[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [Xen-users] Xen and GFS
Thanks again. I was trying to avoid the mount/unmount complexity and use SAN space more efficiently by simply keeping all the xenU domains on the shared file system, but it looks as though that won't work quite like I had hoped. I'm not looking for HA per se (although this is important,) more for flexibility when load balancing the VMs running on all the blades, and safer live migrations from blade to blade. I'm a little nervous about having a LUN up on two boxes at the same time, as I've got some experience with killing file systems this way (in the test lab, anyway.) On Apr 18, 2006, at 1:30 PM, John Madden wrote: On Tuesday 18 April 2006 16:17, Jim Klein wrote:The setup I have is 3 - AMD_64DP server blades w/ 4Gb RAM each, attached to FC SAN. The thought was that I would create a GFS volume on the SAN, mount it under Xen dom0 on all 3 blades, create all the VBDs for my VMs on the SAN, and thus be able to easily migrate VMs from one blade to another, without any intermediary mounts and unmounts on the blades. I thought it made a lot of sense, but maybe my approach is wrong.Not necessarily wrong, but perhaps just an unnecessary layer. If your intentis HA Xen, I would set it up like this: 1) Both machines connected to the SAN over FC 2) Both machines having visibility to the same SAN LUN(s) 3) Both machines running heartbeat with private interconnects4) LVM lv's (from dom0) on the LUN(s) for carving up the storage for thedomU's 5) In the event of a node failure, the failback machine starts withan "/etc/init.d/lvm start" or equivalent to prep the lv's for use. Then xendstart, etc.For migration, you'd be doing somewhat the same thing, only you'd need a separate SAN LUN (still use LVM inside dom0) for each VBD. My understandingis that writing is only done by one Xen stack at once (node 0 beforemigration, node 1 after migration, nothing in between), so all you have to do is make that LUN available to the other Xen instance and you should be set. A cluster filesystem should only be used when more than one node must writeto the same LUN at the same time. John -- John Madden Sr. UNIX Systems Engineer Ivy Tech Community College of Indiana jmadden@xxxxxxxxxxx _______________________________________________ Xen-users mailing list Xen-users@xxxxxxxxxxxxxxxxxxx http://lists.xensource.com/xen-users
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |