Trouble rebuilding array on 3ware 7000 (twe)

[Matt Staroscik]

I need a hand from a disk guru as I am having some problems with a 7000-2 
mirror under 4.10.

My 3ware Escalade 7000-2 has a pair of Maxtor 160GBs on it. I run them 
mirrored. The other day, a drive bit the dust. As the disk was well and 
truly hosed according to the Maxtor utilities, I chucked it and dropped a 
replacement into the RAID. The replacement disk was the exact same Maxtor 
model. (these were all cold swaps by the way, no hot swap cages here yet.)

At boot I hit alt-3 to get into the 3ware BIOS and I started the array 
rebuilding. All seemed well. The 3dm web page indicated a rebuild was in 
progress. Disks were chugging.

After a few minutes, the rebuild stopped and the web page showed "error." 
The console showed unhelpful twe0 Disk Error and Rebuild Halted messages.

At this point I decided to play it really safe. I dropped a spare drive 
into a non-3ware IDE bay and proceeded to save off dumps of all the 
filesystems. All was well except for a few read errors on /usr. These were 
accompanied by a message about a twe controller failure with error 0x40.

I rebooted into single-user, ran fsck -fy and tried the /usr dump again. 
There were less errors, but there were still 2 bad blocks, with the same 
error code.

At this point, the system runs but I have a degraded mirror (with its own 
errors) that I cannot rebuild, and I am not sure why.

- Is it still the case that a 7000-series 3ware cannot successfully rebuild 
under FreeBSD? I found year-old mailing list traffic that mentioned this, 
and suggested a Knoppix Linux live CD as a workaround.

- Or, are the errors on the source drive killing the rebuild process?

- How can I fix the errors in /usr? I ran fsck -fy in single user mode, 
what else can I try?

Many thanks in advance for your ideas. With half a RAID I am running on 
borrowed time. At least a SMART probe of the remaining drive only shows 2 
errors!

Best,
Matt S.