Before you begin
Follow the steps in this KB article for connecting to public or private GKE cluster.
Install kubectl and configure kubectl command line access using steps 1 and 2 in Getting thread dump section in the same article (alternatively, see go/cdf-internal-runbook for same instructions in more easily copy-paste-able format).
Updating memory of a pod
Get the name of the pod of interest: kubectl get pods
To check current configuration of pod: kubectl describe pod/<name of pod>
Modify and save the memory in the yaml: kubectl edit cdapmasters.cdap.cdap.io/<instance-name> -o yaml
Check if the pod restarted successfully: kubectl get pods
For nodes stuck in init:CrashLoopBackoff state
Delete the PV:
Find the name of the PV for the pod we want to restart: kubectl get pvc
Delete the PV: kubectl delete pvc/<pod-pvc-name>
Delete the statefulset:
Find the name of the statefulset for the pod we want to restart: kubectl get statefulset
Delete the statefulset: kubectl delete statefulset/<pod-statefulset-name>
Check if the pod restarted successfully:
kubectl get pods
kubectl describe pod/<updated-pod>