Dave Buckley Posted February 10, 2009 Share Posted February 10, 2009 Ok this is a constantly on/off subject for me as it keeps getting put on hold (with it not being vital). Where I work we have two training rooms, each equipped with 8 computers, which for the most of the time, sit turned off (good for the planet, bad for my rendering speeds) The boss has said I can use them for rendering as we are now starting to take on production work (small stuff but substantial enough to need a render farm) My question is this: I understand how the render farm works at a very base level. 3DS Max is on my workstation, and the training suites will be used as the processing power. They all have Max installed on them. I believe they should all have the server application running on them. I am just unsure as to which machines should have the Monitor and/or Manager applications running on them. It would make sense if the Monitor/Manager applications were on the same machine as the application (my workstation) as this is the one i will be sat at in order to 'monitor' the progress of jobs while i work. However i don't want my workstation to do any rendering. So is it advisable to run both Manager and Monitor applications on my workstation??? Link to comment Share on other sites More sharing options...
mattclinch Posted February 10, 2009 Share Posted February 10, 2009 not really. by all means run monitor on your workstation, as it really is only for viewing and editing the job queue. but i would recommend you run the manager on a dedicated machine that is reliable and not likely to crash/go down/spontaneously combust with you slowing it down with max/photoshop/browsing the web. it should be clean and stable with a decent sized hard drive - if the manager machine crashes you may lose your job queue and have to resubmit all jobs. Link to comment Share on other sites More sharing options...
Devin Johnston Posted February 10, 2009 Share Posted February 10, 2009 If your workstation isn't going to be rendering then it's perfectly fine to run both the manager and monitoring applications from it. The monitoring app doesn't require much ram or processing power and for the most part neither does the manager except when it's handing out rendering assignments to the render nodes. You will need to have all of your maps and models located on a server that all the nodes has full access to and you will need a network drive to write all of the image files to with full write access. I wouldn't recommend using your workstation as the location for all of this, your machine will become bogged down and you won't get much work done. Link to comment Share on other sites More sharing options...
Dave Buckley Posted February 10, 2009 Author Share Posted February 10, 2009 ok so here's what i think i'm going to go for, my workstation as the monitor and also the place where i construct my scenes/create maps etc, local copies of maps and scenes stored on here too, then i am going to have 15 machines downstairs with the server application running, and then the last machine of the 16 in total downstairs as the manager. with regards to maps etc, if i am using a scene file that is local to my workstation and the same with maps, can i not just transfer everything to a networked location at rendertime or do i always have to be working from that networked location??? and also, (sorry if i am jumping ahead, currently working through the install guide for backburner) but can i get server/monitor/manager to respectively load automatically as the machines are turned on, so that all i need to do is turn the machines on?? i don't want to have to go round each individual one and fire up the relevant monitor/manager/server application Link to comment Share on other sites More sharing options...
mattclinch Posted February 10, 2009 Share Posted February 10, 2009 and also, (sorry if i am jumping ahead, currently working through the install guide for backburner) but can i get server/monitor/manager to respectively load automatically as the machines are turned on, so that all i need to do is turn the machines on?? i don't want to have to go round each individual one and fire up the relevant monitor/manager/server application yes, check the help file for running these as a service - bear in mind that for some reason running it as a service means it wont read mapped network drives, known bug i think. Link to comment Share on other sites More sharing options...
BrianKitts Posted February 10, 2009 Share Posted February 10, 2009 I run manager on my everyday machine, it manages running tasks all day and night when I turn on server on my machine as well. Typically I'd love to have a dedicated barebones $200 machine to run manager on but since I don't it runs on mine. Either way I'm not going to sacrafice a good machine at night to run manager as I've never had any problem with running it on a render server machine. It's better having one more render node. Link to comment Share on other sites More sharing options...
Dave Buckley Posted February 10, 2009 Author Share Posted February 10, 2009 cool, thanks for the heads up, one to discuss with the IT guys i think, i'm not too sure about the whole mapped network drives. not too sure if this is going to be an issue for me yet?? still figuring out how it works properly Link to comment Share on other sites More sharing options...
Dave Buckley Posted February 10, 2009 Author Share Posted February 10, 2009 also my workstation is 64 bit MAX, and all the render nodes/servers are going to be 32bit MAX, i have a feeling this is going to be a problem?? Link to comment Share on other sites More sharing options...
Scott Dombrowski Posted February 10, 2009 Share Posted February 10, 2009 (edited) yes, check the help file for running these as a service - bear in mind that for some reason running it as a service means it wont read mapped network drives, known bug i think. I have my server apps running as a service and to find the maps, you are right, they cannot be referenced to mapped drives (i.e. N:\network drive\maps). They have to be referenced in the UNC standard (\\network server\network drive\maps). I've learned to just do that by default when I'm applying textures. I think there's a way to remap your assets using the Asset Tracker... right click on the texture, then choose "resolve path to UNC location" daveb867, if you can, put all of your texture assets on the network somewhere and reference that location when creating your materials. Any new textures you make will then be easily usable for the next project, and all textures will be ready to go for network rendering. Edited February 10, 2009 by nauticus27 Link to comment Share on other sites More sharing options...
