DISCLAIMER: This is a continuing series detailing the painful story of a DIY render farm build. It is terribly technics and
somewhat frustrating. Those who are unprepared for such “entertainment” are advised to ignore these posts.
Today is when I worked out all the details. Apparently scons was doing okay. Further investigation showed that Dr. Queue had built in the src folder, i.e., right next to the source pieces. They were supposed to build in the bin folder. Way back at the very beginning I had written the correct paths to install as one of the preliminary steps to compiling. Since then I’ve burned everything down fifty-seven times. Since I never went back and added those paths the compile made the programs right next to the sources.
It’s OK that I did not, because then Dr. Queue would have been building in usr/local this whole time, which is not really necessary on a Mac. The Dr. Queue instructions insisted on that path because it’s good for Unix. Who knows what you would do on Windows? The answer is usually that you throw Windows out and put Ubuntu or something on that machine so you can get something done…
So I took the binaries out and placed them in the drqueue/bin folder. This revealed more config problems. So I then
- Fixed the paths in .conf files so they point to the proper directories (/drqueue/tmp and /dequeue/logs)
- Fixed environment variables – etc.
- Removed duplicates! There were extra copies of the binaries lying around.
- And the master worked fine now! Once the file was dropped into the Terminal it started running, echoed the correct prompts to the terminal, showed up in the Activity Monitor, and seemed to be waiting for slaves to join!
- I was queued…!
And apparently queued properly, unlike many of the other tests. But if only I could get that GUI going… Just as I lost hope, I found the issue! I needed to have upgraded X11 to X-Quartz. The out-of-date X11 distribution might have been what was causing all the fuss!
Having installed X-Quartz, I started drqman…
drqman started up immediately.
Now I vowed to keep it running forever while I set up the first worker!
Of course I tried to just copy over the Dr. Queue I had just compiled to the Intel machine.
I guess it did not build fat binaries, as it would not boot up on the newer machine. Odd, I was sure I was using “drqman.Darwin.fat” before, and that would indicate a fat binary… but that was the not-so-good distribution I had downloaded…
Or was it?
I was overcome by a terrible sinking gut feeling at that moment. It may have been absolutely true that had I upgraded to XQuartz in the first place that the already compiled Dr. Queue may have worked out of the box, thus saving me all this time and energy. In fact, it was looking increasingly like this was the case.
Okay, there was no time to start whining. I needed to figure out how to install at least OSX 10.8 on the the first worker, “Gog,” and get the farm going.