Repository navigation
Widen the Spot pool for RunsOn runners: more families, capacity-optimized, size ranges - #75822
Merged
Merged
Conversation
Generated-by: Claude Opus 5
boring-cyborg
Bot
added
area:dev-tools
backport-to-v3-4-test
potiuk
requested review from
amoghrajesh,
bugraoz93,
choo121600,
ephraimbuddy,
gopidesupavan,
jason810496,
jedcunningham,
jscheffl and
vatsrahul1001
as code owners
October 27, 2026 20:47
shahar1
approved these changes
Oct 27, 2026
shahar1
enabled auto-merge (squash)
October 27, 2026 20:47
Member
Author
|
(we should only merge after deploying terraform plan). |
Generated-by: Claude Opus 5
Generated-by: Claude Opus 5
potiuk
changed the title
potiuk
disabled auto-merge
October 27, 2026 21:43
Contributor
Backport successfully created: v3-4-testNote: As of Merging PRs targeted for Airflow 3.X In matter of doubt please ask in #release-management Slack channel.
|
github-actions Bot
pushed a commit
to aws-mwaa/upstream-to-airflow
that referenced
this pull request
Oct 27, 2026
…pacity-optimized, size ranges (apache#75822) * Let RunsOn x64 runners use more instance families for Spot Generated-by: Claude Opus 5 * Take RunsOn Spot runners from capacity-optimized pools Generated-by: Claude Opus 5 * Let RunsOn runners take Spot from up to twice the requested size (cherry picked from commit d17f121) Co-authored-by: Jarek Potiuk <jarek@potiuk.com> Generated-by: Claude Opus 5
aws-airflow-bot
pushed a commit
to aws-mwaa/upstream-to-airflow
that referenced
this pull request
Oct 27, 2026
…pacity-optimized, size ranges (apache#75822) * Let RunsOn x64 runners use more instance families for Spot Generated-by: Claude Opus 5 * Take RunsOn Spot runners from capacity-optimized pools Generated-by: Claude Opus 5 * Let RunsOn runners take Spot from up to twice the requested size (cherry picked from commit d17f121) Co-authored-by: Jarek Potiuk <jarek@potiuk.com> Generated-by: Claude Opus 5
vatsrahul1001
pushed a commit
that referenced
this pull request
Oct 28, 2026
…pacity-optimized, size ranges (#75822) * Let RunsOn x64 runners use more instance families for Spot Generated-by: Claude Opus 5 * Take RunsOn Spot runners from capacity-optimized pools Generated-by: Claude Opus 5 * Let RunsOn runners take Spot from up to twice the requested size (cherry picked from commit d17f121) Co-authored-by: Jarek Potiuk <jarek@potiuk.com> Generated-by: Claude Opus 5
vatsrahul1001
pushed a commit
that referenced
this pull request
Oct 28, 2026
…pacity-optimized, size ranges (#75822) * Let RunsOn x64 runners use more instance families for Spot Generated-by: Claude Opus 5 * Take RunsOn Spot runners from capacity-optimized pools Generated-by: Claude Opus 5 * Let RunsOn runners take Spot from up to twice the requested size (cherry picked from commit d17f121) Co-authored-by: Jarek Potiuk <jarek@potiuk.com> Generated-by: Claude Opus 5
vatsrahul1001
pushed a commit
that referenced
this pull request
Oct 28, 2026
…pacity-optimized, size ranges (#75822) (#75825) * Let RunsOn x64 runners use more instance families for Spot Generated-by: Claude Opus 5 * Take RunsOn Spot runners from capacity-optimized pools Generated-by: Claude Opus 5 * Let RunsOn runners take Spot from up to twice the requested size (cherry picked from commit d17f121) Generated-by: Claude Opus 5 Co-authored-by: Jarek Potiuk <jarek@potiuk.com>
1 task done
potiuk
added a commit
that referenced
this pull request
Nov 1, 2026
The static checks of every pull request based on current main fail in `ruff format`, which reformats these two constants, added in #75822. Generated-by: Claude Opus 5
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 free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
On RunsOn, every job asks for one exact size from four x64 families (
family=m7a+m7i+m6a+m6i,cpu=8/ram=32) with RunsOn's default price-capacity-optimized Spot strategy. That leaves the Spot allocator few pools, and in us-east-2 they run short: of 2156 Spot capacity errors in the RunsOn worker log on 2026-10-26/27, 1959 were x64 and 197 arm64. A miss sends the runner to on-demand, and an interruption reruns the whole job on on-demand (719 reruns in those two days). On-demand costs more and its 2000 vCPU quota was full on 2026-10-26 16:00 UTC.This PR widens the pool in three ways:
m8a,m8i,m5a,m5, and the x64 in-memory runners also acceptr8i. All of them have Spot capacity offered in all three us-east-2 zones, at prices comparable to the current families (2xlarge Spot on 2026-10-27: 0.08–0.16 USD/hour against 0.12–0.15). arm64 families are unchanged.spot=co(capacity-optimized) on all RunsOn templates. AWS takes Spot from the pools with the most spare capacity, which lowers both capacity misses and interruptions, at a somewhat higher Spot price than the default.4+8/16+32,8+16/32+64,32+64/128+256, and8+16/32+128for the in-memory jobs (101format()call sites in 18 workflows). Spot can then also come from the next size up.Trade-off: capacity-optimized ignores price, so with ranges a job can land on an instance up to twice the size it needs, at up to twice the hourly price. xdist follows the host's CPUs, so such a job runs twice the test workers and finishes sooner, but job durations become less comparable between runs. The weekly on-demand share report in apache/airflow-ci-infra and the RunsOn daily cost report will show whether the net cost goes down. Each part is its own commit and can be reverted on its own.
Related: apache/airflow-ci-infra#77 upgrades the RunsOn stack to v3.5.1, after which a Spot capacity miss sends only the affected runner to on-demand instead of every job for five minutes.
🤖 Generated with Claude Code