-
Cluster rendering redux
Hi all:
A few months ago I started experimenting with cluster rendering using Compressor and Qmaster, and I posted here mentioning how it seemed to work, but the networked renders took as long (or even longer) than single-machine renders.
Today I had time so I started experimenting again. The one thing that I hadn’t done the last time was to mount the drive with the source files in the other computers, so I did so, but it didn’t make any difference: the cluster gets created, and you can see the submitted job from any of the nodes using Batch Monitor, but the render times are worse than with a stand-alone machine, and the nodes seem to be idle the whole time: basically, it seems like the render is being done only in one machine (the one from which you submit the job). This happens both with a QuickCluster and with a managed cluster.
So my question is: when people talk about “mounting the drive with the source files”, what do they mean exactly? So far I’ve been mounting them in two different ways:
-Using the “Connect to server” option in the Finder, and typing “afp://(IP of the computer)”.
-Going to the “Network” icon in a Finder window, and selecting the server in question.I should mention that I have all my source files in a separate drive from the system drive, and what I’m mounting is the whole drive (not just the folder with the files).
Incidentally: whenever Qmaster is activated, I can see in the “Network” window of the Finder an icon with a name like “192.168.1.38/private/var/spool/qmaster/EBC0D…” (which is the default location of Qmaster’s shared cluster storage), but when I try to click on it to mount it, there’s an error message about how “the system could not find the original resource”, or something like that. I’ve tried to configure the shared cluster storage in a more accesible location, like my home directory, but when I try to mount it I get the same error.