[
https://issues.apache.org/jira/browse/AIRFLOW-4551?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16844643#comment-16844643
]
Emmanuel Brard commented on AIRFLOW-4551:
-----------------------------------------
[~jackjack10] Not really, what you mentioned is really interesting but the
scheduler logs are not necessarily available to end-users and also I don't see
in the code anything special about dealing with DAG parsing exception.
The strak trace I mentioned is already available from the scheduler log itself,
but unfortunately only the exception is made available to the end-user through
the web UI (the famous broken DAG message).
Since the `stractrace` field in the `import_error` table is of type text I
think it would be relevant to add the actual stacktrace and not only the
exception.
> Scheduler captures exception and stracktrace
> --------------------------------------------
>
> Key: AIRFLOW-4551
> URL: https://issues.apache.org/jira/browse/AIRFLOW-4551
> Project: Apache Airflow
> Issue Type: New Feature
> Components: scheduler
> Reporter: Emmanuel Brard
> Priority: Minor
>
> As of today, when the scheduler cannot import a DAG, an entry is
> [added|https://github.com/apache/airflow/blob/97ad71c99373759cc022733b2732c3bd43af983d/airflow/models/dagbag.py#L201]
> to `import_error` table with the exception message as `stacktrace` (later
> displayed on the web-ui).
> This is most of the time not enough to debug complex situation.
> I would suggest to add the "real" stack trace in addition to the exception.
> Using the traceback module should allow for it in a very similar way.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
