason: error
exit code: 137
indicates an OOMKILLED in which case you should check the limit requests CPU/MEM.
Federation problems (NGOT)#
This alert checks the metric prometheus_build_info. This alert is not sent from Prometheus (both "cluster" and "client") but is federated from central Prometheus on kub-53/mon3 and kub-34/mon4 (or, on Staging, kub-10002/mon3 and kub-10010/mon4).
Checks:
- Check if Prometheus (cluster) in
monitoringnamespace is running. - Check if Prometheus (central) in
monitoring-stack-corp-obs-corp-prd(ormonitoring-stack-corp-obs-corp-stg) on kub-53/mon3 and kub-34/mon4 (or kub-10002/mon3 and kub-10010/mon4) are running. - Check in Prometheus (central) the targets (do a port forward, more information here). Check if the target exists. Check its state. If the target is unhealthy, Prometheus may provide an explanation.
- Check if the ScrapeConfig
federate-metrics-momthat defines the federation exists and is correctly defined. - Check the Secret
federate-metrics-momthat defines the credentials to connect to Prometheus (cluster): is it correctly defined ? Are these the correct credentials ?
Tip
The secret is defined in Vault in secret/applications/prometheus/cluster-federate.