Hi, I’m going through the release notes and upgrade notes now. But just as a general question, when upgrading from 0.9.1.1 to latest, is it best to upgrade to each minor version along the way? Or can you skip over minor versions, like going from 0.9.1.1 -> 0.10.0, then 0.10.0 -> 0.11.0?
Would it be crazy to try to go straight to 0.11.0?
It sounds like you are already doing the right thing, which is reading the release notes for all intervening releases, especially the “updating from previous release” sections. You can find them all here: https://github.com/druid-io/druid/releases
I’m not sure off the top of my head if going directly from 0.9.1.1 -> 0.11.0 would work, but it might, especially if you aren’t using one of the features that has had its API evolving during that time (like query time lookups and the kafka indexing service).
And I see that 0.12.0 came out while I was reading through the release notes! Congratulations! So we’ll probably try to take it to that version unless there is some big stumbling block going from 0.11 -> 0.12.
We are in fact using query time lookups, and really hoping that some of the changes there will end the behavior in 0.9.1.1 where when you create or update a single lookup, all nodes dump and reload all their lookups, taking the lookups offline for 10-15 minutes. I saw the addition of versions to lookup configs in 0.10.1 giving me some hope, but it wasn’t clear if it would affect that specific behavior.
Anyway, I’m thinking now to try going to 0.10.1 where there were a lot of lookup changes, then from there to 0.12.0. Is there anything else to look out for regarding lookups? Hopefully I’ll answer my own question as I continue working through the release notes, but if there’s anything you want to call out that’d be helpful.