#celery

malinoff: you can do that if you want to, but not required

anyone got an idea why result doesnt seem to get set until after the task is executed?

why would the result be set before the task is executed?

asksol: as far as i understood an update_state call will write the meta arg into result

asksol: i get the state name updated fr my task, but the meta i did send along isnt stored anywhere

the meta argument should be available in result.result

asksol: so flower wont show it, but programs could extract it

flower doesn't use results at all

asksol: there is a table row that shows them

im currently switchimg some legacy code that stored status updates in the database over to using celery builtin methods

no, flower does not use results, monitors use events

asksol: when i click a task in the tasks tab, flower shows the results - so i thought tht was the actual task result

asksol: so flower will only load the result of a completed task?

ronny_: asksol tries to say that the result you can see in flower is not received from result backend, but from event that celery fired right after the task has finished

i see

so i'll proceed with the rest, thanks for the clarification

so bascially all i do uses the sync api, but flower does all via events, thus i should expect differences inthe data i see due to different perspective

flower subscribes to events, and is completely separated from the result backend. it does not use results for anything. Results are not used for monitoring, since they may be private to the process that requested the task, cannot be truncated and platform specific

Hi there, I just noticed a recurring job circulating through our rabbitmq infrastructure, that looks like the terminating job of a chord.

There seems to also be a bug about this: https://github.com/celery/celery/issues/1700

So I'm hopefull this will be fixed in the future

but: what do I do about the messages circulating in my message queue and clogging it up?

is there any common pattern to ensure some tasks run serial per set of signifikant parameters?

ronny_: take a look at chord's

dwt: doesnt look correct

dwt: i want at a single time no more than one of a certain task with a certain parameter

Well you can use chords to just chain them together that all of them are chained

then no two of them will execute at the same time

use only one worker

while that was a joke, you could create a special queue for this kind of task and only have one worker for it

dwt: then i'd have to findd the old one

ronny_: Well, preferably you create them all at once. Then you can set up the chord and forget about all of it afterwards

dwt: because something like user a on webserver 1 starts it and user b on webserver 2 starts the same task with the same parameters

Too bad.

Then it probably doesn't work

rbu's suggestion looks serviceable though

rbu: 2 tasks of that type may run in parallel if, and only if the parameters are different

(it seems like this is begging for some extra tracking)

It does.

tasks can store their parameters in a memcache/key value store before starting

something like redis

... which you can also use to implement celery over

hmm

how would i create a redis task in a way that it only succeeds if i can also atomically put its key into a mapping where it shouldnt exist

or is there a different approach

hi all. I'm struggling a bit with tasks group. I have a task receiving a report, containing several base64 encoded images. I send each b64 to a subtask, in charge of decoding the image to byes, saving it to the disk, uploading it and returning its url.

How am I to gather the results of all subtasks in the group without using get() ?

When I do, I get a warning "RuntimeWarning: Never call result.get() within a task!"

to be clearer, I need to gather the results of all subtasks in the calling task, and insert these results in DB

Hi - I have a question regarding the python inspect() interface.

Anyone who might be able to help?

can you chain tasks without passing the result of the first task to the second? Or should I just ignore the argument if I don't need it?

frodopwns1: see immutable subtasks