Am 22.01.21 um 12:01 schrieb Shantur Rathore:
Thanks Matthias,
Ceph iSCSI is indeed supported but it introduces an overhead for running
LIO gateways for iSCSI.
CephFS works as a posix domain, if we could get a posix domain to work
as a master domain then we could run a self-hosted engine on it.
Concerning this
you should look at
https://bugzilla.redhat.com/show_bug.cgi?id=1577529.
Ceph RBD ( rbd-nbd hopefully in future ) could be used with
cinderlib and we have got a self-hosted infrastructure with Ceph.
I am hopeful that when cinderlib integration is mature enough to be out
of Tech preview, there will be a way to migrate old cinder disks to new
cinderlib.
PS: About your large deployment, go OpenStack or OpenNebula if you like.
Proxmox clustering isn't very great, it doesn't have a single controller
and uses coro-sync based clustering which isn't very great.
Cheers,
Shantur
On Fri, Jan 22, 2021 at 10:36 AM Matthias Leopold
<matthias.leopold(a)meduniwien.ac.at
<mailto:matthias.leopold@meduniwien.ac.at>> wrote:
I can confirm that Ceph iSCSI can be used for master domain, we are
using it together with VM disks on Ceph via Cinder ("old style").
Recent
developments concerning Ceph in oVirt are disappointing for me, I think
I will have to look elsewhere (OpenStack, Proxmox) for our rather big
deployment. At least Nir Soffer's explanation for the move to cinderlib
in another thread (dated 20210121) shed some light on the background of
this decision.
Matthias
...