[linux-lvm] Messed up metadata for thin pool?

[Matej Cepl]

Hi,

I am using RHEL-7 for my day-to-day work and I have been playing with
thinp for my data (nobody warned me that having root on thinp is
probably a bad idea). I haven't noticed that I have used up all my
metadata space, so I have ended with LVM misbehaving, xfs corrupted, all
good stuff (/var/log/messages is on
https://luther.ceplovi.cz/dav/xfs_logs/messages.gz). I have tried to
recover the thinp following
http://man7.org/linux/man-pages/man7/lvmthin.7.html but although I have
added 112MB to metadata after running lvconvret --repair, and some
restarts, I have got this mess
https://luther.ceplovi.cz/dav/lvs-a-o-devices.txt and

a) vgchange -a y fails with

Check of pool rhel/pool00 failed (status:1). Manual repair required!
2 logical volume(s) in volume group "rhel" now active.

b) another lvconvert --repair ends with

Internal error: Missing allocatable pvs.

c) thin_check /dev/mapper/rhel-pool00_tdata fails with

couldn't stat dev path

(and that's correct it really is not there).

Could anybody point me in the direction of recovering my data, please?

Thanks a lot,

Matěj

[Matěj Cepl]

Post by Matej Cepl
Could anybody point me in the direction of recovering my data,
please?

With a lot of in-between-Christmas-help from Zdeněk Kabeláč
(thank you!) I have ended up with the working system and these
two bugs:

* https://bugzilla.redhat.com/show_bug.cgi?id=1177056
* https://bugzilla.redhat.com/show_bug.cgi?id=1177060

Best,

Matěj