roland.kammerer at linbit.com
Mon Jul 30 10:09:30 CEST 2018
On Fri, Jul 27, 2018 at 01:52:55PM +0100, Yannis Milios wrote:
> One last thing I forgot to mention in the last post is ...
> When creating a VM or CT via PVE webgui it fails with the below:
Okay, that is something for the LINSTOR people to look into. Maybe that
happend because of the "linstor-server" vs.
"linstor-controller/linstor-satellite" confusion and the according
service was not started there.
> Did some investigation on the linstor side and realised as a possible
> problem to be the following:
> root at pve3:~# linstor r c pve2 vm-101-disk-1
> The default storage pool 'DfltStorPool' for resource 'vm-101-disk-1'
> for volume number '0' is not deployed on node 'pve2'.
> The resource which should be deployed had at least one volume
> definition (volume number '0') which LinStor tried to automatically create.
> The default storage pool's name for this new volume was looked for in its
> volume definition's properties, its resource's properties, its node's
> properties and finally in a system wide default storage pool name defined
> by the LinStor controller.
> Node: pve2, Resource: vm-101-disk-1
> If I specify the '--storage-pool drbdpool' option on 'linstor r c pve2
> vm-101-disk-1' , then the resource is being assigned properly to the
> cluster node.
> Could this be the problem that PVE fails as well ?
No. The plugin creates the storage exactly once, after that it only
creates/removes diskless assignments. When it creates the volume, it
uses "--auto-place". This is clever enough to look through your pools
and take the one (per node) that fits best. We assume you only have one
per node in a Proxmox setup. If you have multiple, one will be chosen.
If you do the resource creation manually (which you did in the second
step), then LINSTOR wants to know the storage pool. But "manual
assignment" and using "--auto-place" are different things.
More information about the drbd-user