This issue seems to be happening for me too. If I start a number of workers via celeryd-multi, the worker hangs on a task when it involves calling out to my database. If I start one worker, the tasks are handled fine.

Ask, you mentioned in irc once that the -P eventlet arg should be passed as the first argument when running celeryd (which solved the exact same problem I was having with just one worker...). I noticed that if I check the processes spawned by celeryd-multi, they look like this:

/usr/bin/python -m celery.bin.celeryd --config=celeryconfig -P eventlet -l INFO -n n4.ras --logfile=/home/user/dir/logs/celeryd.log [email protected]

The celeryd-multi command I used: celeryd-multi restart n1 n2 n3 n4 n5 -P eventlet --config=celeryconfig -l INFO --logfile=/home/user/dir/logs/celeryd.log

It would seem that the -P arg is not the first argument, so maybe since this fixed my problem with one worker this could fix the celeryd-multi problem as well?