Question or comments on the 2022 Release Schedule blog post?
I’m a little concerned about 2 major releases coming out so close together, especially since we’re just now preparing to move to OpenSearch 1.x (from ODFE 1.13.3). What upgrade paths/scenarios do you envision supporting? For example, will OpenSearch 1.x => OpenSearch 3.x be an option or will we need to move to OpenSearch 2.x?
Also, while the big themes of these releases are focused on OpenSearch, do you see these releases including significant changes in OpenSearch Dashboards as well?
I’m just trying to get a feel for how much this schedule might impact us. It seems like we’re always playing catch-up. Perhaps we’ve been too conservative but letting things “mature” a bit seemed wise given the magnitude of the effort needed to create the fork.
1 Like
As far as upgrading between OpenSearch 1.x directly to OpenSearch 3.x, I’m not sure that’s entirely planned out and I’ll let others chime in. However, do keep in mind both OpenSearch 2.x and OpenSearch 3.x will be based on Lucene 9, so the differences are not super-duper deep.
re: Dashboards: the post was referring to the project as a whole - so OpenSearch Dashboards would be in the same boat. There has been some talk about decoupling major versions but AFAIK, that hasn’t yet happened.
FWIW, I fully expect folks will be on 1.x for some time. It’s going to take a while for managed services especially do move to the new major versions. So, latest version currency isn’t a full expectation (as it was with ES). This 2.x to 3.x is a little bit of a fast-forward in the project to get everything in a reasonable state. I would expect 3.x onward to be more oriented towards more minor releases (less breakage).
I think the upgrade from 1.x to 3.x is good feedback (rolling and snapshot/restore). We should take that into consideration. Since both 2.x and 3.x are developed on Lucene 9.0, it should be doable. I am concerned about the 2 majors so close together will mean that some will go straight from 1.x to 3.x and we should try to make that easy. I do agree with not waiting on Lucene 9 (2.x) and that we have a lot of critical work to do on the internals to setup our longer term re-architecture that will not be ready until later in the year so it seems somewhat unavoidable to do two majors fairly close together.
Also we need to be super clear about any breaking changes in the core REST APIs - which I assume will be very minimal. These changes are the ones that are the most difficult for folks to absorb as they need to change their application/client interfaces.
1 Like
I kind of have the same concern or question regarding the release plan. Major versions are really close together, although I’m not totally sure about the support policy for opensearch.
Are you planning to support the last minor version of the previous major like in case of ES or is that planned for n-2 as well now that skip upgrade came up?
1 Like