Hey Everyone, I am trying to figure out the upgrade process for Flink operator. Please note that our CI/CD does not allow use of helm so the operator upgrade process using helm in official
doc might not be relevant to us. The manual upgrade process described
here requires job downtime and we would like to avoid that if possible.
Since Flink operator documentation guarantees
backward compatibility, will below steps suffice to ensure a smooth upgrade -
• (kubectl) replace existing CRD with the desired version CRD
• remove existing operator deployment
• deploy the desired version flink operator.
• For each job in cluster
◦ Suspend the jobs with savepoint
◦ delete existing FlinkDeployment resource.
◦ create new FlinkDeployment resource and ensure to start the job by resuming from savepoint.