AWS Thinkbox Discussion Forums

Houdini Distributed sim

Hello everyone,
I searched the forum for this topic, found a hand full of results, but didnt see a solution for it.

Im basically doing the process following houdini’s documentation, up to the point of submitting to HQueue.
Since I plan on using DL, im skipping HQueue parameters, thinking those would not be relevant for DL. (?)

Once i have the out/distributedsim node created, I submit it using the in-app DL submitter. It submits successfully, but altho Ive set 'frames per Task" to cover the entire sim, the monitor only takes frames 0-1. Im assuming 0-1 are the “slice” index. Since everything “frame range” wise is set to 1-250.
Indeed, after submission, I have frame 1 of both slices created on the folder.

How to have the full sequence simed ? Is anyone using dist sims thru deadline successfully ?



I’m no master simulation man, but it looks like to get it across multiple machines you’ll need to use the “Distribute” options. From the docs:

docs.thinkboxsoftware.com/produc … ation-jobs

You shouldn’t need to play with the HQueue settings directly as Deadline should be driving those options for you under the hood. There are some bits you can configure for it in the plugin settings dialog:
docs.thinkboxsoftware.com/produc … figuration

Try working with the Distribute tool and if you hit bumps there I can keep going there.

Considering that distributed simulation is the trickiest part of Houdini, I’ll add a docs issue for fleshing that section out a bit more.

Thanks Edwin,

Ill get to it in a bit.
I didnt see that plugin config on DL had ports and IP address set up.
The path I have set for simtrack.py seems to be valid, I just need to set it on the distnode then, will try that!
Im unsure which IP address to use tho, Ill try machine1_ip:8000 on the node. lets see where that takes me.
Cheers!

I don’t see a place for an IP address in the configuration. Deadline is going to use the first Slave as the controller, so there shouldn’t be any need to specify one manually.

Did you change any of these to be an IP address?

I apologize for my lame communication skills =)
What I mean was using the DL ports into the DistributedSim node inside Houdini.
If you see the last attachment of my first msg, Thats the DistributedSim node on its default settings. You see the IP + port set to: Localhost:5000.
I was thinking about changing that to Slave1_name:8000 or Slave1_ip:8000 , since 8000 is the port set for DL.

Altho that is a HQUEUE setup entry, I dont see anything else I can change. I will triple check all the steps, but im pretty sure I covered the manual, and no luck on distributing it over 2 slaves. They do 1 frame, and stay there (job stays active, but no process is happening after frame 1)

Would you know if changing the HQueue Server parameter i was mentioning above does anything at all with DL ?
Its seems I cant get it work at all…

I also tried filling “Tracker Address” on the /obj/AutoDopNetwork/DISTRIBUTE_flipfluidobject_CONTROLS.
The port is already at 8000.

… Im just having an epiphany moment. I think I have HQUEUE listening to the same port, could that be the problem ?

No luck.
At the moment here is the status quo.
I have to open the scene using UNC paths. \server\drive\folders
If otherwise, I get errors. “Unable to open scene file”

By using UNC paths i get the output of the first frame for each slice on the folder. The job throws no errors, and stays like this. Aka “frozen” on the first frame.

I tried Hqueue, and I have the same error no mater what i use. Tried UNC and mapped drive to a letter, its seems that HQ is unable to open the scene file either way.
I checked the machines can see the file. I can open houdini on the local machine and render or sim from the UI without problem.
Unable to use DL or HQ tho.

Any insight would be most appreciated.
Thanks!

Hi,
I was wondering if anybody has had any luck with this.
I’m kind of in the same situation where I’m trying to submit a distributed FLIP sim to our farm with Deadline 10.

The job gets on the farm on two render slaves/blades, but does not progress. It stays at 0%.
I’m not too savvy concerning servers and the technical part of a farm and such, so I have no clue where to look now.

I would very much appreciate any help.

Eckhart

I am starting from the beginning trying to get hqueue sims working and found this thread. There was some information gleened from this thread, but I obviously hope there is newer documentation of tutorials on how to set this up since the thread has been dead for a year or two.

Are there any newer threads or documentation on how to setup an hqueue partitioned simulation in deadline?

Privacy | Site terms | Cookie preferences