My original message is at the bottom. My problem is resolved, but I'm not sure I completely understand how I cured it. One key piece of information came from Casper Dik, of course. He pointed out that all my mount points are busy because they are being shared by NFS. Neither lsof nor fuser shows this usage. It is necessary to unshare before mounting or umounting. I didn't mention NFS in my original message. Even after unsharing the offending mount point, I couldn't mount it. The breakthrough came when I decided to fsck the partition, just to make sure the disk is really working. It looked like this: $: fsck /dev/dsk/c1t2d0s1 /dev/dsk/c1t2d0s1 is a mounted file system, ignored That's more or less the same baloney I had been seeing. For some reason I decided to umount it by device name, instead of mount point. I don't think I had ever done that before. $:umount /dev/dsk/c1t2d0s1 umount: warning: /dev/dsk/c1t2d0s1 not in mnttab Then I did the fsck again, and it worked! That umount must have accomplished something, in spite of the error message. After this, there was no problem. mount /glast/04 worked with no complaints. I have a hypothesis about what happened. I don't understand the details well enough to know if this is the truth. Suppose the file system got "mounted" at boot time, but there was no entry in mnttab for some reason. For quite a while, I was balked by the busyness of the mount point due to NFS. Even after that was cured, some programs were unable to deal with the file system by name, such as mount, umount, and df. (I don't see why mount should have suffered from this, though.) When I umount'ed the partition by device name, umount complained but did its job. Now mnttab was in agreement with reality, and things could proceed. Thanks also to Steve Maher, Marco Greene, John W. Ballard, Thomas M. Payerle, David Foster, Dan Astoorian, and jbarratt. Several pointed out the use of lsof and fuser, although I mentioned lsof in my message. Most mentioned the importance of mnttab. Original question: > I'm really confused. I can't mount one of my filesystems, and > all the mount points seem to be busy, when I know they aren't. > > This is on an Ultra 2 with Solaris 7. I just installed the new > kernel patch 106541-24 and rebooted. Apparently I installed > the patch without backing up the original files, so I can't back > it out. > > The first symptom I see is the dreaded message: > > $ /etc/mount /glast/04 > mount: /dev/dsk/c1t2d0s1 is already mounted, /glast/04 is busy, > or the allowable number of mount points has been exceeded > > According to the archives, the "number of mount points" part is > probably bogus. > > To check if it's already mounted, I ran df and looked at vfstab. > The string c1t2d0s1 doesn't appear in df, and it only appears > once in vfstab: > /dev/dsk/c1t2d0s1 /dev/rdsk/c1t2d0s1 /glast/04 ufs 4 yes logging > It's surrounded by a bunch of other similar entries. > > So maybe the problem is that the mount point is busy. That seems > awfully unlikely, but it's worth a shot. I did "rmdir /glast/04" and > it went away peacefully. I recreated it, and still mount fails. > > Now I began to wonder about the number of mount points. I decided to > unmount one of the other partitions to see if it freed up a slot. > I couldn't unmount any of them. I got this behavior: > $: /etc/umount /glast/11 > umount: /glast/11 busy > on nearly a dozen file systems. Some of these don't get used at all, > so this seems pretty unlikely. Then I ran lsof, and it tells me that > /glast/11 and its friends are not busy. I believe it, based on the > pattern of use on this computer. > > What's going on? * Patrick L. Nolan * * W. W. Hansen Experimental Physics Laboratory (HEPL) * * Stanford University * _______________________________________________ sunmanagers mailing list sunmanagers@sunmanagers.org http://www.sunmanagers.org/mailman/listinfo/sunmanagersReceived on Wed Apr 23 18:34:39 2003
This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:09 EST