OVM 3.1.1 problems with HP EVA 4400 SAN, G7 hosts

This is a few words about an OVM installation at a new client.  The OVM install is awesome now, talk about a quantum leap from 2 and 3 flat.  The bonding and bridging is super simple and the multipathing worked straight out of the box (when the SAN LUNs were right).

We did have some problems with LUNs coming up on one host and not the other.  Even a rebuild of OVS on the hosts did not fix it – What?  So we had the SAN LUNs created again and then EVERYTHING worked out of the box.  It would have taken < 4 hours to build a highly available VM host infrastructure.  This gives 48 CPUs and 400 GB of RAM available in that very short amount of time.

Great doco from oracle.

http://docs.oracle.com/cd/E27300_01/index.html

quick start guide is great for when things work.  Finding answers when there are problems is much harder!

This was the error I was getting when trying to create the repository on the multipathed LUN

(07/26/2012 04:42:40:368 PM)
OVMAPI_B000E Storage plugin command [storage_plugin_createFileSystem] failed for storage server [0004fb0000090000dbe3a865f23402d3] failed with [com.oracle.ovm.mgr.api.exception.FailedOperationException: OVMAPI_4010E Attempt to send command: dispatch to server: ssydovm03.xxx.local failed. OVMAPI_4004E Server Failed Command: dispatch https://?uname?:?pwd?@192.168.8.63:8899/api/2 storage_plugin_createFileSystem oracle.ocfs2.OCFS2.OCFS2Plugin 0004fb0000050000577f7a4075a0ef54 /dev/mapper/360014380064896ea0001000001c00000 0, Status: OSCPlugin.InvalidValueEx:'The backing device /dev/mapper/360014380064896ea0001000001c00000 is not allowed to contain partitions'
Thu Jul 26 16:42:40 EST 2012
Thu Jul 26 16:42:40 EST 2012] OVMAPI_4010E Attempt to send command: dispatch to server: ssydovm03.xxx.local failed. OVMAPI_4004E Server Failed Command: dispatch https://?uname?:?pwd?@192.168.8.63:8899/api/2 storage_plugin_createFileSystem

This was fixed by creating a new LUN, there was something wrong with the disk.  Nothing wrong with OCFS2 or OVM.

The other multipath problem was also put down to a problem with the SAN LUNs.