I'm setting up a two-node Heartbeat+Pacemaker cluster to run Xen domUs in a fail-over and live-migrate type environment. Storage is a LUN on an (EqualLogic) iSCSI SAN device. I initially wanted to use LVM to manage the storage, with each domU having it's own root and swap LV on a VG dedicated to domUs. I would have another VG that holds service data (eg, Cacti monitoring, Nagios data...). As it's an active/passive cluster only one node would be accessing an LV on either VG at any given time, except for fail-over and live-migration. From what I gather thus far, you could do fail-over with LVM, but not live-migration, because in the case of live-migration you have >=2 nodes accessing the data at once and LVM is not cluster aware.
Now, the options seem to be to use cLVM, or a cluster aware file system such as OCFS2. Most of the information I've found online seems to geared towards DRBR + OCFS2 + iSCSI + LVM + ... + n. I'd rather avoid that amount of complexity.
Is cLVM an acceptable solution to the environment presented, or is there a need for OCFS2 in this setup?
Since no more than one machine should ever access a volume at a time in your proposed setup, clvm should be sufficient. OCFS2 is only necessary if two or more machines need to have a shared volume mounted at the same time. Clvm makes sure that the two xen hosts knows what part of the LUN that a specific volume is located at and unless you start the same virtual machine on both machines at the same time, that particular volume is only accessed by one (virtual) machine at a time. You could consider using XenServer as the OS for your two host servers and add them together in a cluster setup, which will automatically take care of LVM and preparation of the LUN and all that. Automatic fail-over (called HA in XenServer) is only available by purchasing an advanced license. The free version allows clustering, live-migration and so on and means that pretty much everything about Xen has been setup for you.