AWS Thinkbox Discussion Forums

Houdini Distributed Sim and Simtracker

Going nuts here.

I was able to setup a adhoc distributed sim by launching houdini on 2 seperate machines and loading our HIP file with the distributed sim setup.

Entered the Tracker host and port on, and defined manually Slice 1 and Slice 2 on each PC
Hit play on the timeline of both…

Jumped over to the tracker running in verbose mode and saw the output of the nodes talking back and forth to each other.
Walked over to the 2 PC’s running houdini and saw the time lines moving forward and the slices were working. Checked the output path and saw the bgeo.sc files being written with the right naming (…$SLICE.$F5.bgeo.sc)
Great!

Got to the point of implementing this in deadline,
We can get the tasks to launch. task 0 launches running the simtracker, and a render process I can navigate via web browser and see the webserver page running on the node.

Second task launches on another node, I can see in the slave log, that it tells it that it is a different slice, see that it is told the tracker IP and port.
but the nodes NEVER communicate to the tracker.
The node running TASK 0 doesnt even show that it is connected to its self since it is running both simtracker and hython

Firewall is OFF. on all machines.

If we leave it alone, the sims progress, but since slice 0 and 1 dont have any idea whos doing what, the only slice that actually has any particles in it is slice 0 then slice 1 is basically empty cache files.
Slice 0 finishes with time spent on creating the particles
Slice 1 finishes really fast because it doesnt know what slice 0 is doing.

What am I missing here?
Edwin or Russell, Do you guys know what might be going on here?

[attachment=0]Task0.PNG[/attachment]
[attachment=1]Task1.PNG[/attachment]

I think the folks who know the most are going to be the integration team here.

The fact that they work outside of Deadine and the firewall is disabled makes me guess right now that it may be the timing on when the tasks are being picked up. If I recall correctly, slow joiners don’t go very well. Other than that, I’m not sure.

There are a few other threads about Houdini Sims, if only I could @mention Nick here. :smiley:
forums.thinkboxsoftware.com/vie … 11&t=15157
forums.thinkboxsoftware.com/vie … 11&t=15618
forums.thinkboxsoftware.com/vie … 11&t=14629

I’ll run it by the guys and see if they’ve got any other ideas in case the slow start isn’t the issue.

Update: The timestamps there are definitely showing that they started around the same time so that’s likely not the case… The exceptions though are interesting. I’ve checked our code and we do check the Rop’s type, but the error seems to be coming from “hou.session”. For those who use Houdini on the regular, have you seen that exception before? It doesn’t seem to be one of ours.

Privacy | Site terms | Cookie preferences