This content has been marked as final. Show 4 replies
Well, actually there isn't much of a choice here: OCFS2 is the fs that OVM uses and it's not too bad either. As with all cluster fs, OCFS2 is not suited for databases, since latency can get a little high, but since in case of a HA event, the guest would likely restart anyway, one could use some other storage for that.
Actually, I still would suggest taking a closer look at what I had mentioned in your other thread:
[Use OCFS2 to Share Local Storage (DAS) Across OVM Compute Nodes|https://forums.oracle.com/forums/thread.jspa?threadID=2438632&tstart=0]
I still think that this would also solve thus thread. In the end, there is no how-to on how to build that up and you probably will need some help from some experts. I for myself, also didnÄt configure all of that on my own - i left ASM/Cluster part for some contractor to setup and focused on the iscsi-target/initiator setup along with the multipath setup on m VM servers.
Maybe I will do a little write-up, once we get it all sorted out, since we do experience some strage behaviour of our storages boxes, while ASM hammers ob them quite hardly, but I suspect the raid controllers to blame and not the concept… ;)
NFS isn't over complicated. You should have multiple paths to the NFS share being used and the NFS share should be on a system that has multiple NICS bonded together to provide HA. There are bonding modes available to the VM server that provide either load balance/failover or just straight up Fail over.
Either way. Oracle VM already uses OCFS2 on all its clustering connections such as repositories and server pools. It just creates a ocfs2 filesystem "on top" of the NFS filesystem. Its not really "on top". But you should get the picture. Its probabily very similar to the NFS direct connections associated with the Oracle Databases running over NFS.