NFS mount is not responding

[Sean Dale]

I have a filesystem exported (HPUX 11) to another server (also HPUX 11). The connection seems to have been broken. When I try to umount, I get the error saying the filesystem is in use. When I try to use fuser to see what is using it, it does not respond. The mount command shows it as mounted. When I issue bdf, it shows an error that it is not responding and repeats forever. Is there any way I can get the nfs filesystem unmounted without breaking other mounts and without rebooting? Thanks

[Pete Randall · Accepted solution]

You could try stopping and restarting all the NFS services.


Pete

[Mel Burslan]

try stopping your nfs client which should release the nfs mount from the client side

/sbin/init.d/nfs.client stop

[saju_2]

Hi

It is a good idea to re-read te /etc/exports file by executing exportfs -a. Then check the NFS mount status. If it is not resolving do a nfs stop and start.

Regards
CS

[Sean Dale]

When you say client side, do you mean the remote server (the one exporting) or the local server where it is mounted?

[Mel Burslan]

client=local server, i.e., the one it is mounting it.

other one, i.e. the remote one, is called the nfs server in nfs terminology

[Pete Randall]

By the way, I should warn you that I don't have much confidence in this NFS stop and start actually helping. In my experience, the only way to clear this is to re-boot.


Pete

[Sean Dale]

I tried to stop and restart the nfs client. When I then typed umount /mnt/misc I still got the error:

nfs umount: nfs_unmount: /mnt/misc: is busy
umount: return error 1.

[Mel Burslan]

just stop the nfs.client and leave it as is for the time being.

then do a bdf on that mount point. It should return the bdf info from the parent directory level.

If it is still complaining about a stale nfs handle or erorring out, you are out of luck and have to reboot as Pete said. This what I was explaining is kind of a hack. it may work if you are lucky. Sometimes, you just are not that lucky unfortunately.

[Sean Dale]

It looks like I will have to reboot (and it's a production machine). Nobody will be happy about it , but, at least when the machine comes back, it will run at normal speed again. Because it is trying to connect to the failed nfs mount, the server is running very slowly and affecting our Oracle databases. Thank you to everyone for responding quickly with your insight. I really do appreciate it.

Sean

[Mel Burslan]

As I have indicated I am not even knowledgeable in solaris let alone being an expert but a little googling turned out a document like this:

Title:How to Set Password Aging on a User Account
http://docs.sun.com/app/docs/doc/817-1985/6mhm8o5lt?a=view

As I do not have a decent solaris box under my hand to test it, I hjave no clue if it will work or not, but you may want to glance thru it and see if it serves your purpose.

[Mel Burslan]

sorry the previous posting was intended for another question. please disregard

[Sean Dale]

I just remembered that I have a perl script that is run with Cron every 30 minutes that does a bdf (and other things) to make sure the filesystems for Oracle are not close to getting full. I think that script might be what is causing the error to say the filesystem is in use. I am going to try to kill all of the instances of it sitting in the ps -ef queue then try again to umount the filesystem. Thanks again to all who helped.