-
Bug
-
Resolution: Done
-
Critical
-
None
https://bugzilla.redhat.com/show_bug.cgi?id=1731788
"""
Description of problem:
We are using "Celery" distributed task queue with RabbitMQ broker. The Celery uses qpid-proton to connect to RabbitMQ. After the upgrade to latest qpid-proton-0.28.0-1.el7 from EPEL7, our application started leaking lot of memory.
The core file dumped by the "gcore" utility showed lot of memory related to SSL certificates, so I presume the leak must appear somewhere in a code related to that.
We have been able to fix the memory leak in our application by downgrading to older version qpid-proton-0.26.0-2.el7 while keeping all the other packages in the same versions, so we are pretty sure the issue is in qpid-proton or in some incompatibility between Celery and qpid-proton in this version.
Version-Release number of selected component (if applicable):
Version with a leak: qpid-proton-0.28.0-1.el7
Version without leak: qpid-proton-0.26.0-2.el7
Steps to Reproduce:
1. Deploy RabbitMQ with SSL support.
2. Setup Celery worker to connect to rabbitmq using the "amqps://" protocol.
3. Send dummy tasks to the Celery worker.
4. The worker leaks memory slowly
I'm not sure if this is the easiest way to reproduce it. Maybe the Celery is not really important here, but that's where we see the leak.
"""
- is duplicated by
-
ENTMQCL-1715 [python] Client leak affecting Satellite in 0.28
- Closed
- is related to
-
ENTMQCL-1996 [py, win] memory leak from BlockingConnection reconnection
- New