Scott Dombrowski Posted February 10, 2009 Share Posted February 10, 2009 also my workstation is 64 bit MAX, and all the render nodes/servers are going to be 32bit MAX, i have a feeling this is going to be a problem?? This will not be a problem when submitting jobs, as long as you don't add your 64bit system to the pool of server nodes. When you submit a job, there's an option to submit it as 32bit or 64bit. Link to comment Share on other sites More sharing options...
Dave Buckley Posted February 10, 2009 Author Share Posted February 10, 2009 daveb867, if you can, put all of your texture assets on the network somewhere and reference that location when creating your materials. Any new textures you make will then be easily usable for the next project, and all textures will be ready to go for network rendering. so basically just work from a texture library that is on a network to save complication??? this is fine, as i currently work from an external hard drive, but can just copy all texture libraries to a network location and then start working from there Link to comment Share on other sites More sharing options...
Scott Dombrowski Posted February 10, 2009 Share Posted February 10, 2009 so basically just work from a texture library that is on a network to save complication??? this is fine, as i currently work from an external hard drive, but can just copy all texture libraries to a network location and then start working from there That's what works for me. Others may have a different workflow, and I'd be interested to hear what else works, but that may be taking this thread away from setting up Backburner. Link to comment Share on other sites More sharing options...
Dave Buckley Posted February 10, 2009 Author Share Posted February 10, 2009 i am also interested to hear other people's setups, could i not continue to work from my external hard drive, then just zip up all the assets and place them on the network so that all the nodes have access to any assets used in the scene Link to comment Share on other sites More sharing options...
mattclinch Posted February 10, 2009 Share Posted February 10, 2009 I have my server apps running as a service and to find the maps, you are right, they cannot be referenced to mapped drives (i.e. N:\network drive\maps). They have to be referenced in the UNC standard (\\network server\network drive\maps). I've learned to just do that by default when I'm applying textures. I think there's a way to remap your assets using the Asset Tracker... right click on the texture, then choose "resolve path to UNC location" we got around this by running the appropriate server or manager application as a scheduled task at logon rather than a service, as we like our mapped network drives. something to consider. Link to comment Share on other sites More sharing options...
Crazy Homeless Guy Posted February 10, 2009 Share Posted February 10, 2009 (edited) Working from a server with UNC or DFS pathing will save you time, and is the absolute proper way to be set up. You are guaranteed to run into problems on a regular basis if you are not working this way. Also, I do not like running servers in service mode. At least not until you are a 100% sure that everything is working smoothly. Troubleshooting a renderfarm where the servers are in server mode is a bitch. Also, I am not sure if you can see the image rendering while it is in service mode. I constantly check the progress of my renderings while they are processing in backburner. To do this you need to be able to see them on the screen. If the training room is far from your desk, you might want to setup remote desktop connections to the machines. This way you can log into them, and check the progress without getting up from your desk. Since I don't like to run as a service, I simply place a shortcut to the Server application in the Windows startup directory. This way it will auto launch when Windows starts. Another thing to think about... Your farm machines are going to need copies of the plugins that you are using placed on them. There are a couple of ways to go about this. You can create a server directory for plugins, or you can copy the plugins to each machine. If you are using plugins on a server, then you need to repath each Max copies plugin path back to the server. Which ever way you go, you should use scripted Windows batch files to speed this along, and to make things run smoothly. In addition to the note above, if you are rendering with Mental ray, you should path all of your shaders back to the network also. This will save time, and allow all of the machines to see the shaders. Edited February 10, 2009 by Crazy Homeless Guy Link to comment Share on other sites More sharing options...
Scott Dombrowski Posted February 10, 2009 Share Posted February 10, 2009 i am also interested to hear other people's setups, could i not continue to work from my external hard drive, then just zip up all the assets and place them on the network so that all the nodes have access to any assets used in the scene I suppose that might work, but you might run into the issue with the paths not being referenced as UNC paths. It wouldn't hurt to texture up a teapot, archive it, and see what happens. we got around this by running the appropriate server or manager application as a scheduled task at logon rather than a service, as we like our mapped network drives. something to consider. Not a bad idea. I'm in the process of getting IT to install Max and Backburner as part of the standard package when they set up new machines. We have over 300 people in the building, and I start to drool a little when I think about being able to harvest all those CPU cycles at night. I'll suggest that to them and see what they say. Link to comment Share on other sites More sharing options...
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now