When deploying the New Relic Kubernetes integration for the first time in an account, we deploy a default set of alert conditions to your account. The predefined alert policy, named Kubernetes default alert policy, doesn't have a notification channel by default to avoid unwanted notifications.
The alert conditions' thresholds can be customized to your environment and the alert policy updated to send notifications. For more information, see the Infrastructure alerts documentation.
Predefined alert conditions
- Container CPU usage % is too high
-
Setting Value Event type K8sContainerSample
SELECT value (cpuUsedCores/cpuLimitCores)*100
Warning threshold >90% for at least 5 minutes Critical threshold >95% for at least 5 mins - Container memory usage % is too high
-
Setting Value Event type K8sContainerSample
SELECT value (memoryUsedBytes/memoryLimitBytes)/100
Warning threshold >85% for at least 5 minutes Critical threshold >95% for at least 5 mins - Pod was unable to be scheduled
-
Setting Value Event type K8sPodSample
SELECT value isScheduled
Warning threshold Critical threshold isScheduled
= 0 for at least 7 minutes - Pod is not ready
-
Setting Value Event type K8sPodSample
SELECT value isReady
Warning threshold Critical threshold isReady = 0
for at least 10 minutes - Container is running out of space
-
Setting Value Event type K8sContainerSample
SELECT value fsUsedPercent
Warning threshold >75% for at least 5 minutes Critical threshold >90% for at least 5 minutes - ReplicaSet doesn't have desired amount of pods
-
Setting Value Event type K8sReplicaSetSample
SELECT value podsDesired - podsReady
Warning threshold Critical threshold 0
for at least 5 minutes - ETCD open file descriptors
-
Setting Value Event type K8sEtcdSample
SELECT value (processOpenFds/processMaxFds)*100
Warning threshold >75% for at least 3 minutes Critical threshold >90% for at least 5 minutes
Create new alert conditions
To create new alert conditions based on Kubernetes metric data, see Understand and use data.