Automount

[SAM_24]

Hi,
I automounted file system like
/usr/share/man -o ro server1,server2:/usr/share/man.
My client mounted it from server2. On server2 I stopped nfs daemon.
Now my client should access files from server1???
But it says NFS stale file handle. Am I missing anything or it will not work in this way?

Thanks.

[Michael Tully]

The NFS daemons work two ways, one being the server process from server 1 and the client process from server 2. If you kill one, (I think utilising port 2049) you have effectively killed part of the communication, that is why it reports it as stale. Really the only way to fix it is to restart the NFS client on server 2 and export again the mount.

[Pete Randall]

I don't believe there is any sort of fail-over mechanism available in automount. When server2 stopped being available, the result is going to be a stale file handle.


Pete

[Sridhar Bhaskarla]

Your message was not clear to me. If you stopped NFS daemon on server from where you are mouting the filesystem, you will get stale file handles. Are you expecting it to look at server1 since server2 is not available?.

It won't work that way because the mount point will be busy looking for server2 and there is no way it can mount from server1 until you fix the stale issue.

-Sri

[Sunil Sharma_1]

Hi,
NFS client will get error and it won't be able to access mount point.
significence of giving to server name in automounter map will help for load balancing on your server, it doesn't support fail over kind of stuff.

in case of your one NFS server overloaded then new client will access FS through new server.

Sunil

[SAM_24]

Hi all,

This definition is from man page.

*********
Replicated File Systems
Multiple location fields can be specified for replicated NFS
file systems, in which case automount and the kernel will
each try to use that information to increase availability.
If the read-only flag is set in the map entry, automount
mounts a list of locations that the kernel may use, sorted
by several criteria. When a server does not respond, the
kernel will switch to an alternate server.
**********

Isn't therotically client should failover to another server???

Thanks.

[Pete Randall]

Theoretically, it sure sounds that way, however this section of the man page sounds different (if the server crashes):

"The multiple location feature for NFS mounts allows the automountd
daemon to choose the most appropriate server at mount time. While
such a mount is in effect, the daemon does not monitor the status of
the server. If the server crashes, automountd will not select an
alternative server from the list."

I also note that the section of your quote concerning the read-only flag does not appear in my 11i version of man automount. What version are you running?


Pete

[Pete Randall]

I also checked an 11.0 system and the " If the read-only flag is set in the map entry, automount
mounts a list of locations that the kernel may use, sorted
by several criteria. When a server does not respond, the
kernel will switch to an alternate server." quote does not appear there, either. Is this a 10.20 system? Automount on 10.20 was prone to lots of problems which have improved dramatically in 11.x, though it would appear that the feature which you seek has disappeared rather than being enhanced.


Pete