Use a bare raise so the original exception gets propagated.

[jbalogh]

With this change I get full stack traces in the logs. Without it the exception appears to originate from the line with raise e.

[mistercrunch]

I thought e would contain the original exception, this goes into unexplored territories for me. How is it different?

[jbalogh]

Sorry for the delay, our cluster was having a bad day.

Here's a traceback with raise e:

$ ./bin/airflow run dag task.Crash 2015-07-09 --local
Activating the virtualenv.
Logging into: ./logs/dag/task.Crash/2015-07-09T00:00:00
Traceback (most recent call last):
File "bin/airflow", line 156, in <module>
args.func(args)
File "vendor/src/airflow/airflow/bin/cli.py", line 126, in run
job_id=args.job_id,
File "vendor/src/airflow/airflow/models.py", line 755, in run
raise e
ZeroDivisionError: integer division or modulo by zero

And here's a traceback with raise:

$ ./bin/airflow run dag task.Crash 2015-07-09 --local
Activating the virtualenv.
Logging into: ./logs/dag/task.Crash/2015-07-09T00:00:00
Traceback (most recent call last):
File "bin/airflow", line 156, in <module>
args.func(args)
File "vendor/src/airflow/airflow/bin/cli.py", line 126, in run
job_id=args.job_id,
File "vendor/src/airflow/airflow/models.py", line 752, in run
task_copy.execute(context=self.get_template_context())
File "vendor/src/airflow/airflow/operators/python_operator.py", line 48, in execute
return_value = self.python_callable(*self.op_args, **self.op_kwargs)
File "./tasks/task.py", line 186, in execute
self.run(context)
File "./tasks/task.py", line 309, in run
1 / 0
ZeroDivisionError: integer division or modulo by zero

This only affects what's shown in the airflow run context. The log file shows the proper traceback both times.

[artwr]

For future ref for me :
http://stackoverflow.com/questions/4825234/exception-traceback-is-hidden-if-not-re-raised-immediately
http://nedbatchelder.com/blog/200711/rethrowing_exceptions_in_python.html

Seems reasonable.

[mistercrunch]

Ok, nice.