Which is better XEN SAN strategy--Clustered File System (OCFS2/GFS) or use of CLVM?
Hello all, we are trying to set up a XEN-enabled virtualized failover cluster that will access a SAN. We will likely run Xen 4.0 on Debian Squeeze.
We are trying to figure out whether to go with a clustered file system such as OCFS2 or GFS, or else a 'sliced' SAN in combination with a CLVM.
1. Does cluster+CLVM-based failover work better than OCFS2? I had heard there were 'fencing' problems with OCFS2, creating frequent 'dropouts' of (ostensibly) failed nodes. But I don't know how well CLVM-based systems manage failover.
2. Does use of OCFS2 require, effectively, TWO heartbeat systems? One to manage failover of processes from failed node to replacement node, and one to manage failover of SAN resource and file access?
I have heard that
Advantages of OCFS2 are:
ease of installation
Advantages of CLVM are:
faster than OCFS2 (because OCFS2 copy-on-write can be slow)
removes single point of failure represented by OCFS2
Any preferences between these two solutions? Thanks very much in advance.