Is there any supported way besides using a SAN (with one VE on a LUN) to have migration and failover in an OpenVZ cluster? From what I've read, things like NFS will not work, but I haven't seen any definitive information in English.
Is there any supported way besides using a SAN (with one VE on a LUN) to have migration and failover in an OpenVZ cluster? From what I've read, things like NFS will not work, but I haven't seen any definitive information in English.
I did try (using 2.6.26 OpenVZ from Debian) running VE file-systems over NFS (I had vzquota disabled, but don't know if it's required) and it worked, although somewhat slowly compared to local disk performance (which is expected).
I would suggest some kind of block level fail-over instead of NFS (DRBD is already suggested), but than again your NFS NAS might be better than my test environment :-)
I can't really recommend using cluster file-system over shared storage, because I actually tried to make OpenVZ highly available using OCFS2 and HP Eva storage, but it was quite slow (partly because of storage, partly because of OCFS2) and had unpredictable performance characteristics (I think that is storage's fault).
I can't think of any reason why OpenVZ VMs over NFS wouldn't work, although I haven't tried it myself. The problem with using NFS is that you're just moving your single point of failure from the OpenVZ server to the NFS server, so unless that's HA as well, you're not really any better off.
My favourite file storage replication mechanism is DRBD, and that should work rather nicely with something like Heartbeat to manage the failover.