Errors in kernel log

[cyberjock]

Ok, I've confirmed that this is a bug. Ticket created: http://support.freenas.org/ticket/2497

 

[g0del]

Update, mostly for anyone in the future who finds this thread googling for help with a problem. After I started resilvering to replace the drive that I inadvertantly messed up, I started seeing even more issues, on disks that hadn't shown any problems before. At that point I shut down the machine until I could get another power supply for testing.

Today I put in a new, larger power supply. Since then I've seen zero errors showing up in the /var/log/messages file, and the resilver started running much faster than before. The resilver finished and found several CKSUM errors and a single file which had data corruption (luckily an easily replaced file), which I think is to be expected if the drives were silently corrupting data due to power issues.

I'm currently running a scrub which I expect will find and fix more CKSUM errors. Once that is finished, I'll run another scrub. If all the problems were caused by a flaky power supply, then I anticipate that the second scrub will find no errors. If the second scrub does find issues, I'll start checking to see if the bad power might have caused a drive to go bad.

At this point, it looks like the problems were all caused by a power supply issue. I'll probably update again once the scrubs are complete, but that will take several days. I'd like to thank the people who helped me.

 

[cyberjock]

Nice! Power Supply issues and RAM issues are the 2 hardest to identify because its not always obvious that the problems are there.

Congrats on getting your stuff fixed. :)

 

[titan_rw]

Ok, I've confirmed that this is a bug. Ticket created: http://support.freenas.org/ticket/2497

I think I had this problem.

Checked with the 'irc gods', and jpaetzel let me know what was tripping up the replacement. The zfs code sees the 'old' zpool configuration in the first part of the zfs partition (usually p2). The quick wipe from the gui (which is think is done automatically before drive replacement starts) doesn't clear the start of p2, only the very start of the drive, and the very end.

I solved the problem by manually clearing the start of the second partition. Then the replace disk in the gui worked fine.

Assuming you had a 2 gig swap, something like this:?

Code:

dd if=/dev/zero of=/dev/adaX bs=1m seek=2048 count=10

That should clear 10 megs starting at 2048 meg (2gig). 10 megs is probably more than is needed, but it only takes a second anyway. That's by memory, but I'm pretty sure that was what worked.

 

[slushieken]

I think I had this problem.

Checked with the 'irc gods', and jpaetzel let me know what was tripping up the replacement. The zfs code sees the 'old' zpool configuration in the first part of the zfs partition (usually p2). The quick wipe from the gui (which is think is done automatically before drive replacement starts) doesn't clear the start of p2, only the very start of the drive, and the very end.

I solved the problem by manually clearing the start of the second partition. Then the replace disk in the gui worked fine.

Assuming you had a 2 gig swap, something like this:?

Code:

dd if=/dev/zero of=/dev/adaX bs=1m seek=2048 count=10

That should clear 10 megs starting at 2048 meg (2gig). 10 megs is probably more than is needed, but it only takes a second anyway. That's by memory, but I'm pretty sure that was what worked.

This did work for me. I just used it. Saved me from wiping the whole thing - lots of time. Thanks!