The two systems (the worker-manager and the web frontend) really don't talk to each other well enough! By design, really, so that slowness/load in one doesn't also overwhelm the other. The workers really are fire-and-forget, and the individual worker jobs don't know anything about each other really -- like, if I fire an import job right now, the process that spawns has no concept that it's in a queue. It just hangs out until the worker-manager can grab it.
Mostly it's the kind of change that we don't want to spend time on because it'd be a lot of work, a very high-risk change, and it's not needed 99.9% of the time and the .1% of the time it could be useful, it would also add a bunch of load. We have a lot of other stuff we need to be doing instead, and very limited time/resources/peoplepower!
no subject
The two systems (the worker-manager and the web frontend) really don't talk to each other well enough! By design, really, so that slowness/load in one doesn't also overwhelm the other. The workers really are fire-and-forget, and the individual worker jobs don't know anything about each other really -- like, if I fire an import job right now, the process that spawns has no concept that it's in a queue. It just hangs out until the worker-manager can grab it.
Mostly it's the kind of change that we don't want to spend time on because it'd be a lot of work, a very high-risk change, and it's not needed 99.9% of the time and the .1% of the time it could be useful, it would also add a bunch of load. We have a lot of other stuff we need to be doing instead, and very limited time/resources/peoplepower!