You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current CI pipeline takes at least 15 minutes to complete given all checks pass. This can be seen from the workflow runs on #136.
I believe that the CI is running slower than it needs to because there are unnecessary dependencies between jobs. This can be seen in the following excerpt from .github/workflows/build.yml, which I've abbreviated for clarity:
The integration-tests job is one of the slowest jobs in the pipeline, and it doesn't start until test_extensions passes, which in turn doesn't start until build_jupyter-chat passes.
I don't see an obvious benefit to keeping the inter-job dependencies that exist today. One may argue that this reduces runner usage, but this is only true when the CI fails (which is ideally <<50% of the time). One may also argue that this results in seeing fewer failed jobs when the CI fails, but again, this is optimizing for failure.
Proposed Solution
Remove the needs field from all jobs, and allow the CI to run fully in parallel.
The text was updated successfully, but these errors were encountered:
Problem
The current CI pipeline takes at least 15 minutes to complete given all checks pass. This can be seen from the workflow runs on #136.
I believe that the CI is running slower than it needs to because there are unnecessary dependencies between jobs. This can be seen in the following excerpt from
.github/workflows/build.yml
, which I've abbreviated for clarity:The
integration-tests
job is one of the slowest jobs in the pipeline, and it doesn't start untiltest_extensions
passes, which in turn doesn't start untilbuild_jupyter-chat
passes.I don't see an obvious benefit to keeping the inter-job dependencies that exist today. One may argue that this reduces runner usage, but this is only true when the CI fails (which is ideally <<50% of the time). One may also argue that this results in seeing fewer failed jobs when the CI fails, but again, this is optimizing for failure.
Proposed Solution
Remove the
needs
field from all jobs, and allow the CI to run fully in parallel.The text was updated successfully, but these errors were encountered: