This weekend, I decided to spend some time, and money, to aggregate all my hard drives using LVM, and create a massive storage array. It worked, kinda.
I put 2x120GB, 1x80GB, and 1x400GB drives into atlantis, my old windows box, that happens to be a dual processor system running Linux, now. LVM makes it pretty easy to add drives to an existing logical volume, so I copied stuff over the network, and built a 720GB virtual drive, piece by piece. After filesystem overhead [ext3] and rounding, the total size came out to around 661GB. Of course, the fun didn't stop there.
Apparently the IDE controller I was using, a Promise Ultra100, is either buggy itself or has a buggy driver. Silly things like this kept happening when I put the system under load:
Dec 5 23:55:16 atlantis kernel: hdg: task_out_intr: status=0x58 { DriveReady SeekComplete DataRequest }
Dec 5 23:55:16 atlantis kernel:
Dec 5 23:55:16 atlantis kernel: ide: failed opcode was: unknown
Dec 5 23:55:16 atlantis kernel: hdg: status timeout: status=0xd0 { Busy }
Dec 5 23:55:16 atlantis kernel:
Dec 5 23:55:16 atlantis kernel: ide: failed opcode was: unknown
Dec 5 23:55:16 atlantis kernel: PDC202XX: Secondary channel reset.
Dec 5 23:55:16 atlantis kernel: PDC202XX: Primary channel reset.
Dec 5 23:55:16 atlantis kernel: hdg: no DRQ after issuing MULTWRITE_EXT
Dec 5 23:55:16 atlantis kernel: ide3: reset: master: error (0x00?)
Dec 5 22:44:11 atlantis kernel: hdf: lost interrupt
Dec 5 22:44:41 atlantis kernel: hdf: lost interrupt
I ordered a newer Promise IDE controller, an Ultra133 TX2 card, which uses the pdc202xx_new driver, instead of the pdc202xx_old driver. Hopefully this will fix the issue.
Tomorrow I'll be heading to the Sheraton Parsippany Hotel in, well, Parsippany, to attend a Juniper Secure Access Training session. I've heard it's a pretty nice place, and should be a good experience. I hope it doesn't rain, too much.
