#celery

asksol: hey, are you here?

Hi, I have chained two tasks using the `chain` signature. I have also provided the (only) argument for the second task as a return value from the first task. However, I can see in my (sys)logs that only the first task is ever accepted and executed by celery. The second task never shows up. Not exactly sure what I'm doing wrong.

http://pasted.co/3403c431 this is what it looks like, pretty simple 2 lines of code. I tried the pipe syntax as well but to no avail.

Hello everyone. I have a question regarding task registration.

Twitter: crank7

Currently when starting up a worker all tasks in my project are being registered with this worker, even if they are in some required dependency.

Is there a way to control this? I don't want to startup all workers with all tasks, but have different workers who execute different tasks. I know I can solve this with a queue, but I don't want them to load the code in the first place.

notanormalnerd: if you don't import modules contain tasks or autodiscover them using app.autodiscover_tasks - they won't load

notanormalnerd: understanding what imports what will help you to control which tasks are loaded when

Hi. I am needing to control the number of workers per queue, starting more workers, etc.

cyme (2014) tries to do that but is highly outdated

1) is it worth to update the thing or

2) is better to redo the features I need using management signals?

control* signals

malinoff: Thanks for the quick response

malinoff: I will have a look into that

alanjds: use ansible - worker processes management is something orthogonal to celery, it's job for a sanely written process manager like systemd in conjunction with a simple ssh wrapper

alanjds: or you can leverage your cloud provider's apis

w/e

malinoff: I would prefer something less infra-involved.

The use case is less to control machines going up and down and more to control the balance of how much workers are subscribed for each queue and

cyme is supposed to do that, for the docs embed on the repo

my doubt is: For an app that handles the queue subscriptions, is better to control this stuff via celery control signals, or to delegate to an external tool like cyme? (counting the burden of make cyme work again)

alanjds: definitely external tools

Ok... can you elaborate on this?

plz*

The flaw I saw on cyme is the "unsync" between the internal state DB and the real world

talking on the control bus directly could fix that easily, I guess

but I am not the one that worked with the control bus yet on life

alanjds: I'm not talking about cyme. Generally, celery itself shouldn't manage its worker processes, this is something external

and should be managed by external tools

not cyme specifically

brb ~30 minutes

@malinoff for the strace dump

which process should i trace?

malinoff: ok. tkx. As cyme repo in on celery/cyme github org, I would expect it to be the "recommended example" about how to do stuff

hey im working in django 1.9. I previously put `import myapp.tasks` in my `myapp/__init__.py` file... but i need to move it to the `myapp/apps.py` django AppConfig due to application loading... but when I do this I get a very odd error whenever I try to start my celery workers: http://pastebin.com/J7qB7Fc6 it says that the celery application cannot import celery... which I don't really understand. what could be my

problem? when is the appropriate time to import my tasks?

is anyone here able to help?

OK i've put my issue up in github: https://github.com/celery/celery/issues/3349