Hi all,
We've decided to move from VMware to PVE and I'm very happy with PVE so far.
I'd like to ask about storage design - our most common setup is 3 hosts what have (one) storage connected via fiber/hba (to all three hosts). After exploring various options I've end up with ocfs2 as clustered FS (with multipath on each host) published as folder to PVE and qcow2 images to have snapshots.
Is there any better way to set it up to have snapshots and storage attached to all hosts to allow seamless migrations?
With this setup I've noticed some problems which might be related to ocfs2 (as I've not seen in on our another setup where we're using ceph):
We've decided to move from VMware to PVE and I'm very happy with PVE so far.
I'd like to ask about storage design - our most common setup is 3 hosts what have (one) storage connected via fiber/hba (to all three hosts). After exploring various options I've end up with ocfs2 as clustered FS (with multipath on each host) published as folder to PVE and qcow2 images to have snapshots.
Is there any better way to set it up to have snapshots and storage attached to all hosts to allow seamless migrations?
With this setup I've noticed some problems which might be related to ocfs2 (as I've not seen in on our another setup where we're using ceph):
- When I was converting vmdk to qcow2 (via qemu-img convert -p -f vmdk source.vmdk -O qcow2 ../images/id/destination.qcow2) directly to ocfs2 storage the qcow2 image was corrupted somehow and VMs didn't boot - when I converted it to different storage (NFS) and then migrated it to ocfs2 all was fine
- I'Ve setup the HA and added VMs as resources and migrate option in the cluster config but for some reason when I'm rebooting host (to apply updates) and HA kick in to migrate VMs to other hosts the VMs are killed and then started on another server with error message that disk (../images/id/id.qcow2) was not found. When the original server is back online the VMs are migrated without any problems.
