#celery

i use celery with a supervisord. i had started celery (a few days ago) with concurrency set to 10. somehow i see only 2 poolworkers and the mainprocess running now. shouldn't the main process take care of spawning workers to maintain the concurrency set initially?

..and now how do i find why these workers died?

it should definitely replace them, check the worker logs maybe?

you can set the MP_LOG=1 envvar to get more related logging

going to release 3.0.20 tomorrow, would be happy if someone could test the 3.0 branch

asksol: anything I can do to help with celery issue #1129?

hi all. This is showing up under the config section of our flower interface: errorValueError("Unsupported type: <type 'method-wrapper'>",) and it's not reporting on the workers. though they seem to be running. Any ideas for TS?

Diranged: well, you can fix the leak ;)

frogythegreat: probably the configuration was not json serializable

Diranged: I have no idea where it's leaking yet, even replacing it with weakref it leaks 1MB per task

asksol: :P

asksol: yeah you should see the graphs of our celery workers.. ugh..

asksol: thx. I'm pretty sure I haven't changed anything about the configs from the default pip installs. I can see tasks on the tasks tab, and while the queues are listed on the Workers tab, their info in the table is all empty

frogythegreat: I added some serialization support to that recently, e.g. kombu Queue instances etc

askol: ok, pardon the newbness, but not sure I understand?

json has a small set of built-in types, but I have added more types in the development version

o ok

it looks like a bug in conf inspect command

asksol: https://github.com/mher/flower/issues/100

mher, aksol: yeah, that looks like our issue alright

mher, aksol: however I just upgraded flower (and dependencies via pip) from 0.4.3 to 0.5.1 and the problem seems gone.

Although, in that vein; any idea why the workers report as offline when they're fetching tasks from our rabbitmq?

hi everyone! is there a way to specify a two serializers: one for tasks submission, and one for tasks' result ?

s/a two/two

okay just set CELERAY_TASK_SERIALIZER and CELERY_RESULT_SERIALIZER :)

can i specify CELERYD_PREFETCH_MULTIPLIER on the command line?

or better, can I apply CELERYD_PREFETCH_MULTIPLIER with different values to different workers?