Balance loop converting from single to raid-1 (kernel 5.13.3)

Balance loop converting from single to raid-1 (kernel 5.13.3)

[Wei Tang]

Today I was trying to convert a btrfs disk from single to raid-1. It 
gets stuck in this seemingly infinite loop:

Jul 23 08:30:48 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:48 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:48 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:48 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:48 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers
Jul 23 08:30:50 loopforward kernel: BTRFS info (device dm-5): found 1 
extents, stage: update data pointers

Canceling the conversion then restarting it, or rebooting, does not seem 
to do anything -- the same "1 extents" loop always reappears.

# uname -a
Linux loopforward 5.13.3 #1-NixOS SMP Mon Jul 19 08:04:55 UTC 2021 
x86_64 GNU/Linux

Re: Balance loop converting from single to raid-1 (kernel 5.13.3)

[Qu Wenruo]

On 2021/7/23 下午2:39, Wei Tang wrote:
> Today I was trying to convert a btrfs disk from single to raid-1. It
> gets stuck in this seemingly infinite loop:
>
> Jul 23 08:30:48 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:48 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:48 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:48 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:48 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:49 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers
> Jul 23 08:30:50 loopforward kernel: BTRFS info (device dm-5): found 1
> extents, stage: update data pointers

This looks like a runaway balance, but that should be fixed in v5.8.

Thus it's pretty weird.


>
> Canceling the conversion then restarting it, or rebooting, does not seem
> to do anything -- the same "1 extents" loop always reappears.

You can unmount, then mount with "-o skip_balance" mount option to
workaround it.

BTW, when you unmount the fs, could you also catch the kernel messages?
As since v5.8 we added an extra warning to catch such problem.

Thanks,
Qu

>
> # uname -a
> Linux loopforward 5.13.3 #1-NixOS SMP Mon Jul 19 08:04:55 UTC 2021
> x86_64 GNU/Linux