Re: Filesystem goes readonly soon after mount, cannot free space or rebalance

[Asif Youssuff <yoasif@gmail.com>]

Hi all,

Just checking in to see if there are any other ideas or upcoming
possible fixes on the horizon.

Thanks,
Asif

On Wed, Jun 23, 2021 at 12:24 PM Asif Youssuff <yoasif@gmail.com> wrote:
>
> On Wed, Jun 23, 2021 at 5:38 AM Qu Wenruo <quwenruo.btrfs@gmx.com> wrote:
> >
> >
> >
> > On 2021/6/23 下午5:32, Asif Youssuff wrote:
> > > On Tue, Jun 22, 2021 at 5:33 PM Chris Murphy <lists@colorremedies.com> wrote:
> > >>
> > >> On Tue, Jun 22, 2021 at 12:37 AM Asif Youssuff <yoasif@gmail.com> wrote:
> > >>
> > >>> I went ahead and also created two partitions each on the two new usb
> > >>> disks (for a two of four new partitions) and added them to the btrfs
> > >>> filesystem using "btrfs device add", then removing a snapshot followed
> > >>> by a "btrfs fi sync". The filesystem still goes ro after a while.
> > >>
> > >> Yeah Qu is correct, and I had it wrong. Two devices have enough space
> > >> for a new metadata BG, but no other disks. And it requires four. All
> > >> you need is to add two but it won't even let you add one before it
> > >> goes ro. So it's stuck. Until there is a kernel fix for this, it's
> > >> permanently read-only (unless we're missing some other work around).
> > >
> > > Hmm, would this be something that would be fixed in the near term?
> > > Just wondering how long I'd have to leave this as ro.
> > >
> > > Happy to continue to try any other ideas of course, but I'd rather not
> > > destroy the fs and start over.
> >
> > The problem is, I don't see why we can't do anything, including adding a
> > new device.
> >
> >  From the fi df output, we still have around 512M metadata free space,
> > it means unless there is something wrong with globalrsv, we should be
> > able to add new devices, which doesn't require much new space.
> >
> > For the worst case, I can craft a btrfs-progs program to manually
> > degrade all your RAID1C3/C4 chunks to SINGLE.
> >
> > By that you can have tons of space left, but that should be the last
> > thing to do IMHO.
> >
> > Currently I prefer to find a way to stop all background work like
> > balance/subvolume deletion, and try to add two new devices to the array,
> > then continue.
> >
> > I still can't get why "btrfs device add" would success while "btrfs fi
> > show" shows no new device.
> >
> > The new devices to add are really blank new devices, right? Not some
> > disks already used by btrfs?
>
> Yes, absolutely. I even formatted them as ext4 and rebooted in order
> to ensure that btrfs processes would not be acting on them until I did
> btrfs device add.
>
> Once I run btrfs device add, the filesystem type shows up in gparted
> as "unknown" - eg:
>
> /dev/sdk1           2048 15187967 15185920  7.2G 83 Linux
> /dev/sdk2       15187968 30375935 15187968  7.2G 83 Linux
>
> I am also adding the devices like what Paul jones suggested - like:
>
> sudo mount /media/camino/ && sudo btrfs device add -f /dev/sdl1
> /dev/sdl2 /dev/sdk1 /dev/sdk2 /media/camino/
>
> >
> > Thanks,
> > Qu
> > >
> > >>
> > >>
> > >>> Would mounting as degraded make it possible to add the disks and have
> > >>> it stick?
> > >>
> > >> No, that's even more fragile and you're sure to run into myriad
> > >> degraded raid5 bugs, not least of which are the bogus errors. And it
> > >> won't be obvious which ones are important and which ones are not.
> > >>
> > >>
> > >> --
> > >> Chris Murphy
> > >
> > >
> > >
>
>
>
> --
> Thanks,
> Asif



-- 
Thanks,
Asif