Celery reconnecting

[imbolc]

I'm using redis as celerys broker and channels layers backend. After sending a channel message from celery task I see it reconnecting multiple times:

[2018-04-14 08:15:39,995: INFO/MainProcess] Received task: pong[923fcfde-b0f5-4d33-a3db-ed4f28aca71e]
[2018-04-14 08:15:40,004: DEBUG/MainProcess] Using selector: SelectSelector
[2018-04-14 08:15:40,006: DEBUG/MainProcess] Parsing Redis URI 'redis://localhost:6379/2'
[2018-04-14 08:15:40,006: DEBUG/MainProcess] Creating tcp connection to ('localhost', 6379)
[2018-04-14 08:15:40,024: DEBUG/MainProcess] Parsing Redis URI 'redis://localhost:6379/2'
[2018-04-14 08:15:40,024: DEBUG/MainProcess] Creating tcp connection to ('localhost', 6379)
[2018-04-14 08:15:40,042: DEBUG/MainProcess] Parsing Redis URI 'redis://localhost:6379/2'
[2018-04-14 08:15:40,043: DEBUG/MainProcess] Creating tcp connection to ('localhost', 6379)
[2018-04-14 08:15:40,062: DEBUG/MainProcess] Parsing Redis URI 'redis://localhost:6379/2'
[2018-04-14 08:15:40,063: DEBUG/MainProcess] Creating tcp connection to ('localhost', 6379)
[2018-04-14 08:15:40,083: DEBUG/MainProcess] Parsing Redis URI 'redis://localhost:6379/2'
[2018-04-14 08:15:40,084: DEBUG/MainProcess] Creating tcp connection to ('localhost', 6379)
[2018-04-14 08:15:40,104: DEBUG/MainProcess] Parsing Redis URI 'redis://localhost:6379/2'
[2018-04-14 08:15:40,105: DEBUG/MainProcess] Creating tcp connection to ('localhost', 6379)
[2018-04-14 08:15:40,132: INFO/MainProcess] Task pong[923fcfde-b0f5-4d33-a3db-ed4f28aca71e] succeeded in 0.13646206096746027s: None

If I set the same redis database for celery and channels I see much more such reconnections.

celery==4.1.0
channels==2.1.0
channels-redis==2.1.1

[andrewgodwin]

I'm not sure there's anything we can do about this in Channels - I'm not familiar with how Celery uses Redis internally, but I know they use a different library to us to talk to it so I don't know why it would interfere like this.

[imbolc]

Sure, but I was using channels from first version and there wasn't such issue. And after switching to channels-2 I found this. So I thought maybe something would come into your mind, some difference in how the second version works with redis or something like this.

[andrewgodwin]

I'm afraid nothing immediately comes to mind - the difference between Channels 1 and 2 is vast (asyncio, Python 3 only, etc.) that there's no one thing that it could be.

[imbolc]

I though it celery who was reconnecting, but actually it's channels-redis, and it isn't celery-specific. I've created an issue there: django/channels_redis#100