Can this be related the case?
https://bugzilla.redhat.com/show_bug.cgi?id=810082

On 1 Dec 2020, at 10:25, Vinícius Ferrão <ferrao@versatushpc.com.br> wrote:

ECC RAM everywhere: hosts and storage.

I even run Memtest86 on both hypervisor hosts just be sure. No errors. I haven’t had the opportunity to run it on the storage yet.

After I’ve sent that message yesterday, the engine VM crashed again, filesystem went offline. There was some discards (again) on the switch, probably due to the “boot storm” of other VM’s. But this time a simple reboot fixed the filesystem and the hosted engine VM was back.

Since it was an extremely small amount of time, I’ve checked everything again, and only the discards issues came up, there are ~90k discards on Po2 (which is the LACP interface of the hypervisor). Since the fact, I enabled hardware flow control on the ports of the switch, but discards are still happening:

Port        Align-Err     FCS-Err    Xmit-Err     Rcv-Err  UnderSize  OutDiscards
Po1                 0           0           0           0          0            0
Po2                 0           0           0           0          0         3650
Po3                 0           0           0           0          0            0
Po4                 0           0           0           0          0            0
Po5                 0           0           0           0          0            0
Po6                 0           0           0           0          0            0
Po7                 0           0           0           0          0            0
Po20                0           0           0           0          0        13788

I think this may be related… but it’s just a guess.

Thanks,


On 1 Dec 2020, at 05:06, Strahil Nikolov <hunter86_bg@yahoo.com> wrote:

Could it be faulty ram ?
Do you use ECC ram ?

Best Regards,
Strahil Nikolov






В вторник, 1 декември 2020 г., 06:17:10 Гринуич+2, Vinícius Ferrão via Users <users@ovirt.org> написа:






Hi again,



I had to shutdown everything because of a power outage in the office. When trying to get the infra up again, even the Engine have corrupted:



[  772.466982] XFS (dm-4): Invalid superblock magic number
mount: /var: wrong fs type, bad option, bad superblock on /dev/mapper/ovirt-var, missing codepage or helper program, or other error.
[  772.472885] XFS (dm-3): Mounting V5 Filesystem
[  773.629700] XFS (dm-3): Starting recovery (logdev: internal)
[  773.731104] XFS (dm-3): Metadata CRC error detected at xfs_agfl_read_verify+0xa1/0xf0 [xfs], xfs_agfl block 0xf00003
[  773.734352] XFS (dm-3): Unmount and run xfs_repair
[  773.736216] XFS (dm-3): First 128 bytes of corrupted metadata buffer:
[  773.738458] 00000000: 23 31 31 35 36 35 35 34 29 00 2d 20 52 65 62 75  #1156554).- Rebu
[  773.741044] 00000010: 69 6c 74 20 66 6f 72 20 68 74 74 70 73 3a 2f 2f  ilt for https://
[  773.743636] 00000020: 66 65 64 6f 72 61 70 72 6f 6a 65 63 74 2e 6f 72  fedoraproject.or
[  773.746191] 00000030: 67 2f 77 69 6b 69 2f 46 65 64 6f 72 61 5f 32 33  g/wiki/Fedora_23
[  773.748818] 00000040: 5f 4d 61 73 73 5f 52 65 62 75 69 6c 64 00 2d 20  _Mass_Rebuild.-
[  773.751399] 00000050: 44 72 6f 70 20 6f 62 73 6f 6c 65 74 65 20 64 65  Drop obsolete de
[  773.753933] 00000060: 66 61 74 74 72 20 73 74 61 6e 7a 61 73 20 28 23  fattr stanzas (#
[  773.756428] 00000070: 31 30 34 37 30 33 31 29 00 2d 20 49 6e 73 74 61  1047031).- Insta
[  773.758873] XFS (dm-3): metadata I/O error in "xfs_trans_read_buf_map" at daddr 0xf00003 len 1 error 74
[  773.763756] XFS (dm-3): xfs_do_force_shutdown(0x8) called from line 446 of file fs/xfs/libxfs/xfs_defer.c. Return address = 00000000962bd5ee
[  773.769363] XFS (dm-3): Corruption of in-memory data detected.  Shutting down filesystem
[  773.772643] XFS (dm-3): Please unmount the filesystem and rectify the problem(s)
[  773.776079] XFS (dm-3): xfs_imap_to_bp: xfs_trans_read_buf() returned error -5.
[  773.779113] XFS (dm-3): xlog_recover_clear_agi_bucket: failed to clear agi 3. Continuing.
[  773.783039] XFS (dm-3): xfs_imap_to_bp: xfs_trans_read_buf() returned error -5.
[  773.785698] XFS (dm-3): xlog_recover_clear_agi_bucket: failed to clear agi 3. Continuing.
[  773.790023] XFS (dm-3): Ending recovery (logdev: internal)
[  773.792489] XFS (dm-3): Error -5 recovering leftover CoW allocations.
mount: /var/log: can't read superblock on /dev/mapper/ovirt-log.
mount: /var/log/audit: mount point does not exist.




