Itoo Software Forum

Author Topic: Distributed render problems  (Read 1874 times)

JTP Vis

  • Full Member
  • ***
  • Posts: 122
Distributed render problems
« on: June 18, 2015, 06:50:51 PM »
We're having a bit of trouble with railclone and distributed rendering with V-Ray. We are using Max 2014 on the workstation, with railclone 2.4.7. We are distributing to nodes with Max 2015 and railclone 2.4.7. Up until now we have been rendering very successfully with this set up. The problems have occurred when we have been working with a collaborator. They are using railclone 2.3.4, and they created the railclone objects that seem to be causing the failures. It's worth noting at this point the files in question are very large/heavy and complicated. Possibly not able to upload as the resulting .zip would be huge, but if someone from itoo can help I'll do my best to get it to you.

I have been through all the files associated with the master file and checked and double checked the paths for all files/objects/textures etc. and there are none missing.

The manifestation of the problem is that the render starts on the master machine, the process of transferring assets to the nodes is begun and the nodes start loading the assets/scene into their ram. At an un-definable point (at least I don't know how to define it) the slave machine gets a Max error and the loading process crashes out. The v-ray spawner re-loads and the process starts again.

I am not sure what else to try as I have also checked and double checked the nodes for plugins etc. As I said before, normally they render fine. It's just with one or two of this job's files that have been started by our collaborator. It's worth noting that it's not all his files on this job!

I hope someone can shed a bit of light on this, as we are having to resort to rendering on the desktop machine only, which is painfully time consuming.

Regards,

Bill

Michal Karmazín

  • iToo Software
  • Hero Member
  • *****
  • Posts: 2447
Re: Distributed render problems
« Reply #1 on: June 19, 2015, 10:27:49 AM »
Hi Bill,

As you've mentioned, that you are having these troubles just with some scenes it's very hard to say where the problem could be. Would you please save and send us the minidump file of the crash (from any of these failing render nodes)? (You can get the minidump clicking "View report details->View report contents", checking the path of the "3dsmax_minidump.dmp" file and making a copy of it, before closing the 3ds Max Error Report window). Hopefully, it'll point to the problem source.

Meanwhile, I would suggest you to convert these RC objects to instances using RailClone Tools.

If you could send us a stripped scene (one RailClone object should be fine) to support@itoosoft.com, it would be great. In case of big file size, please use a web sharing service as WeTransfer.

Best regards,

JTP Vis

  • Full Member
  • ***
  • Posts: 122
Re: Distributed render problems
« Reply #2 on: June 19, 2015, 01:40:25 PM »
Max did not crash the first time, it just vanished before it started rendering.

However it worked this time and the next so attached are the "Problem details" from the crashes.

I have tried removing what I thought was the problem object from the scene, but with no success. The problem with such a big scene is that it takes a really long time to open the files to edit them, and Isolating the problem could take ages!

I hope the attached text files are of some help.

Bill

JTP Vis

  • Full Member
  • ***
  • Posts: 122
Re: Distributed render problems
« Reply #3 on: June 19, 2015, 04:41:28 PM »
As a further update:

I have gone into the main landscape file which contains most of the railclone and forestpack objects and have done the "instance" thing. That file is now 380MB as all the models are live. I have updated the links to the master file, and started a test render.

Now this is where it get's interesting... or in fact really annoyingly boring: the render node is registered, the v-ray log says that the node is being used and that the scene has been transferred, but the node does absolutely nothing to contribute to the render! It's ram usage has gone up to 6GB, but it's not doing anything else.

Any suggestions?

Bill

p.s. I have to solve this as the render now takes 8h on a single machine as it can't be distributed!

Michal Karmazín

  • iToo Software
  • Hero Member
  • *****
  • Posts: 2447
Re: Distributed render problems
« Reply #4 on: June 19, 2015, 05:47:05 PM »
Hi Bill,

Thanks for all provided details. Unfortunately, provided logs don't give us a hint of possible problem source. Well, as you've mentioned:
Quote
I have gone into the main landscape file which contains most of the railclone and forestpack objects and have done the "instance" thing.
As after conversion into instances our plug-ins doesn't have any influence at all (this way you can render it even on computers without these plug-ins), it's very hard to say for us what's causing your troubles. Usually, the only way to trace it is by cleaning / simplifying the scene and checking which object is causing these troubles. If the problem persist, I would suggest you to contact Chaos Group regarding these troubles as I'm afraid we can't do much more for you in this case. Thanks for your comprehension.

Best regards,