search

GoogleCloudPlatform / flink-on-k8s-operator

[DEPRECATED] Kubernetes operator for managing the lifecycle of Apache Flink and Beam applications.
Apache License 2.0
659 stars 265 forks source link

Flink-operator cluster status keeps 'Creating' for hours #424

Open scape7yu opened 3 years ago

scape7yu commented 3 years ago

I started a flinkcluster that named 'aaa', for example. First, I executed 'kubectl get flinkcluster -n namespace'. The flinkcluster 'aaa' appeared, but the pods did not appear. Then I executed 'kubectl describe flinkcluster aaa -n namespace', the events are: ############################################################# Events: Type Reason Age From Message

Normal StatusUpdate 16m FlinkOperator Job status: Succeeded Normal StatusUpdate 16m FlinkOperator Cluster status: Creating #############################################################

The cluster status were keeping 'Creating' for hours, the pods did not appear.

Then I executed 'kubectl delete flinkcluster aaa -n namespace', and restarted flinkcluster 'aaa', it worked well.

Did anyone have meet this issue? Start, no response, restart, it's OK.

scape7yu commented 3 years ago

It seems because 'aaa' flinkcluster was submitted before, and 'aaa' was finnished and it was existing, if I submitted 'aaa' again, this issue is likely to appear, but not everytime, has probability.

Is it a bug? We'd better use the different name from the flinkclusters that we had submitted.

elanv commented 3 years ago

Flink cluster was deployed via "deployment" resource previously, but now it is deployed via "statefulset" with fixed pod names such as "aaa-0", "aaa-1", etc. This issue might be related to that change.