Closed Bug 587696 Opened 15 years ago Closed 14 years ago

socket timeouts for redis and celery

Categories

(addons.mozilla.org Graveyard :: Public Pages, defect, P1)

defect

Tracking

(Not tracked)

RESOLVED FIXED
4.x (triaged)

People

(Reporter: jbalogh, Assigned: jbalogh)

Details

(Whiteboard: [z])

Otherwise they hang our websites.
Using the devious server from http://gist.github.com/527442. Here's a branch of redis-py that will do the right thing with socket errors: http://github.com/jbalogh/redis-py/commits/connection-errors. Currently it reconnects until it gets a recursion error: http://github.com/andymccurdy/redis-py/issues/issue/62. Then we'll have to set socket_timeout in the redis.Redis connection.
Is this something we need to patch in production?
(In reply to comment #2) > Is this something we need to patch in production? Redis isn't critical since we have it turned off, but we should fix celery since we're using it in prod.
This sets the rabbitmq timeout to 0.1 seconds: http://github.com/jbalogh/zamboni/commit/55765f4a7f And this hijacks amqplib to not clear timeouts (wtf?): http://github.com/jbalogh/zamboni-lib/commit/a140f53d => 5.11.9 since redis can wait.
Target Milestone: 5.11.8 → 5.11.9
The rabbit fix only works for the web side. We don't want celeryd connections to die.
Target Milestone: 5.11.9 → 5.11.8
Switched to my redis branch in http://github.com/jbalogh/zamboni-lib/commit/a97a976 and updated cache-machine to catch socket errors in http://github.com/jbalogh/django-cache-machine/commits/1eb9f2f7383bb. I don't know what to do about celery yet. We want persistent connections in celeryd but fail-fast sockets from the web app.
Target Milestone: 5.11.8 → 5.11.9
Target Milestone: 5.11.9 → 4.x (triaged)
Status: NEW → RESOLVED
Closed: 14 years ago
Resolution: --- → FIXED
Product: addons.mozilla.org → addons.mozilla.org Graveyard
You need to log in before you can comment on or make changes to this bug.