[ovirt-users] How to reset volume info in storage domain metadata LV

[Kuko Armas]

Recently I got a disk snapshot marked as illegal in an VM due to a failed live merge

I fixed the snapshot manually with qemu-img rebase, now qemu-img check shows everything OK

I also set the imagestatus to OK in the engine database, as shown below

engine=# select image_guid,parentid,imagestatus,vm_snapshot_id,volume_type,volume_format,active from images where image_group_id='c35ccdc5-a256-4460-8dd2-9e639b8430e9';
              image_guid              |               parentid               | imagestatus |            vm_snapshot_id            | volume_type 
| volume_format | active 
--------------------------------------+--------------------------------------+-------------+--------------------------------------+-------------
+---------------+--------
 07729f62-2cd2-45d0-993a-ec8d7fbb6ee0 | 00000000-0000-0000-0000-000000000000 |           1 | 7ae58a5b-eacf-4f6b-a06f-bda1d85170b5 |           1 
|             5 | f
 ee733323-308a-40c8-95d4-b33ca6307362 | 07729f62-2cd2-45d0-993a-ec8d7fbb6ee0 |           1 | 2ae07078-cc48-4e20-a249-52d3f44082b4 |           2 
|             4 | t

But when I try to boot the VM it still fails with this error in the SPM

Thread-292345::ERROR::2017-07-06 14:20:33,155::task::866::Storage.TaskManager.Task::(_setError) Task=`2d262a05-5a03-4ff9-8347-eaa22b6e143c`::Unexpected error
Traceback (most recent call last):
  File "/usr/share/vdsm/storage/task.py", line 873, in _run
    return fn(*args, **kargs)
  File "/usr/share/vdsm/logUtils.py", line 49, in wrapper
    res = f(*args, **kwargs)
  File "/usr/share/vdsm/storage/hsm.py", line 3227, in prepareImage
    raise se.prepareIllegalVolumeError(volUUID)
prepareIllegalVolumeError: Cannot prepare illegal volume: ('ee733323-308a-40c8-95d4-b33ca6307362',)
Thread-292345::DEBUG::2017-07-06 14:20:33,156::task::885::Storage.TaskManager.Task::(_run) Task=`2d262a05-5a03-4ff9-8347-eaa22b6e143c`::Task._run: 2d262a05-5a03-4ff9-8347-eaa22b6e143c (u'146dca57-05fd-4b3f-af8d-b253a7ca6f6e', u'00000001-0001-0001-0001-00000000014d', u'c35ccdc5-a256-4460-8dd2-9e639b8430e9', u'ee733323-308a-40c8-95d4-b33ca6307362') {} failed - stopping task

Just before that error I see this operation in the log:
/usr/bin/taskset --cpu-list 0-23 /usr/bin/dd iflag=direct skip=11 bs=512 if=/dev/146dca57-05fd-4b3f-af8d-b253a7ca6f6e/metadata count=1

If I run it manually I get this:
[root at blade6 ~]# /usr/bin/taskset --cpu-list 0-23 /usr/bin/dd iflag=direct skip=11 bs=512 if=/dev/146dca57-05fd-4b3f-af8d-b253a7ca6f6e/metadata count=1
DOMAIN=146dca57-05fd-4b3f-af8d-b253a7ca6f6e
VOLTYPE=LEAF
CTIME=1497399380
FORMAT=COW
IMAGE=c35ccdc5-a256-4460-8dd2-9e639b8430e9
DISKTYPE=2
PUUID=07729f62-2cd2-45d0-993a-ec8d7fbb6ee0
LEGALITY=ILLEGAL
MTIME=0
POOL_UUID=
SIZE=209715200
TYPE=SPARSE
DESCRIPTION=
EOF

So I guess the volume info is cached in the storage domain's metadata LV, and it's still in ILLEGAL status there

Is there a way to force ovirt to update the information in the metadata LV?

Of course I've thought of updating manually with dd, with it seems too risky (and scary) to do it in production 

Salu2!
-- 
Miguel Armas
CanaryTek Consultoria y Sistemas SL
http://www.canarytek.com/