This is a followup on the incident T314835.
To mitigate the issue the flink job was started from yarn, we should move it back to k8s.
We should coordinate with the cleanup done in T316003:
Option1:
- keep using the T314835 subfolder (even for the flink_ha storage folder)
- once the cleanup is done, fully redeploy the service under the root folder
Option2:
- wait for the cleanup to be done and deploy the service using the root folder
Option1 should be preferred if the cleanup is going to take several weeks.
AC:
- WDQS & WCQS flink jobs are running from k8s@codfw