Re: neighbour table overflow

[Brian May <bam@snoopy.apana.org.au>]

I hope nobody minds quoting there private messages in public, but I
couldn't be bothered replying to each person with the same message in
individually.

Perhaps I should only reply to the list and not each user individually?
Like it or not, I am doing it this way now ;-)

I have truncated the quotes to aid readability.

On Sun, Oct 03, 1999 at 12:02:01PM +0200, Alexander Reelsen wrote:
> Morning...
> 
> On Sun, Oct 03, 1999 at 09:28:21AM +1000, Brian May wrote:
> > > > > Do you have enabled automatic routing in your kernel?
> > > > What kernel options is this? I have 2.2.12.
> > > [ ] IP: kernel level autoconfiguration
> > > CONFIG_IP_PNP
> > Yes - I have that enabled.
> Ok. That's normally not a problem. But here it duplicated the routes as well..

Do other people how experience this problem have this option
enabled?

> > I wonder how they got there? As far as I am aware, no user space process
> > touches the devices accept for configuring lo:
> Now that's really strange. The routing table is neither changed in
> /etc/init.d/networking nor in in /etc/init.d/network, but you have dups in the
> rt_cache file?
> I'm quite confused now, but take a look through your init and rc.boot scripts.
> perhap you'll find something you missed before. If not, a mail to linux-kernel
> ml would perhaps the best idea.

On Tue, Oct 05, 1999 at 04:46:37PM +1000, Brian May wrote:
> I run /bin/sh when /sbin/init is run (ie before anything has
> been configured in userland).
> 
> It is initially OK.
> 
> When I ping a remote host, duplicate entries appear.
> 
> Does this make sense???

On Tue, Oct 05, 1999 at 11:41:15AM +0200, Alexander Reelsen wrote:
> On Tue, Oct 05, 1999 at 04:46:37PM +1000, Brian May wrote:
> > When I ping a remote host, duplicate entries appear.
> > 
> > Does this make sense???
> Hm. Sounds as if the autorouting from the kernel is broken. Not really broken,
> because it works, but too good.

On Thu, Oct 07, 1999 at 04:13:04PM +0200, goswin.brederlow@student.uni-tuebingen.de wrote:
> Brian May <bam@snoopy.apana.org.au> writes:
> > Sorry if this is obvious... 
> > 
> > Does anyone know what this message means?
> > 
> > neighbour table overflow
> > 
> > I have been getting it on my potato NFS-Root system. eg when installing
> > netbase. Restarting nis also causes it to display three times. nis
> > client doesn't work, but might be unrelated configuration error (it
> > looks OK to me though).
> > 
> > Is this potato specific?
> > 
> > NFS-Root specific??
> 
> I saw that on NFS-Root too.
> 
> > Linux 2.2.12 specific?
> 
> I think the system uses 2.2.5 for the server and client.
> 
> > Looks like the message is comming from the kernel (it appears
> > in /var/log/kern.log).
> > 
> > Looks almost like automatic routing or something, but my routes (only 2:
> > my network and default) are static.
> 
> The System I saw that on was with a direct 100 MBit crosslink
> connection, no switch or hub.

My guess is that you too have kernel level autoconfiguration enabled.
Do you have duplicate entries in /proc/net/rt_cache?

On Thu, Oct 07, 1999 at 11:08:39AM -0400, Michael Stone wrote:
> Someone on our local LUG list reported: 
> Another question is: when we upgraded linux 5.2 to 6.0, each time, we
> booted the system, we will get multiple messages like that : 
> neighbour table overflow
> 
> Another person reported after an upgrade to redhat 6.0:
> starting NFS statd: neighbour table overflow
> neighbor table overflow [OK]
> 
> So, it's probably not a debian-specifc problem and probably lies in the
> NFS stuff. You might try disabling nfs locking?

I have the mount "nolock" option enabled. I am not sure I trust it
though (from memory there once was a bug that meant its value was
ignored, but that was sometime ago. Hopefully it has been fixed...)

Is this with kernel level IP autoconfiguration?

On Thu, Oct 07, 1999 at 10:23:35AM -0400, Tony Daniello wrote:
> I am experiencing the similar problem with my Linux Installation.  I was
> wondering if you had found out a solution to this problem and if possible
> share the solution.

Yes, I still have duplicate routes in /proc/net/rt_cache.

No, I no longer get that message. I am not sure why? Perhaps I am no
longer doing the right thing to trigger it anymore.
-- 
Brian May <bam@snoopy.apana.org.au>