- Newest
- Most votes
- Most comments
Hi!
- While no dedicated restart option exists, adding/removing/altering an arbitrary configuration override or log level option will force a restart.
Stopping an environment is not an option as there are numerous dependencies in Airflow such as the meta database that cannot be easily stopped. As an alternative, environments can be created and destroyed via API or CloudFormation.
- MWAA is running open source Airflow, either 1.10.12 or 2.0.2, and does not alter it's behavior. If the web server is not being updated to the lack of errors, it is likely that the scheduler responsible for serializing the DAGs has not completed or still sees errors. See your DAG processing logs in CloudWatch (may require INFO level logging). You can also debug locally using https://github.com/aws/aws-mwaa-local-runner
Thanks!
Came across this thread as I'm having the same issues but with Airflow v3.0.6. It seems to be a bug. If initially there were a few stacks of import errors the UI would display them. After fixing them, at least one stack would remain as a false positive and would not go away even after having to trigger a mwaa update (which annoyingly takes over 30 min). So we just got used to having it there. I tried to reduce the error stack by just forcing some import error at the highest level (ex. if it's some error that happened to be in file 3 that was imported in file 2 and file 2 was imported in file 1 then the UI would show 3 errors for the entire stack so I'd misspell an import in file 1, push to s3, wait for airflow ui to show 1 import error only, then fix the misspelled import and push again...this way it's just 1 error that persists as a false positive)
Relevant content
- asked 2 years ago
