BTRFS Warning

[dpikelny]

Hi.

 

Having a bit of trouble running mover.

Unraid version is 6.7.2 (and pre-apologize if this is a noob question).

 

I have about 1.5GB on my cached SSD disk that I'm trying to move on my spinning disk array.

However, when I run mover, I get this - 

Nov 14 10:39:14 DavesTower kernel: BTRFS warning (device sdb1): csum failed root 5 ino 266 off 290033664 csum 0x48136f86 expected csum 0x4c019db8 mirror 1
Nov 14 10:39:14 DavesTower kernel: BTRFS warning (device sdb1): csum failed root 5 ino 266 off 290033664 csum 0x48136f86 expected csum 0x4c019db8 mirror 1
Nov 14 10:39:14 DavesTower kernel: BTRFS warning (device sdb1): csum failed root 5 ino 266 off 290033664 csum 0x48136f86 expected csum 0x4c019db8 mirror 1
Nov 14 10:39:14 DavesTower kernel: BTRFS warning (device sdb1): csum failed root 5 ino 266 off 290033664 csum 0x48136f86 expected csum 0x4c019db8 mirror 1

 

Next - I put the array offline.

And run this in the shell

btrfs check --check-data-csum -p /dev/sdb1

 

Opening filesystem to check...
Checking filesystem on /dev/sdb1
UUID: 26c49477-6bb5-4e95-992a-15fc5f798a42
[1/7] checking root items                      (0:00:00 elapsed, 124 items checked)
[2/7] checking extents                         (0:00:00 elapsed, 107 items checked)
[3/7] checking free space cache                (0:00:00 elapsed, 7 items checked)
[4/7] checking fs roots                        (0:00:00 elapsed, 2 items checked)
[5/7] checking csums against data              (0:00:03 elapsed, 190 items checked)
[6/7] checking root refs                       (0:00:00 elapsed, 3 items checked)
[7/7] checking quota groups skipped (not enabled on this FS)
found 1494335488 bytes used, no error found
total csum bytes: 1456960
total tree bytes: 1622016
total fs tree bytes: 32768
total extent tree bytes: 16384
btree space waste bytes: 126269
file data blocks allocated: 1492713472
 referenced 1492713472

 

No errors.

I'm like cool - I start the array - and run mover - and I get the same csum errors.

 

I've also tried running btrfs without the 'check' switch - but nothing.

 

Any ideas or next steps?

 

Thanks in advance!

 

[JorgeB]

Checksum errors are usually from bad RAM, data could be getting corrupt in RAM, you could try running a scrub but should give the same results as btrfs check --check-data-csum, and run memtest.

[dpikelny]

Thanks for the reply.

 

So far - at 3% memtest progress - I have 1 red bar.  Red isn't good right? 

 

The only fix is to get new ram, correct?  ECC won't do anything?

[JorgeB]

1 minute ago, dpikelny said:

ECC won't do anything?

Is your RAM ECC? With ECC you shouldn't get any RAM errors, even in memtest, either ECC corrects the error or halts the computer if it can't.

[dpikelny]

Sorry - I take that back - not ECC

I've had it running for 2 years so going through order history on various sites...  ugh.

 

Tried cleaning/reseating the ram - still getting red lines.

Time to go to amazon (and hope it's not the slots)....

 

Thank you for your help.

[dpikelny]

Just to close the loop - looks like it's just 1 of my 2 ram chips are bad (thank goodness it wasn't a slot issue).

Doing mover again (with the 1 good ram chip) - and it's now rebalancing a ton of stuff from spinning disks to SSD and then, I guess, presumably back to equal distribution among spinning disks.

 

So far - all without BTRFS errors.

 

Thanks again for your help!

 

[JorgeB]

You're welcome, lucky you're using btrfs at least on cache or data would be corrupted when moved to the array, and it could be some time before you notice the problem.

[dpikelny]

Wow - didn't know that.  /whew!

 

[JorgeB]

Yes, though for completion let me add that if the RAM was already bad when the data was originally written to cache it could still be corrupt, since in that case it could be corrupted in RAM and btrfs would store an already corrupted checksum, to avoid that you'd need ECC RAM, without ECC btrfs can still detect if data is being corrupted on read.