Hi there!

I have a file with a pyro simulation very heavy in voxels count. It's so slow in my machine. As far as I've been reading, I think its possible to distribute pyro simulations in different machines with Hqueue.

So here are my two questions:

1. Is this also possible to distribute my simulation with Amazon's EC2 Cloud? (Slice method I guess?)

2. Since my voxel count is so heavy that can crash my machine, is it possible to simulate each slice separately in my machine?

Thanks in advance!

Ok, I've found that it's not possible to distribute sliced simulations in the cloud. But that info seem to be from 2009. Is it still like that?

I just tried to use Hqueue in H14 Indie and it said it is not available in Indie Bummer

SESI says this may be a bug, they are investigating.

pbowmar
SESI says this may be a bug, they are investigating.

Out of curiosity, did you ever hear back on local Hqueue + Indie?

I tried the latest H14 (248) and it seems to work now, yay!

Great news, thanks!

Sorry, I take it back

hou.OperationFailed: The attempted operation failed.
Error: Indie does not support generating disk files.

pbowmar
hou.OperationFailed: The attempted operation failed.
Error: Indie does not support generating disk files.

That error should be fixed in the latest H14 build.

Did you also update the Houdini version that runs on the HQueue farm?

Cheers,
Rob

Hi Rob,

Using latest 265 build, installed (update mode) Hqueue 265 as well, still get:

The Houdini 14.0.265 environment has been initialized.
PROGRESS: 0/24
Traceback (most recent call last):
File “/opt/software/hfs14.0/houdini/scripts/hqueue/hq_make_ifds.py”, line 4, in <module>
hqlib.callFunctionWithHQParms(hqlib.makeIFDs)
File “/mnt/opt2_900gb/opt/software/hfs14.0.265/houdini/scripts/hqueue/hqlib.py”, line 1848, in callFunctionWithHQParms
return function(**kwargs)
File “/mnt/opt2_900gb/opt/software/hfs14.0.265/houdini/scripts/hqueue/hqlib.py”, line 120, in makeIFDs
_invokeRopAndCatchErrors(rop, project_name, frame, False)
File “/mnt/opt2_900gb/opt/software/hfs14.0.265/houdini/scripts/hqueue/hqlib.py”, line 441, in _invokeRopAndCatchErrors
raise e
hou.OperationFailed: The attempted operation failed.
Error: Indie does not support generating disk files.

Actually nothing to do with Hqueue, just run Hbatch and try to dump IFDs:

/out -> render -V TorusTest
Rendering 24 frames ( from 1 to 24 by 1)
11:02:57 TorusTest frame 1 (1 of 24)
Error: Indie does not support generating disk files.
Render failed.


So sad

Arg. Looks like I was incorrect in stating that Houdini Indie now allows IFD generation. Sorry about that.

So you need to turn off IFD generation in the HQueue Render ROP. In the parameters pane, go to the Mantra Options tab and uncheck Generate IFDs. Then submit a job.

Cheers,
Rob

Unfortunately, that's not really pratical The memory footprint of Hbatch + Mantra together can be prohibitive, requiring you spending more $$ on a larger Amazon machine. Plus it's much slower than dumping out your IFDs from one process then using Mantra to render them, since you then have to run Hbatch for every frame.

Anyway, I guess an issue for higher-ups. I will cease attempting to use Amazon Web Services. I never did get it to work anyway, but that's likely a separate issue.

It's unfortunate these license limitations are so crippling when it comes to HQ; I thought it was just supposed to “work” based on your license type.

I feel the solution is either support lc/nc ifds; or mimic Pixar and make Houdini limitations license-only, and avoid altogether tiptoeing which license supports feature X, and to which degree. Honestly, I think the latter would be best in the long run, since anyone who'd break those agreements is probably just going to pirate the software anyway. Not having software limitations (maybe just keep existing watermarks on renders and in the scene view?) frees up developer resources/time, good PR, opens up for more folks to evaluate and learn all aspects of Houdini, etc…

Submitting an RFE now…

I could not agree more, that having any restrictions beyond watermarks seems counter productive.