Skip to content

Treat incomplete Kata autoscaling metrics as unavailable - #464

Draft
scotwells wants to merge 2 commits into
mainfrom
codex/guard-incomplete-kata-autoscaling-metrics
Draft

scotwells wants to merge 2 commits into
mainfrom
codex/guard-incomplete-kata-autoscaling-metrics

Conversation

@scotwells

@scotwells scotwells commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

Missing or stale Kata readings must not become zero usage in autoscaling decisions. Return a Pod's metrics only when every declared container has complete, valid, recent measurements for its current identity. Reject duplicates and retired containers. Query updates automatically restart the adapter.

Validation: 33 promtool scenarios cover missing containers, stale collection, replacement, invalid or duplicate readings, and genuine zero usage. The adapter overlay renders; CI runs these regressions.

Rollout order: infra #6854 → Kata provider #39 → this change. Kata metrics are unavailable until the prerequisites arrive. The API timestamp remains query time; freshness is checked separately.

Related to datum-labs/kata-provider#32 and datum-cloud/enhancements#919.

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant