blog

On Wednesday I gave a presentation to members of the campus community about my independent study on the renderfarm.  I was grateful to have individuals from Tech Services, the DMA Department, the Computer Science Department, and other areas in attendance.  In the presentation, I covered the basics of rendering computer graphics and how these principles were expanded into our renderfarm.  I then proceeded to cover the different tasks I completed as part of the independent study.  I finished up with a very productive question and answer session, with many helpful questions being raised.

I would like to thank Bob Landon for his assistance in completing this study, as well as Dr. Lehman for helping me to pursue this opportunity.  I look forward to continuing work on the render farm over J-Term and even into next semester.

Thanks for reading!
Andrew Martin

1 hour

53.5 hours total for the project

After about fifteen weeks working on the render farm, I believe that I am about done with this phase of the project.  Looking back at the original proposal for the project, I noticed how much has changed since the beginning of the semester.  From MentalQueue to sleeping render clients, it has been quite an adventure.  I also noticed how much has actually been accomplished.  First of all, I was able to get MentalRay to work with RenderPal, producing a somewhat automatically managed rendering environment.  With much help from Bob, we were able to configure network drives to connect to the shares required for the shaders and project files.  I also began development of a script to install MentalRay and RenderPal to client machines, along with setting environment variables and starting services.  I ran several of my projects through the render farm, including one that spanned several days.  I also taught a friend, who may be doing the render wrangling next semester, how to run the basics of the render farm.  Currently, I am sending several projects from the advanced 3D class through the render farm.

On Friday I spent most of the time on the render farm finishing up documentation, waking computers, troubleshooting why the MentalRay renderers were not working, and starting renders.  All last week and probably well into next week I will be working on setting up student renders and troubleshooting errors.  It seems that I have acquired the temporary role of render wrangler for the semester.

I will be giving my final presentation on Wednesday at 2:00 in Science Hall room 122.  Come then to see and hear more about this project!

Andrew

6 hours

Right before Thanksgiving break I was asked to present to the advanced 3D animation classes on how to properly format projects for the renderfarm.  I spoke for around half an hour for each class about the various steps required to optimize the Maya scenes, as well as how to export projects.  I had the students follow along so that they could see how to prepare their own projects for the renderfarm.  I think that the presentations went fairly well, and I was actually able to address some issues that I had not foreseen by following along with the twenty plus students through the process.

Unfortunately, I did not blog from two weeks ago (which was actually three weeks ago because of our Thanksgiving break).  However, I did not accomplish much that on Friday because of a server failure at Tech Services.  It was not that big of a deal, except for the fact that the renderfarm’s shaders were stored on the server that was the replacement server.  So I spent most of my time on the renderfarm that week cleaning up my scripting, organizing the render console, and remapping the render shaders folder.  i basically found that in order to remap the shaders, we just need to update the links in the rayrc file and put it on the machines that will be acting as render clients.  I have also found a tool that allows for remote command-line execution (using authentication) that could be helpful for pushing these somewhat small updates out.  The tool is not quite working yet, but I have been talking to Bob about it.

Professor Ballinger from the DMA department has asked me to present to his two advanced 3D Animation classes about how to prepare projects for the renderfarm.  In order to do that, I have been talking to him and Bob about the procedures surrounding the process.  Basically, we are discussing the specifics of how the render wrangler is notified of a job that a students wants rendered and the procedure of “waiting in line” for projects that are submitted as higher priority.  I think that this will be a procedure that really develops itself over the tenure of the first official render wrangler next semester.

Coming up next week… documentation!

Andrew

2 hours

I had a lot of fun this week scripting an installation batch script for the renderfarm.  Basically, Bob helped me to break down each of the tasks required to install the MentalRay client, set up the RenderPal client, configure environment variables, and other related settings.  I decided to use DOS batch scripting, primarily since that is what I know and am comfortable with programming in right now.  I got the main tasks outlined, but I have yet to finish the last several settings and actually test the script.

This week, I believe that I am going to look at user restrictions, account setup, and time/resource management.  Also, depending on some initial troubles I have had with rendering projects, I may have to troubleshoot some of the sleeping renderfarm issues, or possibly some of the render client output issues.

Andrew

3 hours

After spending around thirteen hours last week working on the render farm, we have successfully managed to send a sizeable job through the pipeline.  Now, I just need to establish what the bounds of that pipeline are!  Regardless, I was able to configure the render farm last Friday in a way that would actually render frames on around thirty machines.  The only problem now is that the machines are falling to sleep and not waking up when commanded…

I wrote a script last week that copies the RenderPal directory from the server and initializes the service required to run the RenderPal client.  The script also updates computers that already have RenderPal installed, setting network shares up and other such requirements.  My main nemesis continues to be DeepFreeze, since whenever I think I make changes to the machines, the changes revert.  I think that the problem is that DeepFreeze has to be left off for an additional reboot cycle once I make changes, which does not really make sense.  Also, I need to be sure that any Windows updates install while I have DeepFreeze off so that they will not be stuck in a reboot cycle after DeepFreeze is turned back on.

Tomorrow I am planning to work on waking the render farm remotely!

Andrew

13 hrs.

Do I say it?  Will it just break down once I mention it?  Anyhow, I am extremely excited and grateful to say, the render farm is working!  Bob and I got it to a place today where I can render out complete scenes.  This is exciting!  I have already rendered out a homework assignment in HD in around an hour that would have taken much, much longer had I done it on a single machine.

This does not mean that all our problems are gone, however.  I currently have the RenderPal client configured properly on eight out of thirty machines in one of the labs.  This will not be hard to fix, however, since I already know what needs to be done.  Also, output is still not going to the intended location, but I at least know where it is going and can relocate files as needed.  The renderers are still extremely picky about textures, so I imagine that there will be issues once students actually start rendering and realize that there file mappings are sometimes mixed up.  That should not be too difficult to work out though.

Next week, I hope to reconfigure the machines, or else at least write some scripts to help me with the configurations!

Andrew

3.5 hours

Ok, so perhaps mass deployment is not an accurate description of what I actually did last week, but it felt like it!  For the first hour, I went to the Loew-Brenn classroom and edited several settings on about ten machines, necessary to get things to render properly.  I have known about those changes for some time, but I just got around to editing them at this point.  The main changes that I made included turning off Deep Freeze (so that the changes would be saved), supplying credentials for the render shares, specifying another network drive for shaders, and ensuring that the render service would restart if it were accidentally killed.  Also, I had to update the credentials for a couple of services that did not seem to be starting properly.

After that, I took an extremely simple animated scene that I had created and tried to render it, but I still ran into the issues of specifying the proper output directories.  It seems that there are still some major permissions issues, but I still have not figured them out completely.

I am hoping that maybe we can get some time this week to just sit down and really examine where we need to go next.  I still have around twenty computers to reconfigure in Loew-Brenn, as well as obtaining proper output for a single animation.

Andrew

2 hours

I re-exported the Maya files this week, using the simpler project files instead of the shader-heavy versions.  This helped immensely, but there were still a couple of errors that were produced that did not really effect the output.  I was receiving output, but still with the frame numbering issue.  By the end of the time I spent on the project, I thought I was going to get numbered frames, but there were other issues that prevented the renderer from writing files out.  Frustrating!

I think my next plan of action is to create an extremely simple scene, export it, and then get it to render.  This should be fairly straightforward with what I already know.  Then, I plan to try to get the numbered frames to work.  After that, if I have any time left (or the next week) I will work on ironing out the shader links.  I have several students who are waiting to put some work through the render farm whenever it is operational, so there is a little more pressure on getting this operational (not that there was no pressure before…).

Thanks for reading!

Andrew

2 hours