The "--flags expand" option doesn't seem to work. (Qube! 5.1)

[speedtrials]

I'm using Qube 5.1.0 and the expand flag doesn't seem to work.  For example:

qbsub --flags expand --range 1-5 Render -r mr -s QB_FRAME_NUMBER -e QB_FRAME_NUMBER /path/to/mayaSceneFile.ma

In this case, the frames get rendered one at a time on one machine.  However if I use this:

qbsub --cpus 100 --range 1-5 Render -r mr -s QB_FRAME_NUMBER -e QB_FRAME_NUMBER /path/to/mayaSceneFile.ma

then frames get rendered concurrently on multiple machines.  Any ideas?  Also are there any drawbacks to just using

--cpus 9999999

for all the qbsub submissions to use all the available processors?

[eric]

I'm using Qube 5.1.0 and the expand flag doesn't seem to work.  For example:

qbsub --flags expand --range 1-5 Render -r mr -s QB_FRAME_NUMBER -e QB_FRAME_NUMBER /path/to/mayaSceneFile.ma

In this case, the frames get rendered one at a time on one machine.  However if I use this:

qbsub --cpus 100 --range 1-5 Render -r mr -s QB_FRAME_NUMBER -e QB_FRAME_NUMBER /path/to/mayaSceneFile.ma

then frames get rendered concurrently on multiple machines.  Any ideas?  Also are there any drawbacks to just using

--cpus 9999999

for all the qbsub submissions to use all the available processors?

The expand flag will dispatch subjobs until the first frame completes, so if your frames are very short, you could conceivably have a single subjob do all your frames.

If you want to force a certain number of subjobs, then I would go with the second form, however I wouldn't request any more than the total number of licensed hosts. Any more, and you will just waste time dispatching subjobs even after all the frames are completed. They'll run, but they will consume resources until they find out there are no frames to render.

--cpus 9999999

would be very bad. You would lock up your farm until 9 million subjobs ran, and you would have to kill them to get anything going. Even killing 9 million subjobs would take a non-trivial amount of time. Please don't do this.

[speedtrials]

I'm using Qube 5.1.0 and the expand flag doesn't seem to work.  For example:

qbsub --flags expand --range 1-5 Render -r mr -s QB_FRAME_NUMBER -e QB_FRAME_NUMBER /path/to/mayaSceneFile.ma

In this case, the frames get rendered one at a time on one machine.  However if I use this:

qbsub --cpus 100 --range 1-5 Render -r mr -s QB_FRAME_NUMBER -e QB_FRAME_NUMBER /path/to/mayaSceneFile.ma

then frames get rendered concurrently on multiple machines.  Any ideas?  Also are there any drawbacks to just using

--cpus 9999999

for all the qbsub submissions to use all the available processors?

The expand flag will dispatch subjobs until the first frame completes, so if your frames are very short, you could conceivably have a single subjob do all your frames.

If you want to force a certain number of subjobs, then I would go with the second form, however I wouldn't request any more than the total number of licensed hosts. Any more, and you will just waste time dispatching subjobs even after all the frames are completed. They'll run, but they will consume resources until they find out there are no frames to render.

--cpus 9999999

would be very bad. You would lock up your farm until 9 million subjobs ran, and you would have to kill them to get anything going. Even killing 9 million subjobs would take a non-trivial amount of time. Please don't do this.

Thanks for the quick reply.

I upped the resolution so frames took a ridiculously long time to render, but the expand flag still did not work.

I also tried it with --cpus 9999999 and there were no ill effects to the render farm.  Multiple sequential submissions with --cpus 9999999 went thru without any problems.

[eric]

I'm using Qube 5.1.0 and the expand flag doesn't seem to work.  For example:

qbsub --flags expand --range 1-5 Render -r mr -s QB_FRAME_NUMBER -e QB_FRAME_NUMBER /path/to/mayaSceneFile.ma

We have identified an issue with in 5.1 that prevents expand from working correctly. I will be fixed in an upcoming release. Thanks for calling this to our attention.