Is it possible to achieve zero downtime rollouts for new nifi versions? We have thus far been unsuccessful in convincing Nifi nodes to rejoin the cluster after a version update. The exceptions being thrown range from partial flow inheritance to missing nar bundles (we resolved this one).
Before we spend too much time on this, I’d love to know if we are simply barking up the wrong tree here. Upgrade documentation—last updated 2018—states, “You cannot have nodes with different NiFi versions in the same cluster.” https://cwiki.apache.org/confluence/plugins/servlet/mobile?contentId=80453196#content/view/80453196 Is this documentation up to date? What does this statement mean? Is there explicit code that will prevent this from happening, or is the behavior of a cluster with nodes running different versions simply undefined? Thanks in advance. If this discussion is better suited for the developer mailing list, please let me know and I will repost it there.
