In addition to the credentials error exposed in
https://dagster.phacility.com/D8977, I noticed that we don't actually
need to query self._task_metadata to get the cluster name when
terminating. This is because we tag ECS launched pipeline runs with both
the ARN and the cluster. If it's lacking either, we can't terminate it
regardless of what self._task_metadata says.
Details
Details
- Reviewers
• dgibson • johann • max - Commits
- R1:00d73bb34601: Infer cluster from tags
unit
Diff Detail
Diff Detail
- Repository
- R1 dagster
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
python_modules/libraries/dagster-aws/dagster_aws/ecs/launcher.py | ||
---|---|---|
140 | should we check that ["tasks"] is non-empty too to avoid an index error |
python_modules/libraries/dagster-aws/dagster_aws/ecs/launcher.py | ||
---|---|---|
140 | I haven't been able to reproduce a case where the arn and cluster tags both exist, the request succeeds, but the tasks key is empty. I'm not entirely sure what we'd do in that case - we could retry (but for how long?) or we could return False (but that might not actually be accurate). I think for now, I'd prefer to keep it as is and if we start to see this bubble up, we can reconsider. |
python_modules/libraries/dagster-aws/dagster_aws/ecs/launcher.py | ||
---|---|---|
140 | Actually, maybe I'll just return False. |