Hi all, is it possible to submit a job change / upgrade a Kafka job (when submitting to flink) have zero downtime in some form? Any thoughts how people handle this today?

No, there is always some downtime involved

What are the alternatives? To avoid downtime

You can only work on minimizing downtime, by creating something like a secondary cluster on standby with your upgraded stuff and then shutting down your active cluster and starting your new cluster. Or work on a blue/green deployment model

Thanks Martijin. So natively there’s no job resilience unless we do bring up a standby

There is job resilience, but not without downtime

I see! So most of the flink users plan for downtime? Or blue/green deployments?

Blue/green deployments or hot/standby clusters are common yes

Pretty good resource on this:

Thank you this is useful