Summary : Bug in automounter

From: Geert Touquet (Geert.Touquet@uz.kuleuven.ac.be)
Date: Wed Dec 14 1994 - 02:07:39 CST


Hello,

: >
: > Bug ? Our configuration ???
: >
I have received an interesting answer of Matthew Ayres
        definately a bug, i've been talking to sun about this for about
        a month now, they've escalated it to be fixed. so far they have
        sent me 2 new versions of 101318 both of which had to be removed
        because they broke the system, I'm now waiting for news. This
        bug is serious for us because of the performance hit we experience
        when automount is trying to umount the 800+ home directories it
        has built up over the day by failing to unmount users.

         Bug number is 1169257 if you want to talk to sun about it.

Part 1 : summary bug in automounter
>
> Glenn Satchell advise me to install patch 101318: Synopsis: SunOS 5.3:
> Jumbo NIS+ patch, automountd security, autofs and loopback mounts
>
> Some other problems were solved with this patch, but not the problem
> with the automounter. thanks for the advise
>
>
> The answer of Jim McLean-Lipinski is a typically answer :
> would have to say a bug. We see a similar thing at our site, actually
> it gets worse, If someone automounts a non local home directory, and then
> loses their connection, they will not be unmounted and the automounter will
> hange for all other users as well. We did not see this behavior under 4.1.3.
> Only when we went to Solaris 2.x did it show up. Our configuration has not
> changed from when we were running 4.1.3 so that leads me to believe that is
> YASB (Yet Another Solaris Bug) although getting Sun to admit that has been
> fruitless. They claim that we are the only site that is reporting this type
> of problem. They obviously don't read this list because I've counted 3 sites
> here. Not having the staff to really dedicate time to this, I'm just waiting
> to see if it is fixed in 2.4 :(
>
>
>
>
> Thanks to Glenn Satchell, Gene Loriot
> Me too : Jim McLean-Lipinski
> Hearns
> Ian Lee
> Suzanna Vasquezs
>
>
> Thanks
>
> Touquet Geert
>
>
> Original question :
> >
> > We think, there is a bug in the automounter of Solaris 2.3 or a problem
> > of our configuration ? . He doesn't umount the mounts in /home
> > after an appropriate interval . This is ONLY for the users who
> > have their home directory on the local disks.
> >
> > The local machine : pasteur (Solaris 2.3)
> > Our configuration
> > * /etc/auto_master
> > /home /etc/auto_home -rw,soft
> > * /etc/auto_home
> > +auto_home
> > * NIS map auto_home
> > y888888 pasteur:/users/pasteur/c1t3d0s4/y888888
> > z999999 pasteur:/users/pasteur/c1t3d0s4/z999999
> > ....
> >
> > * Automounter takes the following action when a user begin a session
> > mount("/users/pasteur/c1t3d0s4/y888888", "/home/y888888 ", MS_FSS, 10, 0x00000000, 0) = 0
> >
> > * When the automounter umounts in /home, He will always umount the same user
> > (several times). The users who have their home directory not on the
> > local machine, will be unmounted !!
> > Here is the output of truss automountd
> > 3054: vfork() (returning as child ...) = 285
> > 3054: setsid() = 3054
> > 3054: open("/dev/null", O_WRONLY, 0555710) = 7
> > 3054: ulimit(UL_GDESLIM, 0) = 64
