I know that for container HA I must have my containers on a nfs storage. But my question is : why we can't use replicated storage like drbd, or ceph, or sheepdog ? I don't find any info about that.
Openvz does have hooks where you could create scripts to mount/unmount filesystems on startup/shutdown etc.
I've had this idea that using the hooks it would be possible to store containers on drbd and maybe ceph/sheepdog too.
Starting a container the hook would
mount the filesystem on drbd volume.
Stopping a container the hook would unmount the filesystem on drbd.
Sounds simple but doing it right and making live migration work without corrupting the filesystem will take some work.
Then you still have the issue that this is a hack to proxmox so you really should integrate this directly into proxmox so the allocation of storage, formatting the new volume etc etc is all transparent to the proxmox user.
Long story short you can not do this because the benefit of doing this is too small when you consider the time involved to implement it.
So much easier to use NFS, if you need HA NFS you can find dozens of HA NFS using DRBD articles on the Internet.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
