DR without spawer

[QuakeMarine1]

is it possible to start DR via comandline without using spawner ?

I as a professioal renderfarm have many admin to get a DR to start
I must disable the nodes from farm manger to avoid getting other job while DR is running
I must batch the nodes to start the version of Vray I need in the correct bit version and at the end kill the spawner + re-enable for farm manger (I am not using backburner for DR)
furthermore I block a workstation as DR master

I could setup some nodes doing DR only but then I miss that power for other jobs

I whould love to start a DR in commandline as the normal render comand for 3dsmax do to get around all the issues
do Chaos know about and is in SP3 something different for DR?

onced the GI is save to file or brutforece I could run a script to region render the image in comandline
(or split scanline)
but most stills using irridiance map + LC
I can not render the irr/LC to file first using one PC - thats taking too long and client not saving time

any thought about it?

andre

[vlado]

There are no significant changes to the DR in sp3 (except the support for incremental irradiance maps).

Do you want to start the slaves from the command line, or to initiate a DR rendering? Both of these, you can do from the command line.

Best regards,
Vlado

[dbuchhofer]

you're using backburner still for the normal farm use?

I'm assuming you mostly have batchfiles for each of these things? I dont really have a good solution to the different versions thing, unless you maybe script a version switcher, which wouldn't be too hard

Have you looked at the cmdjob in the max regular help file? it lets you send batches to backburner to run locally..

but you can also use it to "block" out the machines you're using for DR, so they wont take on regular backburner jobs

from in the backburner directory:

cmdjob -jobname DaveIsUsingDR -manager Managername -priority 25 -servers render7;render5 cmd.exe

will set them as "running" a job in backburner, you'd really just need a callback script that checked which machines you have checked in your vray_dr.cfg file, and supplied that to the -servers list, with a post render callback to delete the job.

that could fix the having to kill and restart server.exe part of the problem at least.

I haven't gotten into the hassle of having to support multiple max/vray versions yet thankfully!

[QuakeMarine1]

... (I am not using backburner for DR)...

you have overseen

[vlado]

You can start a DR rendering using the 3dsmaxcmd.exe command; you can start and stop render slaves for DR in a number of ways:

(*) register the spawner as a service, and then start/stop that service from the command line;
(*) directly run/kill the V-Ray spawner executable;
(*) directly run 3ds Max in slave mode for DR rendering.

That should pretty much allow you to do anything you need.

Best regards,
Vlado

[QuakeMarine1]

Do you want to start the slaves from the command line, or to initiate a DR rendering?

no and no

I like to do a DR without master and spawner

a common DR is even using backburner a very complicate thing
see the problem section in the forum - its full of posts about DR

a bulletproof way I can image to do DR could be using "backburner split scanline"
onced job is submittet to BB we need 4 steps to do
first pre-render GI / LC and each "split" save the GI to disk
second merge the "split" GI to one file
3rd render in split mode with GI from file
last just stitch each stripe together

no more spawner wich could fail / die / loose connection / timeout / freez / hung / explode / melt ...
BB cares about error detection during render and restart the task (and all the other benefit BB have .. )
easy restarting strip X if node Y miss a texture


I am sure many user have less stress and error in DR then
furthermore render farms can implement that way of render into the manger software

the way you mention could be working but is risky regarding failur tolerance

[vlado]

Well, best to use stripe rendering in that case then. DR simply works in a different way - it requires more tight communication between the render slaves, and it needs a client machine to collect and redistribute the result.

Further on, nothing really prevents you from implementing the workflow that you describe on your own - you have all the tools you need.

Best regards,
Vlado