Automount problems

[dictum9]

Having issues with automount on 11.23.
When I go to the directory, I get nothing in it.
It's a single map file that I am having issues with.
When I restart all the /sbin/init.d NFS daemons, I get this with the nfs.client daemon, what does it mean?



/sbin/init.d#: ./nfs.client stop
umountall: umount : has failed.
umountall: diagnostics from umount
nfs umount: nfs_unmount: /sapmnt/ASP/global: is busy
umount: return error 1.
umountall: umount : has failed.
umountall: diagnostics from umount
umount: cannot unmount /sapmnt/JPD/profile : Block device required
umount: return error 1.
umountall: umount : has failed.
umountall: diagnostics from umount
umount: cannot unmount /sapmnt/JPD/global : Block device required
umount: return error 1.
umountall: umount : has failed.
umountall: diagnostics from umount
nfs umount: nfs_unmount: /sapmnt/ASP/global: is busy
umount: return error 1.
killing biod
killing automountd

[Uday_S_Ankolekar]

Try with fuser -cu /sapmnt and see which process is holding that mountpoint

The filesystem looks like is busy with one of SAP process

-USA..

[dictum9]

I already did and that went nowhere:



/sbin/init.d#: fuser -cu /sapmnt/JPD/global
/sapmnt/JPD/global: fuser: could not find file system mounted at /sapmnt/JPD/global.

[dictum9]

This "Block device required" error is weird, I've never seen it before, and it's not a result of something using it.

[Tim Nelson · Accepted solution]

cat /etc/mnttab to see if the kernel still thinks it is mounted. Every time I have seen a block device required is because the system still thinks the mount is there when it is not.

[Uday_S_Ankolekar]

You may have to clean the File system try fsck on them before umount.

You might have tried fuser -k (to kill)

-USA..

[dictum9]

Yes, the filesystems are still in the /etc/mnttab.
What can I do to make them available?
Do I have to reboot the machine?



#: grep JPD /var/tmp/mnttab.out
sapjpd_map /sapmnt/JPD/global autofs ignore,direct,dev=6000008b 0 0 1192565363
sapjpd_map /sapmnt/JPD/profile autofs ignore,direct,dev=6000008c 0 0 1192565363
/dev/vgsap/lvol1 /sapmnt/JPD vxfs ioerror=mwdisable,largefiles,delaylog,nodatainlog,dev=40010001 0 0 1192565479
/dev/vgsap/lvol2 /usr/sap/JPD vxfs ioerror=mwdisable,largefiles,delaylog,nodatainlog,dev=40010002 0 0 1192565479

[Tim Nelson]

I think a reboot is going to be your last resort, at least for the one that will not let you mount it.

As far as the mounted busy ones you can still try lsof or fuser -u hostname:/mount/point to see if you can ID the process that is still in the filesystem.


I thought I saw in 11iv3 that there is a umount -f option to forcibly unmount a ghost NFS filesystem but alas we are not there yet.

[Dave Olker]

Hi all,

First of all, forcible unmount is available in 11.23 and 11.31 so you don't need to move to 11.31 to get that.

Secondly, I'm still not clear what you're trying to do. Are you wanting to remove the autofs mount points from the system or actually use AutoFS?

The /etc/mnttab entries you posted:

_________________________

# grep JPD /var/tmp/mnttab.out
sapjpd_map /sapmnt/JPD/global autofs ignore,direct,dev=6000008b 0 0 1192565363
sapjpd_map /sapmnt/JPD/profile autofs ignore,direct,dev=6000008c 0 0 1192565363
/dev/vgsap/lvol1 /sapmnt/JPD vxfs ioerror=mwdisable,largefiles,delaylog,nodatainlog,dev=40010001 0 0 1192565479
/dev/vgsap/lvol2 /usr/sap/JPD vxfs ioerror=mwdisable,largefiles,delaylog,nodatainlog,dev=40010002 0 0 1192565479
_________________________


are either AutoFS or VxFS. Are you trying to use AutoFS to perform loopback mounts to the same system? Or are these local VxFS SAP filesystems unrelated to the AutoFS ones? If you want to remove the AutoFS mount entries from the /etc/mnttab file you should be able to by issuing the command "umountall -F autofs". Does that remove the AutoFS entries?

