Hello Everyone, I am using Ververica platform for flink job deployments. I want to do rolling upgrades with zero downtime. But as per the given upgrade strategies by ververica, none of it achieves zero downtime. Can anyone help me on this?

You should reach out to Ververica for support questions on their platform. But any upgrade will always involve some downtime.

@Sudhan Madhavan thanks for using Ververica platform! Would you like to elaborate what exactly your requirement is wrt the zero downtime upgrade?

@Jing Ge When we do a job rolling upgrade from version x to y, it use to suspend the current version x and when once it gets successfully terminated, then the version y will start from the latest savepoint. Our requirement is to avoid this downtime during upgrade from version x to y. Do we support any deployment strategy like blue green deployment or any other sufficient strategy?

@Martijn Visser does native flink has such capabilities

This really depends on the requirements of your Flink application too. For example, you can't start the same Flink job twice if you want to have exactly-once requirements. I think the only thing you can do is minimize downtime. With Ververica Platform, you could consider something like the Forking Deployments mode https://docs.ververica.com/user_guide/application_operations/lifecycle_management/forking.html

@Sudhan Madhavan Flink as a true distributed system won’t support the zero downtime job upgrade out of the box. Zero downtime upgrade, e.g. blue green deployment, etc., is used for online applications and mostly for the stateless service modules. Flink has a vision to cover the online computation scenario, but currently it is still focusing on offline/nearline data processing. A typical Flink job contains data ingestion, compute logic with state management, and sink result. It should be one piece of service within a whole service pipeline(including online services) to serve the business. I was wondering what is your business requirement that leads to the tech requirement to make Flink job be able to zero downtime upgrade instead of leveraging the zero downtime upgrade capability most online applications/services already have.