Re-start build pipeline by approved contributors
mvriel opened this issue · comments
Feature request
Feature category
- Website: improve the way to browse or discover packages.
- Pipelines: improve the way to publish or manage a package.
- Registry: improve the registry server.
- Unity Editor: improve how OpenUPM works with the Unity Editor.
- Documentation
- Others
What problem does this feature solve?
Sometimes the pipeline fails due to a timeout (see https://github.com/openupm/openupm/actions/runs/5783399794), and that means we, as contributors, need to wait until a new merge happens before our package becomes released.
How should this be implemented in your opinion?
When an approved contributor would be able to restart the pipeline, it could at least help by manually restarting the failed pipeline. Another solution could be to have a retry on the step for building the websites; as the NPM install is probably the biggest culprit when it comes to timeouts.
Are you a regular backer or sponsor at patreon.com/openupm?
No, not yet :)
Are you willing to work on this yourself?
Yes, though I do not know when I will have time for it :)
@mvriel, thank you for submitting the proposal.
I'm surprised I didn't receive notification emails for GitHub action failures on occasion. I will manually trigger a rebuild in those cases.
Please note that the example link failed during the build-website-us stage. This stage is a separate process from the package release pipelines. As a result, only the website update has been delayed until the next merge. This situation is not ideal.
When an approved contributor would be able to restart the pipeline, it could at least help by manually restarting the failed pipeline.
It appears that GitHub doesn't provide permissions at that level of granularity.
Another solution could be to have a retry on the step for building the websites; as the NPM install is probably the biggest culprit when it comes to timeouts.
GitHub doesn't natively support the "retry" feature. However, there are actions like https://github.com/nick-fields/retry that add this functionality, replacing the run command. This aligns with our needs, especially considering that failures often occur during the yarn install stage.