Matthew Walker wrote: > Secondly, I have a server that's having major issues with I/O to the root > partition. > Twice in the last 12 hours, it's switched the root FS to read-only, and then > kernel > panicked. The errors that seem to precede the problem are journal I/O errors. > The > partition is on a raid 5 array that is in good health, and I've used smartctl > to check > the health of the underlying drives, which all seem to be fine. My next > suspect is the > 3ware card, and I'd like to flash the firmware on it, but I can't find any > documentation > about whether it's safe to do that on a running configuration. Does anyone > have any > further insight to shed on this, either about how to troubleshoot the 3ware, > or other > items I should check to isolate the problem more accurately?
This could indicate bad RAM. Make sure you're using ECC RAM. It could also indicate noisy power. You could look at it with an oscilloscope, or you could just buy a high quality PSU and see if it makes a difference. If it doesn't, then at least you'll have a nice PSU on hand for emergencies. I'd like to point out that with software RAID, you can plug the drives into any Linux box and use them without trouble. It's much more complex and risky to do that with hardware RAID. That pluggability is extremely valuable in emergencies, so I personally see hardware RAID as much less reliable than software RAID. Shane /* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */
