4500 RAID failure

[mikecrom]

I have a 4500 with 2 expansion units. We had a power failure for 2 hours on Saturday. The system came up to the recovery console, and I chose to reboot. It rebooted, and could only be pinged -- no other access, then rebooted again, and came up. But once the system was up, it indicated RAID failure on md0 -- the 4 internal drives. The other drives were fine.
Looking at the drives from the web interface, the status is OK on all 4 drives. All 4 drives show the same type, 245G MAXTOR, but in the size column it shows 230.67GB for 2 of the drives, and 533.00 MB for the other 2 drives.
Logging on to the system via ssh, I see that hde and hdg are the drives in question (onlly showing hde):

md100: former device hde2 is unavailable, removing from array!
md101: former device hde5 is unavailable, removing from array!
hde1: bad access: block=2, count=2
end_request: I/O error, dev 21:01 (hde), sector 2
hde: unknown partition table
hde: unknown partition table
md: trying to remove hde2 from md100 ...
md: trying to hot-add hde2 to md100 ...
md: hde2 has zero size, marking faulty!
md: hde2 has zero size, marking faulty!
md: trying to remove hde5 from md101 ...
md: trying to hot-add hde5 to md101 ...
md: hde5 has zero size, marking faulty!
md: hde5 has zero size, marking faulty!
hde: unknown partition table

Running fdisk, I see that neither hde nor hdg have any partitioning info.
Running fdisk or sfdisk to partition the drives works -- it shows the same info (cyls, etc.) as hda & hdc, but then the partition map is wiped by something.

So it appears the drives are OK, and just need partitioned. I'm presuming they are the same partition map as hda hdc -- but don't know for sure.

Any ideas how to get these drives back in the RAID array?
Thanks. Mike

[blue68f100]

If you have lost 2 HD you may be sol. Run SpinRite on the failed HD's, it may beable to repair the broken table. This has worked for some. Then replace the Maxtor HD's with some good Interprise WD HD. I have seen many times where Maxtor fail in buckets. I don't trust them or use them.

You should have your units on UPS BU with auto shut down...... and restart. Just needs to have a UPS that has the network card. The latest OS supports UPB cable too.

[mikecrom]

Thanks. Actually the system is on UPS. It's in a server room with a huge UPS, and got hit with the rolling brown-out on Saturday. The UPS, however, only ran 12 minutes!

I'll give SpinRite a try. The drives seem fine, I just don't see why the partition map disappears even after I install new ones. Unfortunately, I'm not in the same locality as the 4500. Do you know if SpinRite can be booted from grub, or even a USB key? (I know, look in the docs...).

Also, Adaptec used to have the latest O/S for these for free with registration.
Now that Overland Storage has taken over -- and dropped the products -- I can't seem to find any upgades. Any ideas? I'm more interested in getting the drives working again at this point, though, so just an aside.

Thanks again.

[blue68f100]

The GOS is only free if you have a service contract.. Other wise you have to purchase it. Overland has released the old SnapOS for free downloads. But it is only for 10yr hardware that does not support Vist or OSX.

No generator at the site for backup power? Where I use to work we had huge UPS and diesel powered generators. When I left we only had 3 gen and another was in the budget for live spare. These Cat generators burned 120gal/hr, 12 cyl quad trubos locomotive engines. The startup seq had these on line and sync in less than 3 min. Power had to be restored for 5 min and stable before it would switch back to provider power. On your ups setup do you have it set to do an auto shutdown and restart. On restart I have mine set not to restart till the batteries have a min of 25% battery level. This normally takes about 15-20 min, but it eliminates failures when the power is bounced. If bounced normally it will die due to weak batteries. Remember the 4500 take 5-6 min to boot, so you need at least that amount before restarting.

Spinrite is setup to run from a CD or Floppy, both are bootable. It can be run from a usb drive if its bootable. This program is written in assembly language. You will need to be at the machine since you must power down and remove the HD's to be checked to a desktop pc. I know it's a pain, I recheck mine once a year. So I detect when a HD is starting to have problems.