SUMMARY: Avoiding NFS clients Hang when NFS server dies

From: Tru64 User (tru64user@yahoo.com)
Date: Wed Jul 23 2003 - 14:19:43 EDT


Thanks to the quick responses from Oisin McGuinness,
Dr. Allan Rollow, David DeWolfe and Blake Brehl, all
offering differing solutions that i have to pick from.

I took the quick way out, adding "bg" to nfs mount
options. Upon testing, found both "bg" & "soft" need
to be inplace for df to work. It still, however, hangs
for approx. 30 secs. before it lets go and retries in
the background (bg).

Other solutions where automounter, autofs (i have
never implemented these, i know they exist out
there....maybe its time i studied them); Adding option
"df -n" which i have done....but it was getting out of
hand......and what would take care of it altogether?
Clustering.....(read below)

********
Interruptable is a useful intermediate option between
        "hard" and "soft" mounts. Attempts to access the
        remote file system will still hang, but they can be
        interrupted. There was a time when "soft" was very
        dangerous because application didn't gracefully
handle
        I/Os failing. Applications that are part of the base
        system should handle this better.

        The only real solution is to not have the problem of
the
        serving host becoming unavailable. The usual
solution
        to that is a cluster of some sort. Tru64 UNIX V5.0A
        and later can support this. Before V5.0A, two hosts
        could be in a failover configuration; one host fails,
        the other host picks up the load. I'm sure the Linux
        community is working on the clustering problem,
though
        it is hard done correctly. Failover might be an
option,
        if full clustering isn't there.

        Mounting inside another directory helps for those
things
        that have to traverse the root file system; pwd.

****

--- Tru64 User <tru64user@yahoo.com> wrote:
> Hi All,
>
> My tru64 boxes (mostly v4.0G) are heavy
> crossmounters
> of disks served on other tru64 hosts and Linux.
>
> Problem is that if any one host goes down, all other
> hosts hang with a df. Additionally, other programs
> that utilize "df" (eg. our backups, ssh --when it
> tries to generate random seed runs a number of
> commands, including df) hang as well.
>
> I have read and implemented two "thought" solutions:
>
> 1. mounting inside a dir i.e.
> mount -t nfs HOSTA:/dirAAA /HOSTA/dirAAA
> instead of:
> mount -t nfs HOSTA:/dirAAA /dirAAA
> It is not suggested to mount nfs dirs under root.
>
> 2.Using the "soft" option for mounting
> (default=hard).
>
> None of these are a cure. Any additional suggestions
> out there?
>
> _Thanks
>
> Richard
>
>
> __________________________________
> Do you Yahoo!?
> Yahoo! SiteBuilder - Free, easy-to-use web site
> design software
> http://sitebuilder.yahoo.com

=====

__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com



This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:49:28 EDT