Steps to reproduce

1. Create a single-attach volume (e.g. AWS) named test.
2. Apply the configuration and wait until the run starts.
3. Apply the configuration for the second time.

type: dev-environment
ide: vscode
retry: true
volumes:
- test:/test

Actual behaviour

The second run fails with VOLUME_ERROR.

Expected behaviour

The second run is retried until the volume is released by the first run.

dstack version

0.19.13

Server logs

Additional information

Some use cases for retrying volume errors:

• Handling temporary backend errors without stopping the run.
• Enqueuing a run until a volume is released by another run.