[BUG] Merge is broken?

[cmgrote]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

The merge into master seems to be broken, having failed multiple times now from previous PR merges. (Doesn't appear to have been successful after some point yesterday.)

Expected Behavior

The merge to master completes successfully and snapshots, container images, etc are all successfully published.

Steps To Reproduce

1. Create a PR.
2. See that it passes required checks (compilation, etc).
3. Merge it.
4. See that the Merge master) action fails.

Environment

N/A (GitHub Action)

Any Further Information?

https://github.com/odpi/egeria/actions/workflows/merge-maven.yml

[planetf1]

Yes am aware

[planetf1]

Build breakage from around Wed 18:30 was following a failed fix for :

#4046.

• dockerhub limits being reached on pulls
• intermittently impacting development, testing, release & some users of dojo labs
• fix also publishes docker images to quay.io
• fix needed prior to next release so that charts can be updated

Unfortunately some typos & errors have led to a series of corrections to the original intended fix.

• some build errors due to a bad fix
• some build errors due to network issues
• some delays getting the PR in due to other dev activity

The impact has been

• Snapshot maven artifacts not being published (some of the failed builds have gone past this stage, so this is partly addressed)
• docker images not being published (in the area of the fix)

The first error was syntax and I should have spotted before merge. Further on in the process gets harder since we only publish and hit the registries during the post-merge builds, and additionally have conditions based on branch, repo owner & event which makes testing more complex.

Possible fixes

• do yaml validation as part of PR build (would have only addressed first issue)
• do a PR docker build & perhaps push to an alternative location (this may help with other docker build errors, but wouldn't have picked up these errors)
• work on a separate repo/fork (again it's not certain this would have helped. It also takes time to setup, though I have tested some changes this way)

A more pragmatic approach may be to

• short term block of other merges (considered)
• optionally optimization of build/test where code isn't changed
• post status to slack (usually do this, didn't this time)

as this would allow a tighter iteration fix loop

However the impact of delayed snapshot/docker builds would effectively be the same, and could have a larger impact on development more broadly.

Hope that explains. Will close once the fixes are complete.

[planetf1]

Fixed