Hard Lockups

[Ruler]

I've been having a problem for quite a while now. I've gone through the whole range of possible problems, from thinking it's a software bug to believing it to be hardware related. I've replaced basically all the hardware in two different machines with brand new and then used (known good) parts. For all intents and purposes, I've swapped the entire machine hardware-wise with another. I always come back to the same problem, even though I have other boxes running with this exact same configuration without any hiccups whatsoever.

I have two Promise UltraTX PCI hard drive controllers in my box to control 4, 400-gig drives in a software RAID to hold the data. I have a spectra-8 capture card to grab the video. Either one works fine - I have pushed 75 gig of data to the RAID with no problem and can record from the camera to the boot drive (connected to the motherboard) until the drive is full. However, when both are active at the same time (when I record video from the camera to the RAID), the machine locks up hard. This will happen sometimes in a matter of seconds and sometimes it takes a few hours before it freezes.

I've already replaced the hard drive controllers with new ones, rebuilt the RAID and filesystem from new, and changes which PCI slots each is plugged into. (This was in thinking that it was an IRQ/DMA problem. I even matched what slots they're plugged into to another machine that has 2 spectra cards with the 2 pci hard drive controllers in. Those drives are WD though, while these are Seagates.) I've run memtest for hours on end and checked the drives with Seagate's SeaTools CD - no problems of course.

I'm frankly at a loss as what to try next. The last thread I started on this topic, when I initially thought it was a ZM problem because it only happened when ZM was up, or more generically when video was being captured, can be seen at http://www.zoneminder.com/forums/viewtopic.php?t=5845 (As you can see, I've been going over this for quite a while; several months before I started that thread even!) The strangest thing to me is that all 3 machines ran for a good period of time before starting to have problems...