(OPEN) [HOU-3998] houdini 19.5.569 vray 6.1 cgru render error

[maya_bug]

update to vray_adv_61009_houdini19.5.569_bbaa8ca_22548_insta ll_linux / vray_adv_61009_houdini19.5.569_d04ea26_22549_insta ll_linux
still not work

[tsvetomira.girginova]

Hi again,

We will contact you in the already opened ticket to arrange a TeamViewer session and take a look.

[len_xie]

same here ,using houdini and deadline, not all machines, older machines only, window10, random
2024-10-16 10:59:20: 0: STDOUT: Finished Rendering
2024-10-16 10:59:52: 0: STDOUT: 10444: Fatal error: Segmentation fault

[bdancer]

> not all machines, older machines only,

What does "older" mean?
Is there a way to figure out which exact machines are crashing and how are they different from machines that are not crashing?
Is result image file written or crashes during rendering?
Are you rendering with Standalone or hbatch/hython?

[len_xie]

by old I mean machines with Intel Xeon CPU E5-2600 Windows 10 64-bit

they randomly show an error with about a 10% chance. I could not find any differences because the machines that errored out could render normally later in the same job.

The image is written successfully; I actually stared at the folder and waited for the error. Once it appeared, I paused the Deadline job, and the image was there. However, the error triggers a requeue of the frame, which wastes a lot of time.

I am rendering using Hython and the Deadline Houdini plugin.

Other machines with Intel Xeon CPU E5-2698 V4 and Windows 10 64-bit never have this problem.

All machines do not have GPUs and cannot open the Houdini interface at all, but they used to render just fine.

The problem first appeared around July of this year, after some V-Ray or Houdini update; I can't remember which. Full error details are attached:

[chakarov]

Hello len_xie, I couldn't reproduce the issue on my side. Can you share a simple scene with the issue?

[len_xie]

I can upload a test scene, but it's really just a box and a camera, if you can't reproduce it with the same cpu, then it could be my system problem, is there a way to use some post frame script to force a normal finish rendering for frames, I can live with the error, but requeuing is too much waste of time.
--------------------------------------------
managed to use a deadline monitor pre task script to compare the output image modification time with current time, if small enough, fail current task , so that it won't waste time rendering it...

[Dubbie99]

I am getting sporadic segfault crashes on deadline AFTER the frame has finished an successfully written out. I am only using VRay CPU. In a sequence of 150 frames sometimes none do this, sometimes 25 will do it. occasionally it happens a couple of times in a row on the same frame. Its not common to any one machine. We use rez to control the environment on all our machines, so we can be sure every machine is running the exact same software. We are on Rocky 9.1.
Its kind of a pain because if you leave jobs unattended running on the farm, sometimes you are rendering 20% longer than you need to due to pointlessly re-rendered frames. If I am watching the jobs, I can just manually mark the failed tasks as complete and everything is ok.
I have not managed to make a test file that does this behaviour. There seems to be no way to predict which scenes will do this and which won't.

I have attached an image of the deadline log.



this image shows the sporadic nature of the crashes:



oops - missed reading this page of results. Can confirm I am seeing the exact same issue as len_xie

[bdancer]

Dubbie99
1. Is this happening with some recent build?
2. If there is a crash stack - is it the same as for len_xie ?

[Dubbie99]

I think this started about 8 months ago maybe? yes the crash is the same as len_xie