Update to UCS 4.2: Virtualization problems

UCS - Univention Corporate Server
, ,
#1

Hello,

I’ve updated the DC Master to 4.2-2 and have now problems with a UCS 4.1-5 VM. I know this thread:
https://help.univention.com/t/qemu-suspend-to-disk-and-live-migration-issues-with-ucs-4-2/6498

I have similar symptoms (VM can’t be rebooted, rootfs goes to readonly, VM goes off automatically).

But the mentioned reasons don’t apply. I’ve no snapshot, no supend-to-disk or live-migration used.

How can I fix this?

I’ve changed the hardware configuration before the upgrade of the host. I’ve also recreated the whole vm.

Kind regards,
SirTux

#2

Ok I found something:

[  489.178839] blk_update_request: I/O error, dev vda, sector 194494297
[  489.179101] blk_update_request: I/O error, dev vda, sector 194494297
[  489.179595] blk_update_request: I/O error, dev vda, sector 133236593
[  489.179638] blk_update_request: I/O error, dev vda, sector 133236593
[  489.181588] blk_update_request: I/O error, dev vda, sector 133237465
[  489.181697] blk_update_request: I/O error, dev vda, sector 133237465
[  489.181862] blk_update_request: I/O error, dev vda, sector 97740209
[  489.181935] blk_update_request: I/O error, dev vda, sector 97740209
[  627.701815] blk_update_request: I/O error, dev vda, sector 148651065
[  627.702038] blk_update_request: I/O error, dev vda, sector 148651065
[  627.702101] blk_update_request: I/O error, dev vda, sector 148651065
[  627.702136] blk_update_request: I/O error, dev vda, sector 148651065
[  627.702147] blk_update_request: I/O error, dev vda, sector 148651065
[  627.702166] blk_update_request: I/O error, dev vda, sector 148651065
[  627.702185] blk_update_request: I/O error, dev vda, sector 148651065
[  627.702203] blk_update_request: I/O error, dev vda, sector 148651065
[  627.702213] blk_update_request: I/O error, dev vda, sector 148651065
[  627.702223] blk_update_request: I/O error, dev vda, sector 148651065
[  638.878711] blk_update_request: 236 callbacks suppressed
[  638.878716] blk_update_request: I/O error, dev vda, sector 124166937
[  638.878774] blk_update_request: I/O error, dev vda, sector 124166937
[  638.878830] blk_update_request: I/O error, dev vda, sector 124166937
[  649.190009] blk_update_request: I/O error, dev vda, sector 59924889
[  649.190190] blk_update_request: I/O error, dev vda, sector 59924889
[  649.190302] blk_update_request: I/O error, dev vda, sector 59924889

Is this a hardware problem? The image is stored in a lvm volume on a mdadm raid.

#3

The problem exist also with the old hardware type “pc-1.1”. Also using non-virtio disk doesn’t help.

EDIT: Seems to be that this helped:

qemu-img check -r all $imagename
The following inconsistencies were found and repaired:

    1170213 leaked clusters
    491848 corruptions

Double checking the fixed image now...
No errors were found on the image.
1575905/1638400 = 96.19% allocated, 20.79% fragmented, 0.00% compressed clusters
Image end offset: 147899940864
#4

Looks like your qcow2 image is corrupted.
We have never seen such corruption before.
Please make sure that your VM still runs okay:

  1. check the filesystem (e2fsck)

  2. check MD5 sums:

    find /var/lib/dpkg/info/ -maxdepth 1 -name \*.md5sums -not -empty -exec cat {} + |
grep -F -v -f <(sed -ne '1~3s,^/,,p' /var/lib/dpkg/diversions) |
(cd / && LANG=C md5sum --quiet --check -) |
sed -rne 's,(.+): FAILED,/\1,p' |
tee /tmp/md5 |
xargs dpkg -S |
cut -d: -f1 |
tee /tmp/debs |
xargs apt-get --no-act --reinstall install

  3. check your applications in the VM still work.

(The qcow2 file format is normally used to store VM images as files. It can be used inside an LVM, but I wouldn’t recommend it, as it adds additional complexity, costs performance, and is just another layer where things can go wrong.)

#5

I haven’t done the second step because I wanted to upgrade to UCS 4.2. And during the upgrade process all package files are replaced. But step 1 and 3 was successful. Many thanks

Mastodon