Capture and properly handle errors.

Handled in several places:

 - In the run() of the ProcessMedia itself for
   handled (BaseProcessingFail derived) errors (best to do these not
   in on_failure because the errors are highlighted in celeryd in a
   way that looks inappropriate for when the errors are well handled)
 - In ProcessMedia.on_failure() for all other errors
 - In the submit view where all exceptions are caught, media is marked
   at having failed, then the error is re-raised.  (The reason for
   this is that users running in "lazy" mode will get errors
   propagated by celery and so on_failure won't run for them.)
This commit is contained in:
Christopher Allan Webber
2011-08-13 12:21:06 -05:00
parent 4a477e246d
commit 6788b4123e
2 changed files with 50 additions and 27 deletions

View File

@@ -28,7 +28,7 @@ from mediagoblin.util import (
from mediagoblin.util import pass_to_ugettext as _
from mediagoblin.decorators import require_active_login
from mediagoblin.submit import forms as submit_forms, security
from mediagoblin.process_media import process_media
from mediagoblin.process_media import process_media, mark_entry_failed
from mediagoblin.messages import add_message, SUCCESS
@@ -102,9 +102,22 @@ def submit_start(request):
#
# (... don't change entry after this point to avoid race
# conditions with changes to the document via processing code)
process_media.apply_async(
[unicode(entry['_id'])], {},
task_id=task_id)
try:
process_media.apply_async(
[unicode(entry['_id'])], {},
task_id=task_id)
except BaseException as exc:
# The purpose of this section is because when running in "lazy"
# or always-eager-with-exceptions-propagated celery mode that
# the failure handling won't happen on Celery end. Since we
# expect a lot of users to run things in this way we have to
# capture stuff here.
#
# ... not completely the diaper pattern because the exception is
# re-raised :)
mark_entry_failed(entry[u'_id'], exc)
# re-raise the exception
raise
add_message(request, SUCCESS, _('Woohoo! Submitted!'))