Do you use DR?

[RSI]

We just began to use it

We use it for stills, and it's a bit painful, at the begining, it was cool (6 quad core), but after 3 weeks of work with it, we can only render on 2 nodes, so we lost contact with other nodes one after the other. Other problem is with the vraymesh, even if they are found on the Z drive, because of too many traffic on network( we have big scenes), they loose them if there are many (gigalan network no server) so we had to copy them everywhere, we also have to relaunch the render sometime, because some nodes don't refresh the textures, or don't see them every time. It have been a big problem for us , and we have lost a lot of time installing and reinstalling softwares, caming at the office in the morning seing the render not finished because only one node work... So we probably gonna format everything when we will have time and try it with a real server (not a map drive) and a fresh new installation.
Even if i think that Chaosgroup is making an amazing work, i regret the lack of tutorials, explanations about DR I don't understand why I have to make at least one backburner job before i can use dr, service are complicated too (administrator rights..., relaunch, unability to render only on servers node to keep working on the manager workstation, unc...)why is it so complicated, why there is no dedicated topics with best dr worflow installation...could it have a option for sending automatically all textures and mesh on each computer for rendering to save our ass at least for the final render, why does the last cluster is always so long...

Sorry i'm just a little deceptioneted about that beecause we bought new workstations for that and for a couple of important clients (we are trying to be more professional) we have and we've been late for one of them and it's the first time.

Best
"Sorry for my english"
Gael
www.rsi-studio.com

[Sascha Selent]

Essential but not stable

Hi.
I use DR almost every time, but I recognized problems.

The slaves do quite often fail to finish the render. Sometimes the render doesn´t finish at all, so I have to kill every slave one by one and restart them one by one. That´s a little painful.

It would be very useful if I could kill all slaves at once from within max and restart them all at once the same way - compare to finalRender, this is a quite good implementation of DR.

Okay, best would be if the slaves do not fail at all

Ahm, I don´t know if this would be possible (I think not), but I ask anyway - it would be great if the rendering of the lightcache could be shared between the slaves. For now every slave has to render it´s own lightcache and after that the rendering of the irradiance-map is shared.

Correct me if I´m wrong, but as far as I understand the lightcache is split into passes, every pass is it´s own thread and after finishing the passes they are merged together.

When I set passes to one, only one core is used. So maybe it could be possible to distribute the passes to the slaves, after every slave finished the pass it sends it back to the master and the master merges them together, distributing the merged result back to the slaves...

Pardon me if I talk naive nonsense, I´m a dumb user and I do not understand the internals. I just let out what is in my stupid little brain.

But nevertheless: DR rocks!

Sascha

[kimblarsen]

We used to use distributed rendering for most all our still work and animation before we aquired our three new Mac Pro dual quad core workstations; the three of them together has almost half of the capacity of the render rack. However I suspect we will continue to use DR for most of our projects even with these powerful workstations.

From time to time we have had problems with the render nodes not finding textures and HDRs, a problem which we have always been able to solve by copying all the assets into the SAME folder as the Max files.

The issue with large network traffic and render servers being unable to find assets (textures, Xrefs) I believe is not unique to Vray since I have discussed this with colleagues having similar problems using other rendering engines such as Brazil.

I would be interested in hearing from the dev team what their thoughts are on these issues.

Cheers,
Kim

[andrewjohn81]

we'd use it here, but there are 4 of us sharing 15 computers. And when DR crashes it tends to crash the entire machine, or at least lock up any max instances. It's just not been worth it. If we really need something like that we find it's faster, right now, to just send as strips.

Having to precalc GI puts a pretty big damper on things also. We test it occasionally, but I had to mark it as a never since we don't use it on any actual projects. It just isn't reliable enough for us yet.

great idea, just haven't been able to get it set up how we would want it. Some of that could be on our part not setting it up properly, but IMO that should be fairly simple as well. We have a lot of IT restrictions as well so some of the issues may be permissions related. I don't know much about that.

[Clifton Santiago]

@sasha: You can batch stop and start the spawner on your DR nodes with psservice.

http://technet.microsoft.com/en-us/s.../bb897542.aspx

I still don't understand why everyone's nodes keep failing on DR, mine almost never fail. I'm thinking the problem is not with the DR...

And with the integration now of DR and BB, its even more brilliant.

[ene.xis]

totally agree!

though i have to say i it took some time spent with the it guy to make work flawlessly...

[3LP]

And with the integration now of DR and BB, its even more brilliant.

What do you mean ?
what's the link between DR and BB ?

[ene.xis]

meanig u can submit dr jobs to the bb!

[Sascha Selent]

meanig u can submit dr jobs to the bb!

Is this new in the beta of SP3? Does not work with my SP2-version on max2008- I tried it but I´ll try it again.

[Clifton Santiago]

Its been functioning since SP1. If you were to search, you might find this helpful thread: http://www.chaosgroup.com/forums/vbu...ght=backburner

[Sascha Selent]

@sasha: You can batch stop and start the spawner on your DR nodes with psservice.

Thank you for the tip. I didn´t bring it to work yet - maybe it´s a problem with my two network-adapters - seems that psservice uses my 192.168.0.30-adapter rather than the 192.168.1.39-adapter where the slaves are to find... But I´ll try further.

I still don't understand why everyone's nodes keep failing on DR, mine almost never fail. I'm thinking the problem is not with the DR...

Hm, I do not use Spawner as a service. I manually start the spawner on every machine. Maybe this can make problems?

[Sascha Selent]

@sasha: You can batch stop and start the spawner on your DR nodes with psservice.

http://technet.microsoft.com/en-us/s.../bb897542.aspx

I still don't understand why everyone's nodes keep failing on DR, mine almost never fail. I'm thinking the problem is not with the DR...

And with the integration now of DR and BB, its even more brilliant.

Hi Clifton.
I do not bring it to run. May you provide a commandline U use? I thought it would be like this...

psservice \\computer -u user -p password stop VRaySpawner 90

...isn´t it? I´m no commandline-pro.

Locally it works, but not on the LAN...

...it seems that it doesn´t try to connect anywhere - the Firewall does not react at all...

[Clifton Santiago]

http://www.chaosgroup.com/forums/vbu...144#post372144

[Sascha Selent]

http://www.chaosgroup.com/forums/vbu...144#post372144

Thanks man, I found out why it did not work. I tried it on a XP Home-machine. XP-Pro is needed.

But nevertheless now I do not get VRaySpawner to work via service. As service VRaySpawner is up, max is up, it starts to load the scene but then hangs up with 50% of the dual-core in usage. Grmpf. I also changed the user for the service to be sure it can access the network-paths, but this had no effect...

Do you have any suggestions?

Thanks man.

[Sascha Selent]

But nevertheless now I do not get VRaySpawner to work via service.

Found out this too. After installing the service the machines needed a restart...