Skip to content

Allow autoscaling to work better with latest kubernetes - #134

Open
tintou wants to merge 4 commits into
collabora:mainfrom
tintou:tintou/autoscale
Open

tintou wants to merge 4 commits into
collabora:mainfrom
tintou:tintou/autoscale

Conversation

@tintou

@tintou tintou commented Nov 6, 2026

Copy link
Copy Markdown

Here is a series of 4 patches:

  • chart: support current Kubernetes versions: autoscaling/v2beta1 is removed in 1.25
  • Drain running jobs on SIGTERM: Make the runner finish the ongoing jobs before stopping
  • chart: scale down HPA slowly
  • Make the maximum number of concurrent jobs configurable: Mostly for completeness, don't hardcode the value of 64

tintou and others added 4 commits November 6, 2026 13:51
Move the HPA to autoscaling/v2 (v2beta1 was removed in 1.25), set a
RuntimeDefault seccomp profile and require Kubernetes >= 1.23.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Stop requesting jobs on SIGTERM/Ctrl-C and wait for running jobs to
finish, so pod rollouts and autoscaler scale-downs don't kill them.
Add a terminationGracePeriodSeconds chart value (default 1h).

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add --max-jobs / RUNNER_MAX_JOBS (default 64) and a max_jobs chart value.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@tintou

tintou commented Nov 6, 2026

Copy link
Copy Markdown
Author

This might also wait for collabora/gitlab-runner-rs#144

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

1 participant