> > 3054: fcntl(7, F_GETFL, 0x00000000) = 1
> > 3054: close(1) = 0
> > 3054: fcntl(7, F_DUPFD, 0x00000001) = 1
> > 3054: ulimit(UL_GDESLIM, 0) = 64
> > 3054: fcntl(7, F_GETFL, 0x00000000) = 1
> > 3054: close(2) = 0
> > 3054: fcntl(7, F_DUPFD, 0x00000002) = 2
> > 3054: close(7) = 0
> > 3054: execve("/usr/lib/fs/lofs/umount", 0xDFFFD604, 0x0002F6E0) Err#2 ENOENT
> > 3054: _exit(2)
> > 285: Received signal #18, SIGCLD [default]
> > 285: siginfo: SIGCLD CLD_EXITED pid=3054 status=0x0002
> > 285: wait() = 3054 [0x0200]
> > 285: umount("/home/y888888 ") Err#16 EBUSY
> > 285: putmsg(3, 0xDFFFD604, 0x00031980, 0) = 0
> > 285: poll(0xDFFFDE60, 3, -1) = 1
> > 285: getmsg(3, 0xDFFFD780, 0x00031980, 0xDFFFD7CC) = 0
> > 285: time() = 784905258
> > 285: open("/etc/mnttab", O_RDWR, 0666) = 7
> > 285: fcntl(7, F_SETLKW, 0xDFFFD61C) = 0
> > 285: fstat(7, 0xDFFFD418) = 0
> > 285: ioctl(7, TCGETA, 0xDFFFD3A4) Err#25 ENOTTY
> > 285: read(7, " / d e v / d s k / c 0 t".., 8192) = 3050
> > 285: stat("/", 0xDFFFD540) = 0
> > 285: stat("/usr", 0xDFFFD540) = 0
> > 285: stat("/proc", 0xDFFFD540) = 0
> > 285: stat("/dev/fd", 0xDFFFD540) = 0
> > 285: stat("/var", 0xDFFFD540) = 0
> > 285: stat("/tmp", 0xDFFFD540) = 0
> > 285: stat("/vol", 0xDFFFD540) = 0
> > 285: read(7, 0x00050854, 8192) = 0
> > 285: lseek(7, 0, 1) = 3050
> > 285: close(7) = 0
> > 285: vfork() = 3056
> > 3056: vfork() (returning as child ...) = 285
> > 3056: setsid() = 3056
> > 3056: open("/dev/null", O_WRONLY, 0555710) = 7
> > 3056: ulimit(UL_GDESLIM, 0) = 64
> > 3056: fcntl(7, F_GETFL, 0x00000000) = 1
> > 3056: close(1) = 0
> > 3056: fcntl(7, F_DUPFD, 0x00000001) = 1
> > 3056: ulimit(UL_GDESLIM, 0) = 64
> > 3056: fcntl(7, F_GETFL, 0x00000000) = 1
> > 3056: close(2) = 0
> > 3056: fcntl(7, F_DUPFD, 0x00000002) = 2
> > 3056: close(7) = 0
> > 3056: execve("/usr/lib/fs/lofs/umount", 0xDFFFD604, 0x0002F6E0) Err#2 ENOENT
> > 3056: _exit(2)
> > 285: Received signal #18, SIGCLD [default]
> > 285: siginfo: SIGCLD CLD_EXITED pid=3056 status=0x0002
> > 285: wait() = 3056 [0x0200]
> > 285: umount("/home/y888888 ") Err#16 EBUSY
> > 285: putmsg(3, 0xDFFFD604, 0x00031980, 0) = 0
> > 285: poll(0xDFFFDE60, 3, -1) = 1
> > 285: getmsg(3, 0xDFFFD780, 0x00031980, 0xDFFFD7CC) = 0
> > 285: time() = 784905260
> > 285: open("/etc/mnttab", O_RDWR, 0666) = 7
> > 285: fcntl(7, F_SETLKW, 0xDFFFD61C) = 0
> > 285: fstat(7, 0xDFFFD418) = 0
> > 285: ioctl(7, TCGETA, 0xDFFFD3A4) Err#25 ENOTTY
> > 285: read(7, " / d e v / d s k / c 0 t".., 8192) = 3050
> > 285: stat("/", 0xDFFFD540) = 0
> > 285: stat("/usr", 0xDFFFD540) = 0
> > 285: stat("/proc", 0xDFFFD540) = 0
> > 285: stat("/dev/fd", 0xDFFFD540) = 0
> > 285: stat("/var", 0xDFFFD540) = 0
> > 285: stat("/tmp", 0xDFFFD540) = 0
> > 285: stat("/vol", 0xDFFFD540) = 0
> > 285: read(7, 0x00050854, 8192) = 0
> > 285: lseek(7, 0, 1) = 3050
> > 285: close(7) = 0
> > 285: vfork() = 3058
> > .....
> >
> > Bug ? Our configuration ???
> >
> >
> >

-- 
-- 
Geert Touquet            E-mail : Geert.Touquet@uz.kuleuven.ac.be     
                         Mail :   Universitaire Ziekenhuizen Leuven
                                  Kapucijnenvoer 33
Phone  : +(32) 16 33 29 77        B-3000 Leuven, Belgium



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:09:17 CDT