On Mon, Jun 14, 2010 at 7:22 AM, Brian Kroth <[email protected]> wrote: > > I don't believe OCFS2 can currently support any logical volume manager > other than a simple concatenation (and even then it's with extreme > caution). The overhead involved in the lower software layer doing > striping needs to somehow be coordinated among all the nodes in the > cluster else all fs consistency guarantees provided by the SCSI layer > are lost.
Hi, Brian, and thank you for your reply. First, from what I have read, OCFS2 should work fine over any software RAID that does not provide redundancy (i.e., concatenation or RAID-0 is fine; RAID-1 or RAID-5 is not). Second, I saw similar results using cLVM2, which the Suse 11 SP1 say should work with OCFS2 (integrated cluster stack). At least, I am pretty sure I saw similar results, but I forgot to capture the logs so I am not 100% certain. Finally, I ran that "fsck,.ocfs2" from a single node while the md device unmounted on all nodes, and it gave me an "I/O error". I do not see how other nodes could cause that to happen when they are not touching the device at all. - Pat _______________________________________________ Ocfs2-users mailing list [email protected] http://oss.oracle.com/mailman/listinfo/ocfs2-users
