2

Having massive issues with 4 brand new 3TB WD Red drives I just got. Trying to setup raidz2 using "native zfs" with all 4 on ubuntu 14.04.1, which works initially until I start trying to copy data over to the mounted volume.

I've run badblocks on all 4 drives and everything came back OK. I've run SMART extended tests on all 4 drives and all OK.

Thought maybe the idle3 timer might of been causing issues, as I did notice the drives going to sleep quite often, even while I was trying to copy data to them (wtf?) so I grabbed the wd5741 tool, which if I recall correctly didn't work, but idle3-tools successfully stopped the drives going into standby - ever.

to date, syslog is still report errors like:

ata7.00: exception Emask 0x10 SAct 0x0 SErr 0x1910000 action 0xe frozen
ata7: SError: { PHYRdyChg Dispar LinkSeq TrStaTrns }
ata7.00: failed command: READ DMA

end_request: I/O error, dev sdc, sector 377981880

ata5.00: device reported invalid CHS sector 0

ata6.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
ata6.00: failed command: WRITE DMA EXT

for all 4 volumes, ata5-8 and /dev/sdc-f around when the processes accessing those drives start to lock up.

Once these drives lock up, any processes touching them lock up, even just trying to list directory contents (that haven't been listed recently) of the volume will freeze.

Only 2 drives have ever reported the "WRITE DMA EXT" error, but I suspect that is from when I've had to forcefully reset the machine because the devices were "busy" (doing something I couldn't detect nor stop..)

Other threads I've checked, most people are replacing the drives to solve the problem, but they are finding errors via SMART checks - I'm not.

I've had to destroy and re-create the ZFS pool about 5 times now trying to solve the issue. Reinstalled ubuntu server once with minimal software.

Running pretty low on ideas after 2 days of effort..

UPDATE: Ran a copy overnight to a ZFS pool with the 2 drives that haven't had any write issues yet. It froze. I can't see when or how much it copied because when I try to run ls or sudo mount or pretty much anything I get file not found errors. So unless the kernel decided it was a good idea to unmount the system drive, I'd say my SATA controller or borked, which means the motherboard is probably on its way out.. That being said, I did figure out this morning that it's at least 5 years old, 2 of which were spent running 24/7 so it could be time to replace it :( Will keep this post updated with any other news.

3
  • 1
    Are you in a position to exclude the controller card as the issue? I had one that sounds similar to your problem and it was the card that was faulty. Sep 10, 2014 at 9:28
  • 1
    What type of hardware is involved here? Which controller is in use? Also, RAIDZ2 with 4 disks doesn't make sense. Could you try RAID mirrors?
    – ewwhite
    Sep 10, 2014 at 12:49
  • Just using desktop/gaming hardware, so on-mobo SATA controller. 4 disks should be fine? It's the min for raidz2.. Just looking for some redundancy when a drive dies. Don't really have any other controllers to test with =/ If that were the case though, wouldn't the OS be having major issues too? Because it seems fine as long as the main processes don't touch the frozen drives when they halt.
    – shousper
    Sep 10, 2014 at 15:13

1 Answer 1

0

Turns out my SATA Controller is definitely dead. I've swapped out everything else I possibly could, but even on boot it is occasionally just not detecting drives are connected, so it's hard to point at anything else.

Thanks to those that commented and tried to help :)

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .