Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.
Sign upcancel_consumer might lead into 'Unrecoverable error': Message for queue 'xxxx' without consumers. #2953
Comments
ask
closed this
in
celery/kombu@c53bed2
Dec 8, 2015
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
|
also depends on celery/py-amqp@b6d2567 |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
matkeng
Dec 8, 2015
unsubscribe
Sent from my iPhone
On 9 Dec 2015, at 5:03 am, Ask Solem Hoel notifications@github.com wrote:
also depends on b6d2567d9b9e9449fec9d5a3950b1d4ed55b214c
—
Reply to this email directly or view it on GitHub.
matkeng
commented
Dec 8, 2015
|
unsubscribe Sent from my iPhone
|
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
kheraankit
Jun 23, 2016
@ask is it possible to backport these changes or publish a new package by bumping incremental version?
Context:
The fix doesn't seem to be available with kombu package 3.0.35.
I was able to get it though by installing the latest snapshot
pip install --upgrade https://github.com/celery/kombu/zipball/master#egg=kombu
[amqp-2.0.2 kombu-4.0.0a1 vine-1.1.0]
our current celery version is 3.1.17 and it has dependency on kombu
Requires-Dist: kombu (>=3.0.24,<3.1) so installing a major version release --- kombu-4.0.0a1 might not be a good idea at this point, thoughts?
kheraankit
commented
Jun 23, 2016
•
|
@ask is it possible to backport these changes or publish a new package by bumping incremental version? Context: pip install --upgrade https://github.com/celery/kombu/zipball/master#egg=kombu our current celery version is 3.1.17 and it has dependency on kombu Requires-Dist: kombu (>=3.0.24,<3.1) so installing a major version release --- kombu-4.0.0a1 might not be a good idea at this point, thoughts? |
alekibango commentedDec 7, 2015
i really need some workers to be able to get out of most queues in some cases.
When i use some prefetch and concurrency,
it is really possible to crash worker by using cancel_consumer to stop listening to a queue.
When such task from cancelled queue comes in,
this problem should be recoverable, not unrecoverable one.
Just warning in log should be ok, task should be returned into its queue (with late_ack).
Possibly related (affected) issues: #2564, #1452