NEXT is a machine learning system that runs in the cloud and makes it easy to develop, evaluate, and apply active learning in the real-world. Ask better questions. Get better results. Faster. Automated.
changed requirements file for docker image (no longer celery==3.1.25)
Added a celery.worker_ready function that debug prints to console (basically just mimicking the celery 3 behavior (e.g. sync_worker_2@2affb02ed780 ready.), which is no longer the default)
updated celery cli options in next_worker_startup.sh to reflect new new version of celery. (this is what was causing it not to work before)
Copying some comments from @dconathan over from Slack:
I have moved to celery 4. I have noticed getQuery has been pushing .5-1s delays for no apparent reason and I haven't gotten around to debugging it... this could explain that.
Tomorrow I'll rebuild with celery 3.1.25 and see if that fixes it...
Just reverted back to celery 3.1.25 and getQuery delays are back down to sub .1s range… so almost certainly seeing a slowdown due to celery 4
I was wondering about it. Almost all the extra time was “enqueued” time according to the charts.
...
I guess there’s one reason not to switch to the latest and greatest
as per #159 this updates to celery 4.0
Changes:
celery==3.1.25
)sync_worker_2@2affb02ed780 ready.
), which is no longer the default)next_worker_startup.sh
to reflect new new version of celery. (this is what was causing it not to work before)Note that I didn't mess with any of the options in
next.constants
- they are transitioning from upper case to lower case but our constants as is will be supported until celery 5 (?). I tried running their "automatically fix it" script and it broke things so I didn't bother... see http://docs.celeryproject.org/en/latest/whatsnew-4.0.html#step-2-update-your-configuration-with-the-new-setting-names