Gamma/Lut differences on server nodes?

[MegaPixel]

Just wondering if anyone here who has experimented with LWF (Linear Workflow) has ever encountered problems with network servers returning darkerthen usual frames? I have restored the defaults for gamma/lut on my system but it seems like every now and then (and it isn't consistant), my render servers will produce different frames then my own computer does? Thoughts on this?

Dave

[andyp7]

Are you sure its the gamma? Perhaps your render nodes aren't picking up the irradiance maps or light caches. Assuming you are using that method for your lighting.

If not, double check you have "Load enable state with max files"

[MegaPixel]

Thanks, as it turned it was related to not beign able to find the .vrmap file. I tried to do both the IR precalc and animation pass through the backburner by setting up a dependancy on the precalc before the anim. pass started. Not sure why, but when the anim. pass starts, and can't seem to find my .vrmap file even though I specify exactly where it will be. I recall maybe hearing something about how you shouldn't have a drive letter in the path but a computer name instead? Is there an issue with that do you know?

Thx , Dave

[andyp7]

We never used to be able to put the irradiance maps or lc on the main server and have the render nodes pick them up.

We used to have to put the saved irradiance maps into a local directory on each machine (e.g: c:\rendercache). But since we got our new render drive it works fine. I think its because the path is very small. For us its r:\rendercache so the servers all pick it up.

This is only our experience though, perhaps something else causes the machines not to pick up the irradiance maps.

[jujubee]

Make sure you have the same scripts (colorcorrect, other mtl scripts) on all of your nodes. This I think is the main cause.

[MegaPixel]

So in other words... it's not possible to send an entire project off to the backburner and walk away - that is, a project the requires a precalced IR map? What you guys are telling me is that I still have to baby sit the process, wait for the IR map to precalc first, then copy it around to every node in the farm before launching the final render pass on the backburner?

I just recently reconfigured my backburner system so that I didn't have to copy all files locally to every server too. Guess I'm gong back to the basics

[timmatron]

On our last big animation, we never had to copy the GI maps to the local machines. They could all read them off the server, and believe me our server at the time was nothing special.

[MegaPixel]

Could you run through real quick how your network is configured? Or even better, your process for sending your job off to the farm, assuming you are sending it from a workstation and not on the manager itself?

[timmatron]

Well it was pretty straightforward. Just make sure that all of your computers have mapped the server to the same drive letter, like z,x,w, whatever. If they are not all named the same, they will all be looking in different places.

Then we usually just calculated the GI on one computer and saved the GI maps to the server. Then switch the IRmap and LC to 'from saved file' and load up the appropriate file on the server. Then you just send off the job through the network, and really that should be all there is to it. Of course make sure you are saving all the frames to the server and not the local drives.

[MegaPixel]

Well, unfortunatly, you just described exactly how I've been doing it. I have an "R" drive mapped to all render servers so that every machine sees and writes frames to the same location. After I send off my precalc job, I then switch my ir map to "from file" and select the file on the server where it will be saved, even though the file technically doesn't exist yet. I uncheck "don't render final image" setup my frame file destination and check netrender. I then set a dependancy on the job to wait to start until after the precalc job is finished.

I wonder if there is a problem with 15 pcs suddenly hitting the network drive at once looking and accessing information in the same exact spot...

[timmatron]

We weren't sending the GI calc. as a network job. We would calculate it on one machine locally, make sure the GI maps saved out, and then made sure that the render settings were then set to "from saved file" and that "don't render final image" was turned off. THEN we sent that job to the network.