MD raid1 rebuild events

[skyking]

I leave my home server hooked up to a KVM port, logged in. Every month or so, I'll see some drive errors over there in the console. It triggers an array rebuild I can find in daemon.log . I have not checked the disks yet but the infrequency of the problem leads me to believe the hardware will check out OK. I'm thinking disk controller problem?

 

[Nothinman]

What are the drive errors that get logged?

 

[skyking]

They were up on the console, I'll look around for the errors.

 

[Nothinman]

Depending on the distro they should also be in /var/log/kern.log or /var/log/messages.

 

[skyking]

ata2: Status: {DRDY ERR}
Err: {UNC}
Exception EMask 0x0 SAct 0x0 SErr 0x0 action 0x0
BDMA stat 0x24
CMD 25/[buttload of hex] tag 0 DMA524288 in

Lather rinse repeat. I guess I'll run a USB CD-rom diagnostic at it and get an advance RMA, looks like a funky block.

 

[Crusty]

I used to see similar errors with a pair of Samsung Spinpoint 1TB drives on an nforce 4 SATA chipset. Turns out that particular controller was incompatible with those drives, I moved them over to a different board and have had zero problems, so don't rule that out before you RMA a drive.

 

[Nothinman]

Yea, buggy firmware on either side can cause weird issues.

 

[skyking]

Yea, buggy firmware on either side can cause weird issues.

I hope that isn't the case, as this a really neat little fileserver on a via embedded chipset. That may be the whole problem. I know it is incompatible with the 3.0gb spec but these WD green drives have a jumper to set them at 1.5, and it is the reason I went with the WDs. I really don't want to start over or abandon the project as it only uses 28 watts with two drives running, and serves up anything I need.

 

[joetekubi]

One time (in 5 or 6 years), I started getting some disk errors on a RAID 1 software raid. I had to manually "break" the array, and ran fsck on each one individually. One physical disk had some bad sectors, and when I restarted the software raid, the other disk mirrored perfectly. Worked out well for me, but a bit nerve racking until I saw the magic "UU" via cat /proc/mdstat

HTH,
joe

 

[skyking]

I ran badblocks -sv /dev/sda and sda passed OK.
sdb is borking out with those errors, so I may really have a bad disk on my hands.

 

[Crusty]

Does smartctl show any errors for the drives?

 

[skyking]

I ran
smartctl -a /dev/sda
smartctl -a /dev/sdb

In both cases down in the middle I read

SMART Error Log Version: 1
No Errors Logged

I don't know if I have to run a specific test or if it is grabbing the standard logs from the drive. I'll read up a bit more about that.