I'm gradually migrating my Alpha server over to an Opteron box running RHEL4. This involves various disk juggling including hot-plugging SCSI disks on the RH box. All the disks are in external shelves. Every now and then, some of the filesystems on this box spontaneously unmount. Until now, the unmounts seem to have been associated with adding/removing disks on the RH system and I've been logging the mount situation every 5 mins to try and get a clue why. However, during the the most recent unmount (5 filesystems unmounted) I wasnt juggling any disks, but I see this in the messages log within the 5 minute window during which the 5 filesystems disappeared. Is this significant?: Jul 6 18:44:15 ls1 kernel: mtrr: type mismatch for fd000000,800000 old: write-back new: write-combining At the time, I was killing and restarting rhnsd because of an rpm lockup due to a stalled up2date. This wasnt updating anything, just installing xemacs, and I had to kill all processes which had /var/lib/rpm/__db* open as reported by lsof. The 5 disks (one filesystem per disk) happen to be on 3 different Adaptec SCSI adapters, and in 3 different Storcase disk shelves, so I dont suspect an adapter fault or a disk-shelf fault. So far, Gooling and Bugzilla searching has shown up nothing, so any clues/suggestions would be greatly appreciated. Cheers, Terry.
