Re: 5.6-5.10 balance regression?

["Stéphane Lesimple" <stephane_btrfs2@lesimple.fr>]

>> unfortunately the problem is no longer reproducible, probably due to
>> writes happening in meantime. If you still want a btrfs-image, I can
>> create one (unfortunately only without data as there is confidential
>> data in it), but as the problem is currently no longer reproducible, I
>> think it probably won't help.
> 
> That's fine, at least you get your fs back to normal.
> 
> I tried several small balance locally, not reproduced, thus I guess it
> may be related to certain tree layout.
> 
> Anyway, I'll wait for another small enough and reproducible report.

This is still reproducible on my FS, and I have the btrfs-image.
I can easily upload it somewhere, but of course I understand downloading
an image of 51G can be impractical.

An other way might be: as I know which block group is causing the problem,
as per the dmesg, maybe I can dump only the part of the metadata relevant
to this block group?

In any case I can run commands on this system, compile a custom btrfs-progs
or a custom kernel with whatever you want me to try, and reboot as many times
as necessary (this is not a production server).

I know it fails in relocate_block_group(), which returns -2, I'm currently
adding a couple printk's here and there to try to pinpoint that better.

Regards,

Stéphane.