SUMMARY: Problem with NIS/Automount

From: Atmane Alouache (aatmane@yahoo.fr)
Date: Thu May 18 2000 - 14:11:07 CDT


Hi Admins,

I apologize guys because I forgot to thank some of
you.
Then I want to thank :
David Evans
Bismark Espinoza
Adrian Saidac
Z.K. Zachariah
Stefan Voss
Ken Robson
Carsten B. Knudsen

for you help and all your answers.

My original post was :
----------------------
Hi Admins,

On our NIS server, one of the automount maps holds>
this line :
/files -ro,soft,intr
hostA(1),hostB(3):/export/files
The aim of this is to ensure the mount point "/files"
will be always available if hostA goes down.
Actually, if hostA goes down, hostB should take over.
But when hostA is down it does not work and the
message "Stale NFS file handle" is displayed.

Does someone understand this problem ?

Regards

Answers :
---------

Yes. It means somewhere else on your machine there is
a hard mountto hostA.David

David Evans

---

Automount cannot provide automatic fail-over. Once it mounts a filesystem, it will not automatically bind to the alternate server.

Bismark Espinoza ---

You should know - this is a prtty simple nsf mount The parantesis are saying that hostB should respond first having the priority 3 (higher is faster). If B goes down you should wait for the timeout to kick in.

Adrian Saidac ---

I don't think that automount is capable of changing the state of the world dynamically. What happens in reality is that your old mount points may still be in the mounted state when host A goes down in which case when it shifts to host B it detects the stale NFS file handle. This has to be removed by either rebooting the servers or finding out all the processes using the file handle and killing them. I don't know any tools that can do this directly. The closest one you can work with is lsof utility from sunfreeware.

Z.K. Zachariah ---

Hi,

the book "managing NFS and NIS" by Hal Styern (O'Reilley) says this about replicated servers: Note that there is no dynamic load balancing. Once a client performs a mount from a server, it continues to use that server until it unmounts the filesystem. It cannot choose a different server before unmounting its first choice. (Chapter 13: The Automounter, Advanced Map Tricks, Replicated Servers, p. 345 in the 1st editition).I think, this explains your problem.

Stefan Voss ---

Yes, this only works for fresh mounts, machines that already have the filesystem mounted have implicit references to the first server.

Ken Robson ---

Hi Atmane,

Actually, it only means that *new* mounts will be from hostB if hostA is down; machines that have /files mounted when hostA dies will not try to remount from hostB.

regards,

Carsten B. Knudsen ---

___________________________________________________________ Do You Yahoo!? Achetez, vendez! votre prix! Sur http://encheres.yahoo.fr



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:14:08 CDT