Pending dependencies require manual release at long queues


#1

I’m curious as to what the exact process for releasing dependencies is. We render tiled vrscenes with dependent assembly jobs submitted at the same time all via the vrscene plugin. Everything works great.

Sometimes, when we have a whole bunch of jobs queued (all same priority), we’ll notice that the available slaves will jump onto the next full job and start rendering those instead of jumping onto the assembly job, because the assembly jobs haven’t been released yet. Occasionally, this will build up and after more than 10 or 20 minutes, we’ll have to “release” the pending assembly jobs and then the slaves will jump back and assemble the tiles.

I was just curious as to what the process is for determining when a job is released and if there is maybe something we are doing on our end that is causing this build up to occur.


#2

There is a pending job scan that runs on an interval. You can view the settings in Tools > Configure Repository Options > House Cleaning > Pending Job Scan.

Normally a Slave would do this scan by default, if they are busy then they cannot run it. A great solution to this problem would be to run Pulse in your farm. Now a specific machine will be tasked to doing Housecleaning, Pending Scans and Repo Repair versus relying on your Render nodes.

docs.thinkboxsoftware.com/produ … pulse.html


#3

Wow! Yeah, I actually never even knew that’s what pulse was for. We have a small’ish farm, so I just assumed it wasn’t needed. Fired it up now and we’ll see how it goes. Thanks, this is actually super useful!