/var seems to be completely trashed.




The only time that I’ve seem something like this was faulty hardware. But nothing shows up on logs, as far as I know.




After forcing repairs with -L I’ve got other issues:




mount -a
[  326.170941] XFS (dm-4): Mounting V5 Filesystem
[  326.404788] XFS (dm-4): Ending clean mount
[  326.415291] XFS (dm-3): Mounting V5 Filesystem
[  326.611673] XFS (dm-3): Ending clean mount
[  326.621705] XFS (dm-2): Mounting V5 Filesystem
[  326.784067] XFS (dm-2): Starting recovery (logdev: internal)
[  326.792083] XFS (dm-2): Metadata CRC error detected at xfs_agi_read_verify+0xc7/0xf0 [xfs], xfs_agi block 0x2
[  326.794445] XFS (dm-2): Unmount and run xfs_repair
[  326.795557] XFS (dm-2): First 128 bytes of corrupted metadata buffer:
[  326.797055] 00000000: 4d 33 44 34 39 56 00 00 80 00 00 00 f0 cf 00 00  M3D49V..........
[  326.799685] 00000010: 00 00 00 00 02 00 00 00 23 10 00 00 3d 08 01 08  ........#...=...
[  326.802290] 00000020: 21 27 44 34 39 56 00 00 00 d0 00 00 01 00 00 00  !'D49V..........
[  326.804748] 00000030: 50 00 00 00 00 00 00 00 23 10 00 00 41 01 08 08  P.......#...A...
[  326.807296] 00000040: 21 27 44 34 39 56 00 00 10 d0 00 00 02 00 00 00  !'D49V..........
[  326.809883] 00000050: 60 00 00 00 00 00 00 00 23 10 00 00 41 01 08 08  `.......#...A...
[  326.812345] 00000060: 61 2f 44 34 39 56 00 00 00 00 00 00 00 00 00 00  a/D49V..........
[  326.814831] 00000070: 50 34 00 00 00 00 00 00 23 10 00 00 82 08 08 04  P4......#.......
[  326.817237] XFS (dm-2): metadata I/O error in "xfs_trans_read_buf_map" at daddr 0x2 len 1 error 74
mount: /var/log/audit: mount(2) system call failed: Structure needs cleaning.




But after more xfs_repair -L the engine is up…




Now I need to scavenge other VMs and do the same thing.




That’s it.




Thanks all,

V.




PS: For those interested, there’s a paste of the fixes: https://pastebin.com/jsMguw6j








On 29 Nov 2020, at 17:03, Strahil Nikolov <hunter86_bg@yahoo.com> wrote:



Damn...

You are using EFI boot. Does this happen only to EFI machines ?
Did you notice if only EL 8 is affected ?

Best Regards,
Strahil Nikolov






В неделя, 29 ноември 2020 г., 19:36:09 Гринуич+2, Vinícius Ferrão <ferrao@versatushpc.com.br> написа:





Yes!

I have a live VM right now that will de dead on a reboot:

[root@kontainerscomk ~]# cat /etc/*release
NAME="Red Hat Enterprise Linux"
VERSION="8.3 (Ootpa)"
ID="rhel"
ID_LIKE="fedora"
VERSION_ID="8.3"
PLATFORM_ID="platform:el8"
PRETTY_NAME="Red Hat Enterprise Linux 8.3 (Ootpa)"
ANSI_COLOR="0;31"
CPE_NAME="cpe:/o:redhat:enterprise_linux:8.3:GA"
HOME_URL="https://www.redhat.com/"
BUG_REPORT_URL="https://bugzilla.redhat.com/"

REDHAT_BUGZILLA_PRODUCT="Red Hat Enterprise Linux 8"
REDHAT_BUGZILLA_PRODUCT_VERSION=8.3
REDHAT_SUPPORT_PRODUCT="Red Hat Enterprise Linux"
REDHAT_SUPPORT_PRODUCT_VERSION="8.3"
Red Hat Enterprise Linux release 8.3 (Ootpa)
Red Hat Enterprise Linux release 8.3 (Ootpa)

