blkback errors in dmesg

[Demi Marie Obenour]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Occasionally, when my system is under heavy load, it will freeze for a
few seconds. During this time, the pointer is still reasonably
responsive, but nothing else is. When the freeze ends, I get these
entries in dmesg:

Feb 16 13:37:23 dom0 kernel: xen-blkback: Scheduled work from previous purge is still busy, cannot purge list
Feb 16 13:37:23 dom0 kernel: xen-blkback: Scheduled work from previous purge is still busy, cannot purge list

What is the meaning of these entries? Is it a red herring?
- --
Sincerely,
Demi Marie Obenour (she/her/hers)
Invisible Things Lab
-----BEGIN PGP SIGNATURE-----
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=Jjcr
-----END PGP SIGNATURE-----

[Ulrich Windl]

On 2/16/22 22:06, Demi Marie Obenour wrote:
> Occasionally, when my system is under heavy load, it will freeze for a
> few seconds. During this time, the pointer is still reasonably
> responsive, but nothing else is. When the freeze ends, I get these
> entries in dmesg:
>
> Feb 16 13:37:23 dom0 kernel: xen-blkback: Scheduled work from previous purge is still busy, cannot purge list
> Feb 16 13:37:23 dom0 kernel: xen-blkback: Scheduled work from previous purge is still busy, cannot purge list
>
> What is the meaning of these entries? Is it a red herring?
>

My guess is that the relation of block I/O requests to block device
speed is bad (i.e.: to much I/O for the slow disk). If the I/O is caused
by paging (swap), you might want more RAM or trim down your VMs to use
less RAM.

Despite of that I think there was a recent Xen change limiting block
queues (or something like that).

Regards,
Ulrich