Question : Problem: 3ware 9550 RAID 5 randomly dropping disks

I am having consistent but random RAID errors from my 3ware 9550SXU-4LP.

A little background: I was given 3 old appliances that had these cards with 2 drives in a mirror. They are dual Xeon 2.8 machines so I decided to upgrade one to RAID 5 to make a file server. I replaced the mirror with 4 Seagate 640GB ATA drives. I began receiving errors at random times that one of the disks had become disconnected (but not always the same drive). After going through the cabling with no effect I switched to a completely different appliance, only to have the same problem.

Thinking the appliance's PSU might be too weak for 4 drives, I built a new system around the 9550 with a P4 motherboard and 620 W PSU. I ran SMART tests on all the drives, and I am confident they are good. I used all new cables.

I still get the same problem, it's as if one of the drives gets bumped or unplugged.

Last night, after encountering the error and rebuilding the RAID for the umpteenth time (it always rebuilds just fine), I broke down and swapped in the 3rd 3Ware controller. At 3 AM, the RAID suddenly decided to lose a hard drive again.

I have tried new power supplies, a different motherboard, new cables, I have tested the drives using SEATOOLS and SMART utilities, and I still have the same problem with ALL 3 RAID cards.

Either the 9550 sucks, there is a stupid reliability setting I am missing (I've turned off NCQ and write caching), 3Ware and Seagate products don't work together, or I'm cursed.

These are supposed to be good cards, but I have wasted dozens of hours and not a little bit of money trying to get a simple RAID 5 to work reliably.

Curiously, all 3 cards seem to work just fine with the original 80 GB Hitachi's in a mirror.

The Seagates are all set for SATA II. These "free" RAID cards have been a very expensive gift. I really wish I had never heard of them. However, I would rather not drop $2K on a new server.

Answer : Problem: 3ware 9550 RAID 5 randomly dropping disks

It's working now.