Run integration tests after prod deployment#487
Merged
Conversation
The compactor e2e suite is too invasive to run against arc-cbr-production on every manual deploy: - Patches the cluster-wide `node-compactor` Deployment env vars (e.g. `min_node_age=3600`, `taint_rate=0.25`, `fleet_cooldown=90`, and briefly scales replicas to 0 in A6), altering compaction behavior for every managed NodePool — not just the test target — for ~15 minutes. - Drains the target NodePool and provisions up to ~5 r5.24xlarge nodes via Karpenter, competing with live CI capacity and adding EC2 cost. - A runner crash before teardown can leave the compactor stuck on test config or at replicas=0 until the next deploy. Recent runs (24752486783, 24761648999) also show `test_karpenter_deletes_empty_ tainted_nodes` reproducibly timing out, so the suite currently blocks prod rollouts on what is at least partly a test-side flake. arc-staging already exercises the same suite via the pre-merge workflow, so staging coverage is preserved. Mirror the existing `run_integration` gating pattern: add a `run_compactor` input to `_osdc-deploy.yml` (default `true`) and opt out explicitly from `osdc-deploy-prod.yml`.
tofu plan — arc-cbr-production✅ Plan succeeded · commit Plan output |
Contributor
|
In prod we should not run any of the tests. Compactor can put the cluster in a unhealthy state (potentially), janitor as well. Only lint, integration and potentially workload tests should run IMO. |
jeanschmidt
reviewed
Apr 22, 2026
jeanschmidt
reviewed
Apr 22, 2026
|
|
||
| - name: Run smoke tests | ||
| run: just smoke "${{ inputs.cluster }}" | ||
|
|
Contributor
There was a problem hiding this comment.
add integration tests here we had this, this is important, we should run on staging and on production
Smoke tests are noisy and block prod on unrelated failures; compactor e2e tests are too disruptive for prod. Both are validated against arc-staging pre-merge. Integration tests now run on prod against pytorch-canary via CANARY_GITHUB_TOKEN.
huydhn
changed the title
huydhn
had a problem deploying
to
osdc-production
April 22, 2026 20:59 — with 
GitHub Actions
Failure
GitHub Actions
Failure
jeanschmidt
approved these changes
Apr 22, 2026
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
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.
Run integration tests after prod deployment and skip other post-deployment tests that manipulating the clusters
Testing
https://github.com/pytorch/ci-infra/actions/runs/24802393183