OpenVZ containers must be stored on a real file system.
Maybe we could use a cluster file system on top of iSCSI or DRBD but that just adds unnecessary complexities and reduces performance.
I was thinking of a how to use OpenVZ on top of DRBD and came up with a simple idea.
Should not be too difficult to implement this either.
When creating the new container a volume would be created on the iSCSI or DRBD lvm group.
A new file system would be created on the new volume.
That file system would be mounted and the container created on that new file system.
Whenever the container is stopped the file system would be unmounted
When the container is started the file system would be mounted, then the container started.
During live migration, at the right moment, the file system would be dismounted on the source node, then mounted on the destination node.
That would ensure that the file system is only ever mounted on one node at a time.
Does this seem like a viable solution?
Maybe we could use a cluster file system on top of iSCSI or DRBD but that just adds unnecessary complexities and reduces performance.
I was thinking of a how to use OpenVZ on top of DRBD and came up with a simple idea.
Should not be too difficult to implement this either.
When creating the new container a volume would be created on the iSCSI or DRBD lvm group.
A new file system would be created on the new volume.
That file system would be mounted and the container created on that new file system.
Whenever the container is stopped the file system would be unmounted
When the container is started the file system would be mounted, then the container started.
During live migration, at the right moment, the file system would be dismounted on the source node, then mounted on the destination node.
That would ensure that the file system is only ever mounted on one node at a time.
Does this seem like a viable solution?