Skip to content

Conversation

@d4l3k
Copy link
Member

@d4l3k d4l3k commented Jul 19, 2024

This allows restarting a single role when the scheduler supports specific role restarts. Not all schedulers support this and this PR doesn't add support for it in any of the existing open source schedulers.

Volcano doesn't support multiple roles and thus we can't support role level restart with the very limited action policy https://pkg.go.dev/volcano.sh/apis@v1.9.0/pkg/apis/bus/v1alpha1#Action

This follows the pattern in #922. Role is sufficiently generic so is a valid fit for OSS specs.

Test plan:

pytest torchx/specs/test/api_test.py

@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Jul 19, 2024
@d4l3k d4l3k requested a review from kiukchung July 19, 2024 20:56
@d4l3k d4l3k merged commit 6112e03 into main Jul 19, 2024
@d4l3k d4l3k deleted the d4l3k/retry_role branch July 19, 2024 22:01
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants