On Fri, Jun 08, 2007 at 11:23:42AM +0300, Pasi Kärkkäinen wrote: > On Thu, Jun 07, 2007 at 10:29:54AM -0400, Tom Sightler wrote: > > Has anyone out there played around with the root on multipath support in > > RHEL5? I just did an install on one of our IBM HS21 blades connect via > > fiber channel to our EMC Claiion CX700 using the "linux mpath" install > > method. The install went great and everything seemed completely fine. > > There were a few things I had to clean up, like telling multipathd to > > actually start by default, and removing the blacklist > > from /etc/multipath.conf, but for the most part it was a breeze. > > > > However, in looking at the way RHEL5 creates the multipath devices > > during the initrd process, it seems to me this might be a huge weakness. > > The init file in out initrd image has the following section: > > > > echo Making device-mapper control node > > mkdmnod > > mkblkdevs > > rmparts sde > > rmparts sdc > > rmparts sdg > > rmparts sda > > dm create mpath0 0 75497472 multipath 1 queue_if_no_path 1 emc 2 1 > > round-robin 0 2 1 8:0 1000 8:96 1000 round-robin 0 2 1 8:32 1000 8:64 > > 1000 > > dm partadd mpath0 > > rmparts sdh > > rmparts sdb > > rmparts sdf > > rmparts sdd > > dm create mpath1 0 10485760 multipath 1 queue_if_no_path 1 emc 2 1 > > round-robin 0 2 1 8:48 1000 8:80 1000 round-robin 0 2 1 8:16 1000 8:112 > > 1000 > > > > My reading of this means that the multpath devices are created during > > boot time using static device names/numbers. The problem with this is > > that, in most SAN environments, there is no way to guarantee that these > > devices will always be discovered in the same order. Doing something as > > simple as removing an adapter or adding a new LUN can drastically change > > the device names and major/minor numbers (or perhaps even an actual SAN > > path failure). > > > > This feels somewhere between extremely fragile and completely broken. > > Am I misinterpreting this? > > > > Yes, it's broken. > > Just add another LUN to both HBAs and you lose the root device multipath.. > It just doesn't get set up anymore. > > Hopefully this will be fixed soon.. >
Btw did you guys figure out what would be the correct fix for this? Wouldn't be so difficult to hack temporary custom mkinitrd.. -- Pasi _______________________________________________ rhelv5-list mailing list [email protected] https://www.redhat.com/mailman/listinfo/rhelv5-list
