Skip to content

[9.1] [Build] Split tasks by arch for increased parallelism (#223471)#226889

Merged
kibanamachine merged 1 commit intoelastic:9.1from
kibanamachine:backport/9.1/pr-223471
Jul 7, 2025
Merged

[9.1] [Build] Split tasks by arch for increased parallelism (#223471)#226889
kibanamachine merged 1 commit intoelastic:9.1from
kibanamachine:backport/9.1/pr-223471

Conversation

@kibanamachine
Copy link
Copy Markdown
Contributor

Backport

This will backport the following commits from main to 9.1:

Questions ?

Please refer to the Backport tool documentation

## Summary
Closes elastic/kibana-operations#110

This is a continuation of elastic#217929. Previously a `Task` created two
artifacts of a specific flavor, one for `x64` and one for `ARM64`. These
were still running serially, even though the `Task` itself was
parallelized. This PR splits each architecture into its own `Task` so we
can increase parallelism. The result is a 14% further decrease in time
taken for the build step; 53 min down to 46 min. Between the two PRs,
the artifact build time is reduced 57%!

### Testing
https://buildkite.com/elastic/kibana-artifacts-snapshot/builds/6413

### Next steps
- Moving the ARM artifact builds to an ARM native machine would likely
see the next biggest reduction because they take 2-3 times longer than
their x86 counterparts. Though this would likely be a decent amount of
work.
- Some of the `global` tasks could be grouped into parallel tasks, but
there is some dependency between those tasks and the gains would be
minimal.

(cherry picked from commit b51fab2)
@kibanamachine kibanamachine added the backport This PR is a backport of another PR label Jul 7, 2025
@kibanamachine kibanamachine enabled auto-merge (squash) July 7, 2025 19:33
@kibanamachine kibanamachine merged commit e4432f0 into elastic:9.1 Jul 7, 2025
16 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

backport This PR is a backport of another PR

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants