What footages do you use? Most such crashes I have heard of depend on media that requires conforming/ indexing and is handled by MediaCore. Also, do the clients have their own file repository or do you access some the files from a shared location? Could just as well be too many clients accessing the same files then...
We have a render farm running, with the farm being controlled via Deadline.
Not all render jobs seem to cause the issue, but some. They all access the files, from a high speed file server.
This crash issue, has only started since we updated some of the plugins, and when we moved to CS4 on the farm.
Now I have read in some of the threads, that plugins that use open GL might be a cause of the issue?
Have you seen that before?
As a summary
Farm is Xp SP2 dual core AMD CPU's, and the boxes have 2Gb ram each
They are using the fully patched and licenced version of CS4 Production
If OpenGL should be the issue, then we are talking Red Giant and Boris plug-ins, right? Which ones do you use then? Quite possible that they can't do their magic due to insufficient graphics card memory, though that mostly affects MB Looks (due to its separate UI and buffer requirements). The otehr plug-ins ofte nonly use pixel shaders and partial acceleration and are less critical in terms of doubling the required memory. In theory they should even run on cards with not so much memory (256MB only), assuming there are enough shader units. Hard to tell from the distance. I'd definitely create some test projects to hunt this down...
Actually we use both on the farm
But is there any way to get a better log from CS4 of the crash ?
All we get is the normal Windows termination box.
Surely there needs to be some sort of Log in CS4, which will help me nail it down?
I'll check on the card models, when I can next remote in, in a few hrs time.
Those machines are running an onbaord Nvidia 6100 chipset, we are going to try a beefy video card in one of them to see if that helps.
You can try to use the After Effects Log.txt method. Create an empty file in your user home directory where the other AE prefs are stored, and it will log its "ticks" to that file. The file itself only holds a hundred or so steps, but that should be enough to contain the actual event of the crash. Likewise, you can use Extra per Frame Info in your render queue settings to get some more detailed render logs. Other than that, you could of course try and untwirl the render info to see which comps and layers are being processed. Unfortunately there is no way to pipe that info into a text file, so you literally have to stare at the screen al lthe time...
Ok, We have narrowed down the plugin to Knoll light factory 2.5
We run it in command line, and we get the following;
Pure virtual function call
Then CS4 crashes.
Now, I have tried your logging idea, and all we get is a blank txt file, so its crashing before it can log the error.
Now, we have tried this with different video cards (both nvidia cards) and we have the same issue.
I have updated the C++ runtime library on the box (2005 to 2008) with the same result.
I have put in a request to knoll, and I am waiting on a response.
In the mean time, any other ideas?
Do you disable the GPU option in the plug-in when saving the project? The render engine should respect that. You can also render projects using older versions of Knoll. The effect controls are compatible, so e.g. a 2.5 project can still be rendered in 2.0. This might also be an option. Other than that - yes, RG was working on those issues, but so far nothing has come of it. Looks like this will have to wait until they release v3...