[PVE-User] Live migration of LXC containers causes container disk corruption.

[Thomas Lamprecht]

Hi,

On 1/22/18 11:29 AM, Carles Xavier Munyoz Baldó wrote:
> Hi,
> I have a two nodes Proxmox cluster with a directory storate with DRBD in
> Primary/Primary mode. I can live migrate kvm virtual machines without
> problems, but I have problems when I migrate a LXC container from one
> node to the other.
> 
> The contaniner shutdowns correctly in the source node and starts fine in
> the destination node, but sudenly in the destination Proxmox node begin
> to appear logs like these ones and the file system in the container is
> remounted read-only:
> [...]
> kernel: [440900.380974] EXT4-fs error (device drbd2):
> ext4_validate_block_bitmap:384: comm loop1: bg 9770: bad block bitmap
> checksum
> 
> kernel: [440900.383052] EXT4-fs (drbd2): Delayed block allocation failed
> for inode 55312404 at logical offset 45876746 with max blocks 2 with
> error 74
> 
> kernel: [440900.384651] EXT4-fs (drbd2): This should not happen!! Data
> will be lost
> 
> kernel: [440900.384651]
> 
> kernel: [440900.387703] print_req_error: I/O error, dev loop1, sector
> 209987952
> 
> kernel: [440900.389506] Aborting journal on device loop1-8.
> 
> kernel: [440900.394396] EXT4-fs error (device loop1):
> ext4_journal_check_start:60: Detected aborted journal
> 
> kernel: [440900.396204] EXT4-fs (loop1): Remounting filesystem read-only
> 
> kernel: [440902.709123] EXT4-fs warning (device loop1): kmmpd:190: kmmpd
> being stopped since filesystem has been remounted as readonly.
> [...]
> 

DRBD 9 is not under our control on Proxmox VE anymore [1], while they
reversed their license change, they decided to have their own repository
where the respective packages get maintained[2].

So, I at least, cannot really help you here, I'm afraid.

Maybe other users here actively using a DRBD setup can help you out, else
it could be a good option to contact the DRBD user list or ask LINBIT for
support.

cheers,
Thomas

> It seems that the virtual disk for the container becamos corrupt because
> the solution to the problem is copy the disk to a new one and assign
> this new one to the container.
> 
> Why is happening this?
> Am I doing something incorrect for migrate one container from one node
> to the other in the cluster?
> 
> Thank you in advance for your help.
> 
> 

[1]: https://forum.proxmox.com/threads/drbdmanage-license-change.30404/
[2]: https://forum.proxmox.com/threads/drbdmanage-license-change.30404/page-2#post-154900