I have a 400 frame animation using Irradiance Map and QMC for lighting. I set the lighting to multi-frame incremental, and turned on Distributed Rendering. I linked it up with 5-7 other machines.
Slave machines are all running 3DS Max 8 with SP3, and Vray 1.5 RC2. I started Vrayspawner and began the rendering.
First couple of frames work perfectly, with many buckets on the screen corresponding to the network of processors available. All slave machines switch from vraydummy to vrayscene as is expected.
A while later (not sure exactly when, since I left for the night), all the slave machines are back to vraydummy and the primary control computer is running solo, even though nothing changed. I restart the primary computer’s rendering again, verifying it is linked to all the slaves, yet the slaves still do not pick it up.
It takes a full restart on all slaves and the primary control computer to get it running. However, since I run these overnight, that is not possible. This has occured the last 3 nights in a row on all workstations.
Has anyone else seen this issue and can comment? I REALLY need this running, as I have about 5 minutes of animations to get finished by next week.
========================== Starting new session ==========================
Got this module name: C:\Program Files (x86)\Autodesk\3dsMax8\vrayspawner60.exe
Got this INI file name: C:\Program Files (x86)\Autodesk\3dsMax8\VRaySpawner.ini
*** ALERT! Could not open INI file: “C:\Program Files (x86)\Autodesk\3dsMax8\VRaySpawner.ini”. Using defaults.
--
Dumping current settings…
VRaySpawner temp dir: “C:\Program Files (x86)\Autodesk\3dsMax8\network”
Application name (MAX/VIZ): “C:\Program Files (x86)\Autodesk\3dsMax8\3dsmax.exe”
--
Process id is 2856 (0xb28)
Executing ““C:\Program Files (x86)\Autodesk\3dsMax8\3dsmax.exe” -server C:\Program Files (x86)\Autodesk\3dsMax8\network\vraydummy60.max”
Process id is 2856 (0xb28)
Executing ““C:\Program Files (x86)\Autodesk\3dsMax8\3dsmax.exe” -server C:\Program Files (x86)\Autodesk\3dsMax8\network\vraydummy60.max”
========================== Ending session ==========================
Could it have anything to do with the “Executing…” happening twice during a single session? The second time it happened was about 4 hours after the first time.
For animations I would suggest not using DR at all and just let one computer render each frame individually using net render. It is probably faster that way, trouble free, and is the typically the way most companies render animations.
Do your GI calc. one one computer, save the GI solutions, and then send it out to the network.
Or is there a special reason you want to use DR? If you don’t have any moving objects in the animation, a saved Irradience+Light Cache is the best way to go.
Using DR to precalculate IM+lc for an animation (using incremental add to current frame) in multiple computers instead of a single one sounds good to me…
Have u tested it? I cant do it as i cant access any other computer but mine.
That’s an interesting way to use DR, but it would conflict with some rendernodes backburning, as DR is not aware of the rendernode status…
Hehe, no I haven’t tried it either. I don’t see why it wouldn’t work though. We haven’t installed 1.5 yet though, so I’m not sure what the new DR is like.
The new DR is SUPERSWEET. Buckets identify themselves by machine names instead of IP addresses (back to the sweet 1.09 way of things). It’s also super-stable -- even when I cancel a DR, the machines pick right up again when I restart. It’s a godsend to have it fully functional again.
DR also now works in region/crop render modes, which is huggeee. Really, so many little things have been fixed/improved in 1.5 -- I’m really glad I got to shake Vlado’s hand at siggraph.
If anyone tries to precalc an animation using DR, let me know how it works out -- this sounds like an efficient way of doing it, I just assumed I was asking too much.
I have read documentation that says that Incremental Add is not supported through DR for Vray 1.5. Although it should not hurt the precalc of your Irradiance Map, I am not sure it will speed it up any.
i know vlado said it doesnt work. but ive done about 3 projects using incremental add to current with DR. however that was with the internal builds, not 1.5
Is it impossible to make backburner be aware of DR jobs assgined to rendernodes? I am forced to watch out what’s going on in my renderfarm cos sometimes backburner and DR jobs overlap on a single node, leading to crashes and inestabilities. Or maybe the opposite way, force DR to check if the node is rendering something else…