Default resource limits for clusters lead to OOMKilled for client-pods with default image
mbauer83 opened this issue · comments
Michael Bauer commented
When running the example configuration (azure) against a cluster with sufficient available resources, the resource limits given in the example configuration lead to CrashLoopBackoff
-events for the client pods due to OOMKilled.
PS C:\azure-aks> kubectl get pods -n logging -w
NAME READY STATUS RESTARTS AGE
es-master-log-es-cluster-default-0 0/1 Pending 0 1s
es-client-log-es-cluster-664b8d9f76-lvk6t 0/1 Pending 0 1s
es-client-log-es-cluster-664b8d9f76-rr7kh 0/1 Pending 0 1s
es-master-log-es-cluster-default-0 0/1 Pending 0 1s
es-client-log-es-cluster-664b8d9f76-lvk6t 0/1 Pending 0 1s
es-client-log-es-cluster-664b8d9f76-w5tnn 0/1 Pending 0 1s
es-client-log-es-cluster-664b8d9f76-rr7kh 0/1 Pending 0 1s
es-client-log-es-cluster-664b8d9f76-w5tnn 0/1 Pending 0 1s
es-master-log-es-cluster-default-0 0/1 ContainerCreating 0 1s
es-client-log-es-cluster-664b8d9f76-lvk6t 0/1 ContainerCreating 0 1s
es-client-log-es-cluster-664b8d9f76-rr7kh 0/1 ContainerCreating 0 1s
es-client-log-es-cluster-664b8d9f76-w5tnn 0/1 ContainerCreating 0 1s
es-data-log-es-cluster-default-0 0/1 Pending 0 1s
es-data-log-es-cluster-default-0 0/1 Pending 0 1s
es-data-log-es-cluster-default-0 0/1 ContainerCreating 0 1s
es-client-log-es-cluster-664b8d9f76-lvk6t 0/1 Running 0 5s
es-client-log-es-cluster-664b8d9f76-w5tnn 0/1 Running 0 5s
es-client-log-es-cluster-664b8d9f76-lvk6t 0/1 OOMKilled 0 9s
es-client-log-es-cluster-664b8d9f76-w5tnn 0/1 OOMKilled 0 9s
es-client-log-es-cluster-664b8d9f76-lvk6t 0/1 Running 1 11s
es-client-log-es-cluster-664b8d9f76-w5tnn 0/1 Running 1 11s
es-client-log-es-cluster-664b8d9f76-lvk6t 0/1 OOMKilled 1 16s
es-client-log-es-cluster-664b8d9f76-w5tnn 0/1 OOMKilled 1 16s
es-client-log-es-cluster-664b8d9f76-rr7kh 0/1 Running 0 16s
es-client-log-es-cluster-664b8d9f76-w5tnn 0/1 CrashLoopBackOff 1 17s
es-client-log-es-cluster-664b8d9f76-lvk6t 0/1 CrashLoopBackOff 1 18s
es-client-log-es-cluster-664b8d9f76-rr7kh 0/1 OOMKilled 0 20s
es-client-log-es-cluster-664b8d9f76-rr7kh 0/1 Running 1 22s
[...]
Increasing memory limits by 512Mi to 1536Mi allows the cluster to start.
Steve Sloka commented
Many of the examples were driven off of demoing locally in minikube or something but would need to be tweaked for a real-world use case. We should probably have some docs on that.