It seems like there are some global locks in mongo still, we seem to have corresponding locking going on between deadline 6 and 7… We deleted around 6000 jobs from deadline 6, and at the same time, the deadline7 db also locked up:
Should that be happening?
It is definitely possible that the databases could affect each other if they are competing for things like I/O bandwidth or CPU usage. What’s interesting is that it’s the Limit collection in v7 that is affected. We’re still investigating if there is anything specific we’re doing with the limits database that would explain this correlation.
Cheers,
Ryan
It could be that the limits collection is so overused, that that’s where the symptoms during a global lock arise. The other collections are not trying to lock anything constantly.
( Note, that i find it worrisome that the limits collection has any locking whatsoever with a queue that has no renderable jobs and a completely idle farm … :-\ )
Yeah, there must be a bug here, since the slave should only be trying to get limits for jobs that it could actually dequeue.
Unless maybe it’s something in housecleaning…