[root@kontainerscomk ~]# sysctl -a | grep dirty
vm.dirty_background_bytes = 0
vm.dirty_background_ratio = 10
vm.dirty_bytes = 0
vm.dirty_expire_centisecs = 3000
vm.dirty_ratio = 30
vm.dirty_writeback_centisecs = 500
vm.dirtytime_expire_seconds = 43200

[root@kontainerscomk ~]# xfs_db -r /dev/dm-0
xfs_db: /dev/dm-0 is not a valid XFS filesystem (unexpected SB magic number 0xa82a0000)
Use -F to force a read attempt.
[root@kontainerscomk ~]# xfs_db -r /dev/dm-0 -F
xfs_db: /dev/dm-0 is not a valid XFS filesystem (unexpected SB magic number 0xa82a0000)
xfs_db: size check failed
xfs_db: V1 inodes unsupported. Please try an older xfsprogs.

[root@kontainerscomk ~]# cat /etc/fstab
#
# /etc/fstab
# Created by anaconda on Thu Nov 19 22:40:39 2020
#
# Accessible filesystems, by reference, are maintained under '/dev/disk/'.
# See man pages fstab(5), findfs(8), mount(8) and/or blkid(8) for more info.
#
# After editing this file, run 'systemctl daemon-reload' to update systemd
# units generated from this file.
#
/dev/mapper/rhel-root  /                      xfs    defaults        0 0
UUID=ad84d1ea-c9cc-4b22-8338-d1a6b2c7d27e /boot                  xfs    defaults        0 0
UUID=4642-2FF6          /boot/efi              vfat    umask=0077,shortname=winnt 0 2
/dev/mapper/rhel-swap  none                    swap    defaults        0 0

Thanks,


-----Original Message-----
From: Strahil Nikolov <hunter86_bg@yahoo.com>
Sent: Sunday, November 29, 2020 2:33 PM
To: Vinícius Ferrão <ferrao@versatushpc.com.br>
Cc: users <users@ovirt.org>
Subject: Re: [ovirt-users] Re: Constantly XFS in memory corruption inside VMs

Can you check the output on the VM that was affected:
# cat /etc/*release
# sysctl -a | grep dirty


Best Regards,
Strahil Nikolov





В неделя, 29 ноември 2020 г., 19:07:48 Гринуич+2, Vinícius Ferrão via Users <users@ovirt.org> написа:





Hi Strahil.

I’m not using barrier options on mount. It’s the default settings from CentOS install.

I have some additional findings, there’s a big number of discarded packages on the switch on the hypervisor interfaces.

Discards are OK as far as I know, I hope TCP handles this and do the proper retransmissions, but I ask if this may be related or not. Our storage is over NFS. My general expertise is with iSCSI and I’ve never seen this kind of issue with iSCSI, not that I’m aware of.

In other clusters, I’ve seen a high number of discards with iSCSI on XenServer 7.2 but there’s no corruption on the VMs there...

Thanks,

Sent from my iPhone


On 29 Nov 2020, at 04:00, Strahil Nikolov <hunter86_bg@yahoo.com> wrote:

Are you using "nobarrier" mount options in the VM ?

If yes, can you try to remove the "nobarrrier" option.


Best Regards,
Strahil Nikolov






В събота, 28 ноември 2020 г., 19:25:48 Гринуич+2, Vinícius Ferrão <ferrao@versatushpc.com.br> написа:





Hi Strahil,

I moved a running VM to other host, rebooted and no corruption was found. If there's any corruption it may be silent corruption... I've cases where the VM was new, just installed, run dnf -y update to get the updated packages, rebooted, and boom XFS corruption. So perhaps the motion process isn't the one to blame.

But, in fact, I remember when moving a VM that it went down during the process and when I rebooted it was corrupted. But this may not seems related. It perhaps was already in a inconsistent state.

