Proxmox 4.4 and GFS2

[chs_voks]

Hello. I've got 4-node PVE cluster and Storwize v3700 iSCSI 10g storage.

We would like to have the ability to take snapshots, live migration and thin provisioning.
I don't want to use Ceph because software defined storage over storwize not a best idea.
I cant use LVM-Thin because it is not shared.

I see the only solution is to use some cluster filesystem with qcow2 disks. OCFS2 is not supported but what is about GFS2? Can I use it with PVE 4.4? As I know GFS2 need to install cman with pacemaker which will conflict with PVE cluster.

Or maybe there's a better solution?

 

[mir]

Hello. I've got 4-node PVE cluster and Storwize v3700 iSCSI 10g storage.

We would like to have the ability to take snapshots, live migration and thin provisioning.
I don't want to use Ceph because software defined storage over storwize not a best idea.
I cant use LVM-Thin because it is not shared.

I see the only solution is to use some cluster filesystem with qcow2 disks. OCFS2 is not supported but what is about GFS2? Can I use it with PVE 4.4? As I know GFS2 need to install cman with pacemaker which will conflict with PVE cluster.

Or maybe there's a better solution?

Your question has been asked a lot on this forum so try the search functionality.

Apart from this you, IMHO, have two choices:
1) LVM (thick): Expose LUN's through both controllers of the Storwize and connect using multipath. On top of multipath create you pv and vg for lvm volumes. This is the recommended proven way. Means only raw images and no snapshot.
2) GlusterFS: Expose LUN's through both controllers of the Storwize and connect using multipath. On top of multipath create your gluster bricks volumes. This is the recommended but lessor proven way. Raw and Qcow2 images. Snapshot available if using Qcow2 image.

 

[chs_voks]

Your question has been asked a lot on this forum so try the search functionality.

Apart from this you, IMHO, have two choices:
1) LVM (thick): Expose LUN's through both controllers of the Storwize and connect using multipath. On top of multipath create you pv and vg for lvm volumes. This is the recommended proven way. Means only raw images and no snapshot.
2) GlusterFS: Expose LUN's through both controllers of the Storwize and connect using multipath. On top of multipath create your gluster bricks volumes. This is the recommended but lessor proven way. Raw and Qcow2 images. Snapshot available if using Qcow2 image.

mir, thanks for your answer. I've read the forum and all information about 3 or 2 PVE versions. GlusterFS works in FUSE this means poor performance and actually it is SDS over hardware storage too.

What is the current status GFS2 in Proxmox 4?

 

[mir]

GlusterFS works in FUSE this means poor performance and actually it is SDS over hardware storage too.

AFAIK gluster does not use FUSE in proxmox for KVM - QEMU has build-in support for gluster. For LXC it might be a different story.

What is the current status GFS2 in Proxmox 4?

I think latest version is 3.8