Satellite Rendering

[A.Wilkins]

Hi there!

I’m having some problems setting up Satellite Rendering on my network. I have followed a guide on softimage.wiki and everything seems to be set up perfectly, yet the master is still not utilizing the slaves.

I have done the ”Configure satellite distributed rendering” on both the master and the slaves, and made sure they are both using the same port.

The “raysat3_6_53_22server.exe” service is running on both the master and the slaves.

In the windows firewall, the ports have been added as an exception.

We are using a license server and have sufficient licenses.(we are using Advanced Academic)
Sofimage opens and works on both the master and the slaves.

The .ray3hosts file has included the slaves, in the format: computer name:port number.

The “ C:\windows\system32\drivers\etc\services “ includes the “mi-raysat” with the correct port number.

I can use the ping command ”ping <hostname of slave>” and the master can connect to the slaves.

I have used the telnet command ”telnet <satellite hostname> <port>” and get a blank screen, which is apparently correct. So, the port is working.

So, I’m now at a loss as to what to do to get this to work. Any help or suggestions would be greatly appreciated.

Thanks.

[A.Wilkins]

Is it possible that the Softimage Advanced Academic license does not allow for satellite rendering?

When I said the "master", I meant it as in the computer in which i configured the .ray3hosts file to include the slaves and is the one in which I am trying to render off. Is it possible that you have to use the license server?

I'm not getting any error messages at all. Which is most annoying, cause I have no idea at what point this is not working.

[A.Wilkins]

Hello again

Problem solved on another forum!

“go into the setenv file and add the line

set MI_FORCE_OLD_BSP=1”

So, if anyone gets the same problem. There you go!

[Hirazi Blue]

Well, I''m no expert, but this just doesn't make any sense to me:
The environment variable "MI_FORCE_OLD_BSP" shouldn't have any bearing on satellite rendering, as it's "merely" a "switch" between two acceleration methods in Mental Ray, the "old" BSP and the new BSP2.

If setting said environment variable does the trick, that's obviously nice, but adding "MI_FORCE_OLD_BSP" to your setenv.bat shouldn't influence your satellite rendering in that way at all. There are some other environment variables that influence distributed rendering as stated in the manual.

But maybe (and quite obviously) I am missing something here...

[xsisupport]

My understanding is that BSP2 introduced some socket communication issues, and mental ray uses sockets for communication between the satellite clients and server. By switching back to old BSP, you sidestep those socket issues that cause problems for satellite rendering.

[Hirazi Blue]

Ah, thanks for clearing that up...
Does that mean all the "newer" MR-features (like IP and such) are lost to satellite rendering?

[Firebird]

Just to add here.

The same satellite issues where also in pre Bsp2 times when you switch on dynamic bsb. then you will get the same error even with bsp 1 so i think that they somehow use this dynamic thing or something similar in the bsp2 so satellite just don't work.

[xsisupport]

Ah, thanks for clearing that up...
Does that mean all the "newer" MR-features (like IP and such) are lost to satellite rendering?

Yes. From the docs:
"...certain newer mental ray features do not work properly with this MI_FORCE_OLD_BSP setting, including irradiance particles, assemblies, the ambient occlusion cache, and some final gathering optimizations."

And since animated ICE instances use assemblies, that means you lose them to when you switch off BSP2 (unless Softimage sliently decides to use BSP2, which I believe it does in some cases).

[A.Wilkins]

Wow, thanks for the great information!

So, im getting "socket communication issues" then? Does anyone know of a way to fix this?
Or, if there is another way to get the BSP2 to work? Some sort of work-around?

But, besides that, I have done some tests and it seems irradiance particles, ambient occlusion cache and ICE instances are working fine. So, maybe its as xsisupport says and Softimage is just using BSP2 anyway. Which is fine by me!

[A.Wilkins]

The problem is, I do seem to be getting some crashes when rendering with shadows. No shadows works fine.

All I did was create a sphere and a light. Turn on shadows. Subdivide the sphere twice and duplicated it, and Softimage just closes whenever I render. It closes as soon as I get to about 15 spheres and I turned off satellite rendering and this doesn't happen.

I was told that using this Force_old_BSP command might cause alot of crashes and is actually slower. So, I would rather not use it unless I have to.

[Hirazi Blue]

I read the remark I guess you are refering to, by bronco67 on the XSIBase, differently:
BSP2 is slower and crashes a lot...
(I can't confirm or deny this,
but reading the differnt forums there seem to be a lot of people
disliking the whole BSP2-acceleration a lot) ;)

[A.Wilkins]

Yeah, I've been reading about this so much my brain could explode.
I have managed to fix the problem I was having when satellite rendering with shadows. I was running Softimage version 7.0, so I upgraded to 7.5 and this seems to have fixed that problem.

I still have to use the "set MI_FORCE_OLD_BSP=1" command, because Softimage 7.5 was crashing almost instantly without it (when satellite rendering). But, atleast now I dont seem to having anymore problems with it. Who knows what will pop up though, so I think i'll just carry on testing for awhile. And fervently pray I dont get anymore problems.

Has anyone had any major features not work for them when rendering with satellite? Cause, I really dont want to find out later on, down the line, that an important feature will not work. I have tested the features xsisupport said "irradiance particles, assemblies, the ambient occlusion cache, and some final gathering optimizations" and all seems to be working fine. Maybe along the lines of stereoscopic passes?

Any information would be great! thanks!

[Mathaeus]

If you are rendering sequences, I'm afraid you should avoid satellite rendering at all, with any kind of BSP, any MentalRay, not just the one with XSI.

For defaults of old BSP, you should set the BSP depth and leaf size too, so that would be, just in this order:

set MI_FORCE_OLD_BSP=40 10

[Firebird]

I read the remark I guess you are refering to, by bronco67 on the XSIBase, differently:
BSP2 is slower and crashes a lot...
(I can't confirm or deny this,
but reading the differnt forums there seem to be a lot of people
disliking the whole BSP2-acceleration a lot)

I would take a step away of such comments because its just not 100% true it always depends on what scene you are working on (heavy geo, heavy shaders).

Thats why i always test my scenes with both bsp´s and on most of my work with is heavy geo etc. bsp2 is in ~60 % of the scenes faster then bsp 1 only when i render simple masks or constant black white main passes i use bsp 1(was faster in my tests).

But i can imagine if you render more "simple stuff" like a lot of passes with just a few elements in every one then bsp2 can really be slower.

Also displacement was most of the time faster with bsp2....10 sec here 5 sec there then add it up to 2000 frames . there you go, time saved.

So i would say test what works best and use that.

Sure this just comes to play when you are not want to or are able to use satellite rendering.
And hope they will integrate a switch into the UI.
I cant really say much about more crashes didn't really experienced that..... had crashes with both of them

[Hirazi Blue]

Okay my comment probably should have read:

I read the remark I guess you are refering to, by bronco67 on the XSIBase, differently:
"BSP2 is slower and crashes a lot..."
(I can't confirm or deny this,
but reading the different forums (especially XSIBase!) there seem to be quite a few people
disliking the whole BSP2-acceleration)

By which I still do not claim to agree with anyone, I merely try to distill other people's comments on this... ;;)