Anyway, here's the mount options:

Host1:
192.168.10.14:/mnt/pool0/ovirt/vm on
/rhev/data-center/mnt/192.168.10.14:_mnt_pool0_ovirt_vm type nfs4
(rw,relatime,vers=4.1,rsize=131072,wsize=131072,namlen=255,soft,noshar
ecache,proto=tcp,timeo=100,retrans=3,sec=sys,clientaddr=192.168.10.1,l
ocal_lock=none,addr=192.168.10.14)

Host2:
192.168.10.14:/mnt/pool0/ovirt/vm on
/rhev/data-center/mnt/192.168.10.14:_mnt_pool0_ovirt_vm type nfs4
(rw,relatime,vers=4.1,rsize=131072,wsize=131072,namlen=255,soft,noshar
ecache,proto=tcp,timeo=100,retrans=3,sec=sys,clientaddr=192.168.10.1,l
ocal_lock=none,addr=192.168.10.14)

The options are the default ones. I haven't changed anything when configuring this cluster.

Thanks.



-----Original Message-----
From: Strahil Nikolov <hunter86_bg@yahoo.com>
Sent: Saturday, November 28, 2020 1:54 PM
To: users <users@ovirt.org>; Vinícius Ferrão
<ferrao@versatushpc.com.br>
Subject: Re: [ovirt-users] Constantly XFS in memory corruption inside
VMs

Can you try with a test vm, if this happens after a Virtual Machine migration ?

What are your mount options for the storage domain ?

Best Regards,
Strahil Nikolov






В събота, 28 ноември 2020 г., 18:25:15 Гринуич+2, Vinícius Ferrão via Users <users@ovirt.org> написа:








Hello,



I’m trying to discover why an oVirt 4.4.3 Cluster with two hosts and NFS shared storage on TrueNAS 12.0 is constantly getting XFS corruption inside the VMs.



For random reasons VM’s gets corrupted, sometimes halting it or just being silent corrupted and after a reboot the system is unable to boot due to “corruption of in-memory data detected”. Sometimes the corrupted data are “all zeroes”, sometimes there’s data there. In extreme cases the XFS superblock 0 get’s corrupted and the system cannot even detect a XFS partition anymore since the magic XFS key is corrupted on the first blocks of the virtual disk.



This is happening for a month now. We had to rollback some backups, and I don’t trust anymore on the state of the VMs.



Using xfs_db I can see that some VM’s have corrupted superblocks but the VM is up. One in specific, was with sb0 corrupted, so I knew when a reboot kicks in the machine will be gone, and that’s exactly what happened.



Another day I was just installing a new CentOS 8 VM for random reasons, and after running dnf -y update and a reboot the VM was corrupted needing XFS repair. That was an extreme case.



So, I’ve looked on the TrueNAS logs, and there’s apparently nothing wrong on the system. No errors logged on dmesg, nothing on /var/log/messages and no errors on the “zpools”, not even after scrub operations. On the switch, a Catalyst 2960X, we’ve been monitoring it and all it’s interfaces. There are no “up and down” and zero errors on all interfaces (we have a 4x Port LACP on the TrueNAS side and 2x Port LACP on each hosts), everything seems to be fine. The only metric that I was unable to get is “dropped packages”, but I’m don’t know if this can be an issue or not.



Finally, on oVirt, I can’t find anything either. I looked on /var/log/messages and /var/log/sanlock.log but there’s nothing that I found suspicious.



Is there’s anyone out there experiencing this? Our VM’s are mainly CentOS 7/8 with XFS, there’s 3 Windows VM’s that does not seems to be affected, everything else is affected.



Thanks all.



_______________________________________________
Users mailing list -- users@ovirt.org
To unsubscribe send an email to users-leave@ovirt.org Privacy
Statement: https://www.ovirt.org/privacy-policy.html
oVirt Code of Conduct:
https://www.ovirt.org/community/about/community-guidelines/
List Archives:
https://lists.ovirt.org/archives/list/users@ovirt.org/message/VLYSE7HC
FNWTWFZZTL2EJHV36OENHUGB/


_______________________________________________
Users mailing list -- users@ovirt.org
To unsubscribe send an email to users-leave@ovirt.org Privacy Statement: https://www.ovirt.org/privacy-policy.html
oVirt Code of Conduct:  https://www.ovirt.org/community/about/community-guidelines/
List Archives:
https://lists.ovirt.org/archives/list/users@ovirt.org/message/CZ5E55LJMA7Y5XUAIXBH2FMGYSUU27EV/









_______________________________________________
Users mailing list -- users@ovirt.org
To unsubscribe send an email to users-leave@ovirt.org
Privacy Statement: https://www.ovirt.org/privacy-policy.html
oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/
List Archives:
https://lists.ovirt.org/archives/list/users@ovirt.org/message/CFOSOGER2VD56UO7FNK63FV4CXQTXNHB/