Loosing renderslaves during DR jobs

[Phantom-Studio]

Hi all,

Since Vray 3.4, we are loosing some of our renderslaves during Distributed Rendering jobs...
About the job : just a still image 5250 * 3000px, nothing crazy.

Our renderslaves are running VRay standalone on Fedora 23 x64 (minimal).
Our workstation are running the same version of Vray but Workstation on Windows 7 x64.

The strange thing is that the process vrayslave isn't crashing at all.
The 2nd trange thing is this line in the log :

Average rays per light cache sample: -nan (min 2147483647, max 0)
Please find below the log :

log-vrayDR.zip

Any help would be great!
Thanks

Have a good day.
Regards.

Jonathan

[svetlozar.draganov]

Would it be possible to describe in details what exactly happens, or even better to capture it on video?
We've got a similar behavior a few times in our environment where the buckets disappeared one at a time after their completion without any errors printed in the log file etc.
The buckets eventually join the rendering again once all the buckets vanished.
As far as I know we haven't had a chance to distinguish what exactly happened since it only happened a few times.

Does this behavior appears with every scene or just with particular ones?
Does it happen on a very very simple scene, just basic primitive and a dome light?

Is that version 3.40.01 or 3.40.02?
If you've only tried one of them would it be possible to install the other one and see if the issue is still there?

Have you tried to disable Light Cache?

[Phantom-Studio]

Well,

In order to answer your question :

1) It appear on every scene
2) Not on a very simple scene
3) All version since VRay 3.10.01
4) Didn't try to disable Light Cache

It seems to appear on big resolution renders (at least 5K).

But we encounter 3 types of problems :

1) As you described
2) As you described but with vrayslave's process crashing
3) The buckets are not appearing at all ("renderslave is not responding" even if they received and start the job, and no missing assets) and no error in the log file

For the third one, we've found a workaround :
Simply bake the Light Cache.

Hope this help.

Have a good day.
Regards.

Jonathan

[svetlozar.draganov]

Thanks for the confirmation.
Let's first try to reproduce the issue into our environment since this will allow us to troubleshoot it more effectively.
In order to do that would it be possible to send us one of the scenes that reproduces the issue every time along some information about your network render setup like number of machines, render managers and whatever you are using. If we have to perform any specific actions besides replicating the network setup and hit render button please let us know what we should do.

FTP for the scene upload could be provided if you need, it just request it on support@chaosgroup.com along a link to this tread.

It would be helpful to find out if the issue exist or not when LC is disabled, so if it is possible please run one more tests with LC switched off.