Uh oh!
There was an error while loading. Please reload this page.
Reduce canary schedule frequency (AMD 4→2/day, ARM 8→2/day) - #68821
Merged
Conversation
The scheduled canary runs the full matrix on all versions and is by far the most expensive run type. AMD ran 4x/day and ARM 8x/day; halving AMD alone saves ~1,900 compute-h / 2 weeks (~6% of AMD CI), and cutting ARM 8->2/day removes 75% of the ARM canary schedule. The two crons are interleaved (AMD 01:58 & 13:58, ARM 07:28 & 19:28) so a full-matrix canary still runs roughly every ~6h, alternating architecture - so the 'trust earned on main' backstop keeps a ~6h effective cadence for arch-agnostic regressions (12h for arch-specific ones) while the scheduled compute drops substantially. It also relieves runner-pool queuing, since each canary is the heaviest job-flood. check_ci_workflows_in_sync.py expected blocks updated to match.
potiuk
requested review from
amoghrajesh, ashb, bugraoz93, gopidesupavan, jason810496 and jscheffl
as code ownersJune 22, 2026 04:31
bugraoz93
approved these changes
Jun 25, 2026
bugraoz93
left a comment
Contributor
There was a problem hiding this comment.
New hours :) Sounds good, most issues already surface in new PRs as well
Uh oh!
There was an error while loading. Please reload this page.
Contributor
Backport successfully created: v3-3-testNote: As of Merging PRs targeted for Airflow 3.X In matter of doubt please ask in #release-management Slack channel.
|
potiuk added a commit
that referenced
this pull request
Jun 26, 2026
…day) (#68821) (#69039) The scheduled canary runs the full matrix on all versions and is by far the most expensive run type. AMD ran 4x/day and ARM 8x/day; halving AMD alone saves ~1,900 compute-h / 2 weeks (~6% of AMD CI), and cutting ARM 8->2/day removes 75% of the ARM canary schedule. The two crons are interleaved (AMD 01:58 & 13:58, ARM 07:28 & 19:28) so a full-matrix canary still runs roughly every ~6h, alternating architecture - so the 'trust earned on main' backstop keeps a ~6h effective cadence for arch-agnostic regressions (12h for arch-specific ones) while the scheduled compute drops substantially. It also relieves runner-pool queuing, since each canary is the heaviest job-flood. check_ci_workflows_in_sync.py expected blocks updated to match. (cherry picked from commit 5226a1b) Co-authored-by: Jarek Potiuk <jarek@potiuk.com>
karenbraganz pushed a commit
to karenbraganz/airflow
that referenced
this pull request
Jun 30, 2026
…#68821) The scheduled canary runs the full matrix on all versions and is by far the most expensive run type. AMD ran 4x/day and ARM 8x/day; halving AMD alone saves ~1,900 compute-h / 2 weeks (~6% of AMD CI), and cutting ARM 8->2/day removes 75% of the ARM canary schedule. The two crons are interleaved (AMD 01:58 & 13:58, ARM 07:28 & 19:28) so a full-matrix canary still runs roughly every ~6h, alternating architecture - so the 'trust earned on main' backstop keeps a ~6h effective cadence for arch-agnostic regressions (12h for arch-specific ones) while the scheduled compute drops substantially. It also relieves runner-pool queuing, since each canary is the heaviest job-flood. check_ci_workflows_in_sync.py expected blocks updated to match.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The scheduled canary runs the full matrix on all versions and is by far the
most expensive run type. AMD ran 4×/day and ARM 8×/day.
This takes both to 2×/day, interleaved so a full-matrix canary still
runs roughly every ~6 h, alternating architecture:
58 1,13 * * *→ 01:58, 13:5828 7,19 * * *→ 07:28, 19:28Savings: halving AMD alone is ~1,900 compute-h / 2 weeks (~6 % of AMD CI);
cutting ARM 8→2/day removes 75 % of the ARM canary schedule. It also relieves
runner-pool queuing — each canary is the heaviest job-flood (~269 jobs).
Trade-off (latency, not correctness): the canary is the "trust earned on
main" backstop. Interleaving keeps a ~6 h effective cadence for arch-agnosticregressions; an architecture-specific regression now has up to a ~12 h
detection gap (each arch runs 2×/day) instead of ~6 h / ~3 h. Cron-only change,
no test logic touched.
Was generative AI tooling used to co-author this PR?
Generated-by: Claude Code (Opus 4.8) following the guidelines