Why do we need switchover from oldMaster to masterCandidateName in case of single node cluster?

[danholesinsky]

Hi team,

plz, I would like to ask - I use zalando postgres operator with one master (without replicas) and I want to move this master from one node to another during k8s api update (triggered by kubectl cordon node). My question is - why do you need switchover in this code https://github.com/zalando/postgres-operator/blob/master/pkg/cluster/pod.go#L283 in case of single node cluster? This will only lead to 5 retries, because there is no old master, see log:

msg="moving pods: node \"/ip-10-6-22-122.eu-west-1.compute.internal\" became unschedulable and does not have a ready label: map[\"rds\":\"postgres\"]" pkg=controller
msg="starting process to migrate master pod \"debug-pgop/pg-pa-debug-pgop-0\"" cluster-name=debug-pgop/pg-pa-debug-pgop pkg=cluster
msg="migrating single pod cluster \"debug-pgop/pg-pa-debug-pgop\", this will cause downtime of the Postgres cluster until pod is back" cluster-name=debug-pgop/pg-pa-debug-pgop p
msg="moving pod \"debug-pgop/pg-pa-debug-pgop-0\" out of the end-of-life node \"ip-10-6-22-122.eu-west-1.compute.internal\"" cluster-name=debug-pgop/pg-pa-debug-pgop pkg=cluster
msg="pod \"debug-pgop/pg-pa-debug-pgop-0\" has been recreated" cluster-name=debug-pgop/pg-pa-debug-pgop pkg=cluster
msg="unsubscribing from pod \"debug-pgop/pg-pa-debug-pgop-0\" events" cluster-name=debug-pgop/pg-pa-debug-pgop pkg=cluster
msg="pod \"debug-pgop/pg-pa-debug-pgop-0\" moved to end-of-life node \"ip-10-6-24-254.eu-west-1.compute.internal\"" cluster-name=debug-pgop/pg-pa-debug-pgop pkg=cluster
...
msg="switching over from \"pg-pa-debug-pgop-0\" to \"debug-pgop/pg-pa-debug-pgop-0\"" cluster-name=debug-pgop/pg-pa-debug-pgop pkg=cluster
msg="subscribing to pod \"debug-pgop/pg-pa-debug-pgop-0\"" cluster-name=debug-pgop/pg-pa-debug-pgop pkg=cluster
msg="making POST http request: http://10.6.4.52:8008/failover" cluster-name=debug-pgop/pg-pa-debug-pgop pkg=cluster
msg="unsubscribing from pod \"debug-pgop/pg-pa-debug-pgop-0\" events" cluster-name=debug-pgop/pg-pa-debug-pgop pkg=cluster
msg="could not failover to pod \"debug-pgop/pg-pa-debug-pgop-0\": could not switch over from \"pg-pa-debug-pgop-0\" to \"debug-pgop/pg-pa-debug-pgop-0\": could not make request: Post \"http://10.6.4.52:8008/failover\": dial tcp 10.6.4.52:8008: connect: no route to host" cluster-name=debug-pgop/pg-pa-debug-pgop pkg=cluster
...
msg="could not move master pod \"debug-pgop/pg-pa-debug-pgop-0\": could not migrate master pod: still failing after 5 retries" pkg=controller
msg="0/1 master pods have been moved out from the \"/ip-10-6-22-122.eu-west-1.compute.internal\" node" pkg=controller
msg="failed to move master pods from the node \"ip-10-6-22-122.eu-west-1.compute.internal\": could not move master 1/1 pods from the \"/ip-10-6-22-122.eu-west-1.compute.internal\" node" pkg=controller

Is there any other way how to trigger switchover for single master pod cluster without waiting for 5 minutes?

Thank you