On Fri, Oct 26, 2018 at 10:07 PM Markus Stockhausen <stockhausen(a)collogia.de>
wrote:
we observed high count of getVolumeSize() calls in vdsm.log Looking
at the logs it seems as if this is done every minute for all disk that
are currently running on a node.
Is this frequency normal and if yes why?
Yes, this is expected.
Vdsm is monitoring volume size every 60 seconds, based on this
configuration:
https://github.com/oVirt/vdsm/blob/cbdbee979f0bfc7a7dd447af0d69c5aa9e0f8b...
The monitoring is needed to update engine about disk actual size when
using thin provisioning.
- With block storage, when vdsm detect that a disk is too full, it will
extend the disk with help
of the SPM host.
- With sparse disks on file storage, the disk actual size will grow as the
VM writes more
data to the disk.
Engine show the actual disk size based on the size reported by vdsm, using
data collected by
these getVolumeSize calls.
This check can probably optimized to run only on file based thin
provisioned disks instead of
any disk. Please file a bug to optimize this is you think it is an issue.
Nir