Upgrade to Fedora 30 consistently corrupts bcache storage - is there some kind of emergency brake possible during upgrade?

[Rolf Fokkens]

Hi All,

There's this issue: https://bugzilla.redhat.com/show_bug.cgi?id=1708315

Would have been better if I discovered it during Alpha or Beta, but 
unfortunately I didn't. Now we have the situation that any bcache user 
upgrading to Fedora30 involuntarily currupts his storage.

I tried to limit the impact by using a %pretrans script in bcache-tools:

%pretrans
_bcache=`find /sys/block/ -maxdepth 1 -name "bcache[0-9]*" | wc -l`
[[ $_bcache -gt 0 ]] || exit 0
cat >&2 << EOF
bcache will likely corrupt Fedora 30 systems beyond repare.
https://bugzilla.redhat.com/show_bug.cgi?id=1708315
EOF
exit 1

But this only blocks upgrading/installing bcache-tools. The rest 
(including the kernel) is still upgraded so after a reboot the storage 
will get corrupted anyway. Of course adding this to the kernel package 
would do the trick, but I'm not sure if that is an option.

Any suggestions on how to save people from running into this are very 
welcome!

Cheers,

Rolf

[Tomasz Torcz]

On Sun, May 12, 2019 at 01:09:40PM +0200, Rolf Fokkens wrote:
>  Hi All,
>  
>  There's this issue: https://bugzilla.redhat.com/show_bug.cgi?id=1708315
>  
>  Would have been better if I discovered it during Alpha or Beta, but
>  unfortunately I didn't. Now we have the situation that any bcache user
>  upgrading to Fedora30 involuntarily currupts his storage.
>   
  This is huge!  Accidentaly, after upgrade to F30 I experience
two or three I/O hangs with bcache, so I temporarily detached cache
device and (knock on wood) there are no problems (my rootfs is btrfs
raid10 over bcache devices).

  I suggest:
  1) getting this bug into F30 Common Bugs page, with hint that
  detaching cache device before upgrade could help.
  2) complement rhbz#1708315 with pointers to upstream report.


-- 
Tomasz Torcz            There exists no separation between gods and men:
xmpp: zdzichubg(a)chrome.pl   one blends softly casual into the other.

[Rolf Fokkens]

Thx! I followed your suggestions!

On 5/12/19 7:59 PM, Tomasz Torcz wrote:
>  On Sun, May 12, 2019 at 01:09:40PM +0200, Rolf Fokkens wrote:
> > Hi All,
> >
> > There's this issue: https://bugzilla.redhat.com/show_bug.cgi?id=1708315
> >
> > Would have been better if I discovered it during Alpha or Beta, but
> > unfortunately I didn't. Now we have the situation that any bcache user
> > upgrading to Fedora30 involuntarily currupts his storage.
> >
>     This is huge!  Accidentaly, after upgrade to F30 I experience
>  two or three I/O hangs with bcache, so I temporarily detached cache
>  device and (knock on wood) there are no problems (my rootfs is btrfs
>  raid10 over bcache devices).
>
>     I suggest:
>     1) getting this bug into F30 Common Bugs page, with hint that
>     detaching cache device before upgrade could help.
>     2) complement rhbz#1708315 with pointers to upstream report.
>
>