Regards,

Dave

[dictum9]

The problem is that the automounted directories are not available. Everything else is available via the automount.
The automount server is exporting the filesystems.



#: l /sapmnt/JPD/global
total 0

#: l /sapmnt/JPD/profile
total 0

[Dave Olker]

Has this ever worked or is this a new installation? If it used to work, when did it stop working? What changed?

If you want to troubleshoot the problem I'd want to see a copy of all your AutoFS maps, your entire /etc/mnttab file, and I'd want you to collect a debug logfile from automountd.

[dictum9]

Automount has worked on this box, but this particular directory in the map file is problematic. I copied the map file from a pre-existing one and something is wrong.

I configure the map files and then restart the /sbin/init.d/*nfs* daemons. In this case, one of the daemons cannot get restarted - that looks to be the source of the problem.

[dictum9]

I will try:

umountall -F autofs

[dictum9]

umountall does not work either, what can I do?


/sbin/init.d#: umountall -F autofs
umountall: umount : has failed.
umountall: diagnostics from umount
umount: cannot unmount /sapmnt/JPD/profile : Block device required
umount: return error 1.
umountall: umount : has failed.
umountall: diagnostics from umount
umount: cannot unmount /sapmnt/JPD/global : Block device required
umount: return error 1.

[Dave Olker]

Is automountd running when you're trying to remove the AutoFS placeholders? What happens when you issue the command:

# /sbin/init.d/nfs.client start

Do you see any error messages? Can you then use the "umountall -F autofs" command?

[dictum9]

Still doesn't work.
Does it need a reboot?


/sbin/init.d#: /sbin/init.d/nfs.client start
starting NFS CLIENT networking

starting up the rpcbind
rpcbind already started, using pid: 15020
starting up the BIO daemons
/usr/sbin/biod 16
Reading in /etc/exports
starting up the Status Monitor daemon
rpc.statd already started, using pid: 15050
starting up the Lock Manager daemon
rpc.lockd already started, using pid: 15068
Starting up the AutoFS daemon
/usr/sbin/automountd
Running the AutoFS command interface
/usr/sbin/automount
mounting remote NFS file systems ...
mounting CacheFS file systems ...


/sbin/init.d#: umountall -F autofs
umountall: umount : has failed.
umountall: diagnostics from umount
umount: cannot unmount /sapmnt/JPD/profile : Block device required
umount: return error 1.
umountall: umount : has failed.
umountall: diagnostics from umount
umount: cannot unmount /sapmnt/JPD/global : Block device required
umount: return error 1.

[Dave Olker]

Looks that way.

[dictum9]

After the reboot, I get this.
all the map files look good.


/sbin/init.d#: automount -v
automount: /sapmnt/JPD/global mounted
automount: /sapmnt/JPD/profile mounted
automount: no unmounts

/sbin/init.d#: ll /sapmnt/JPD/profile
/sapmnt/JPD/profile unreadable

[Dave Olker]

Does the NFS mount succeed when you do the "ll" command? What does /etc/mnttab say? Is the filesystem mounted from the server you expect?

[dictum9]

I don't think that NFS is successful, all I get is "Unavailable" when I do ls on the directory.
What does it mean?

The two directories involved are:

/sapmnt/JPD/global and
/sapmnt/JPD/profile



/sbin/init.d#: grep JPD /etc/mnttab
sapjpd_map /sapmnt/JPD/global autofs ignore,direct,dev=60000009 0 0 1193694051
sapjpd_map /sapmnt/JPD/profile autofs ignore,direct,dev=6000000a 0 0 1193694051

[Dave Olker]

Sounds like it's time to collect a debug automountd log file. Please send a SIGUSR2 signal to the running automountd process (i.e. kill -17 ), reproduce the failing NFS mount problem, and then send another SIGUSR2 to the automountd process. The debug logging data will be written to the /var/adm/automount.log file. Hopefully the log file will give some hints as to why the NFS mount is failing on this client.

If you need help analyzing the log file you can either post it to this thread or send it to me at dave.olker@hp.com.

Regards,

Dave

[dictum9]

Thanks for the offer... Turns out it was not exported correctly on the NFS server.

The line contained

-anon=65xxxx

I removed everything after and including 'anon' and re-exported. Then everything became visible.