We've been having trouble lately with mantra jobs that were reading bgeo files from disk over NFS during the render, over the grid farm. Essentially, when traffic got heavy(a dozen or so systems, most duals, all burning on frames), it appeared that the geo load would be incomplete or unsuccessful, and mantra would render corrupt frames. Changing this to “Use Geometry as is”, and the problem appears to have gone away. It seemed that having the data buried inside the IFD, while not as efficient for hscript generation, makes for more reliable renders than having to page out in the middle of things. So far, no failures.

Now, I know there's a million variables here, but I'm just wondering: has anyone else had troubles like this with v8?

We're all Linux, SUSE 9.x mostly(except for the central server, which is SUSE Enterprise 9 and quite beefed up). Giggie network switcher, although some of the machines are 100BT.

Thoughts, experiences appreciated.

Cheers,

J.C.