On Fri, 2004-01-30 at 23:58, Matt Alexander wrote:
> On Fri, 30 Jan 2004, Craig White wrote:
> 
> > On Fri, 2004-01-30 at 20:08, Matt Alexander wrote:
> > > Any Sonicwall users out there?  Have you found a way to keep your SSH
> > > connections from being dropped after a few minutes of inactivity?
> > ---
> > I will always resort to the dirty method if forced - an occasional
> > ping...
> > 
> > ping -c 1 192.168.x.x
> > 
> > put it in a cron script that runs every few minutes or in a loop that
> > you can run while connected to keep it from dropping.
> 
> 
> I found a solution...  
> 
> echo 50 > /proc/sys/net/ipv4/tcp_keepalive_time
> 
> Normally tcp_keepalive_time is set to 2 hours, but I set it to 50 seconds
> because the Sonicwall apparently will tear down the connection if it's
> been idle for 60 seconds.
---
Now you've got my interest because this might very well be adaptable to
a problem that I am having.

I have a setup where I have 2 rack mount servers and 1 rack mount NAS
system. Stuff is from Dell and I found out too late to adjust my
thinking that the NAS was Win2K Server based. So here's where the fun
begins.

Being my typical irascible self, I jumped on the Services For Unix
features included with the NAS system and they seem to largely work. But
after some period of inactivity, I haven't nailed down the amount of
time, access to the nfs mounts will have as much as a 45 second pause,
whether from the directly connected Linux system or a Windows client
attaching to the Samba server on the Linux system which pulls in the nfs
mount off the NAS/Win2K server.

My mount(s) look like this...
nas:/media              /home/filesystems/samba/media   nfs    
defaults,rsize=8192,wsize=8192,timeo=14,intr    0 0

so when I absorbed the resolution of your issue, it occurred to me that
I was probably in need of some type of adjustment in the 'keepalive'
category. Now the default for nfs mounts is UDP and not TCP. ls -l
/proc/sys/net/ipv4 offers no utility for udp adjustments.

So it seems that I could make the mount via tcp instead of udp and using
the keepalive option much as you have done above, try to keep the
connection live.

Has anyone dealt with things similar as this? Is it common to have
keepalive issues on NFS mounts? Is it a good idea to use TCP based NFS
mounts (forgetting for a moment that this is a Win2K machine that I am
mounting from)? I was going to go to Dell with this issue but I have
been buried by more burning issues but eventually, they are gonna beat
me up about the occasional 45 second holding pattern which always occurs
for the first person in the office in the morning and infrequently
during the day if the office has become inactive for some 10 or 15
minutes.

Craig