[System] Adding condition to 0.9.2 package => 0.9.3#674
[System] Adding condition to 0.9.2 package => 0.9.3#674ycombinator merged 2 commits intoelastic:masterfrom
Conversation
|
@ycombinator so is the plan to let the automation put in the |
@fearful-symmetry yep, exactly! That way we end up following all the usual processes and have a record of all changes in the |
💚 Build Succeeded
Expand to view the summary
Build stats
Test stats 🧪
Trends 🧪 |
fearful-symmetry
left a comment
There was a problem hiding this comment.
Assuming this is good. In the future we might want to find a less multi-step way to do this, since we're definitely going to see more cases where a stack release is constrained to certain integration releases.
|
Welp, we have a problem! CI is failing with this error: This is because the package registry distribution used for tests by CI contains a newer version of the Even if we merged this PR with the failing CI 😬, I don't think that's going to help us because I suspect we'll see the same error on the I'm wondering now if we should instead make a PR directly to https://github.com/elastic/package-storage/tree/snapshot with a new folder for @mtojek do you have any opinions or other ideas? Please see the PR description to understand what we are trying to accomplish with this PR. |
|
I chatted with @jen-huang off-PR about the error we're seeing in CI: Since we are trying to install the @jen-huang informed me that this is, in fact, possible with the Fleet API by passing |
|
Pinging @elastic/integrations (Team:Integrations) |
mtojek
left a comment
There was a problem hiding this comment.
I'm wondering now if we should instead make a PR directly to https://github.com/elastic/package-storage/tree/snapshot with a new folder for 0.9.3 (copied from 0.9.2) with the necessary condition change. The down side of this approach is that we won't have a copy of the 0.9.3 system package in the integrations repo and we'd bypass any checks that normally happen on integrations repo PRs. 🤔
I met a case in the past where there was condition on Kibana version in the package. As no user could install that package I simply modified the broken source in the package-storage. I understand that in this case it's possible to install the package, so your idea is correct, hence LGTM for this PR and the fix in the elastic-package.
I would also re-consider fixing broken packages directly in the package-storage in such emergency cases. It's much easier to handle it operationally.
|
Thanks for weighing in, @mtojek . I understand your point about broken package versions: because such versions cannot be installed anyway by the user, we can assume no one has installed them and, therefore, a fix to them will not require a re-install. So we can safely fix them directly in When you did this in the past, did you also fix the source package version in the |
In my case it was an issue with incompatible import of Kibana dashboards: elastic/package-storage@0767f7a . I fixed it only in the package-storage (must be compatible with Kibana 7.10), because the Apache package's version has been bumped up in Integrations together with bumped up Kibana's constraint (7.11.0 - https://github.com/elastic/integrations/blob/master/packages/apache/manifest.yml#L12). As we're working with latest master in Integrations, I couldn't alter the Git history there. |
d4234d7 to
300e96c
Compare
Right, this is what I'm facing with this PR here as well. It's why this PR is so messy (so many changes!) and why we will need a follow up PR to restore the I think, in general, we need to have a policy that covers how to fix older versions of packages so we are fixing them consistently and all package authors are aware of it. I'll file a separate issue to take this discussion off this PR. |
|
So just to recap, the next steps with this PR are as follows:
@fearful-symmetry @ChrsMark Please take note of the above, as you two come up as recent editors of the |
This reverts commit b1b35b1.
This reverts commit 6255798.
Created issue for discussion: #682. |
This is done now. The PR to
This is also done now. The PR to revert this PR here is up: #681. @mtojek @fearful-symmetry I've requested your reviews on both PRs above. |
What does this PR do?
This PR effectively performs two steps:
systempackage back to the0.9.2version of the package (commit SHA: 1f5013cconditionto theloaddata stream and bumps up the package version to0.9.3However, due to step 1, the diff in this PR is very large. To get a true-r sense of the changes (i.e. just step 2), check out this PR and run
git diff 1f5013cef1a441e3d76adcfe687b035c1390617c -- packages/system/.Immediately following up this PR will be a PR that undoes step 1, therefore bringing the
systempackage back to it's current state in this repo, i.e. as of commit SHA 293d243.Why is this change important?
The latest released
0.9.xversion of thesystempackage is0.9.2. We need to fix an issue in this series (by adding theconditionfield). For this, we need to release0.9.3of the package with this change.Checklist