Re: 'Socket destroy delayed'

Sheldon E. Newhouse (
Wed, 3 Apr 1996 22:55:00 -0500

Andrew C. Esh writes:
> Paul Slootman wrote:
> > FWIW:
> >
> > I can reliably get a "socket destroy delayed" by pinging a non-existent
> > host on the local network, and hitting ctrl-c after a couple of seconds.
> > This then shows up almost immediately in the syslog:
> >
> > Apr 1 15:51:56 pcpaul kernel: Socket destroy delayed
> > Apr 1 15:52:06 pcpaul kernel: Socket destroy delayed
> ...
> > This is on 1.3.73. I've already patched up to 1.3.80, but I'm waiting
> > for some process to finish its brute-force calculation before
> > rebooting... It has already run for more than a week.
> This is reproducible on my 1.3.82, but I only get one syslog entry,
> as I did with .79.

I had many problems with reports of (socket destroy delayed (r=0,w=252)
when I was running NFS and two networks. Since I dropped the NSF, the
problem has disappeared.

Here is my setup. I have run Slackware 3.0 and RedHat 3.0.3, both
using kernels 1.3.67, 1.3.78, 1.3.80. All have the same problem.

I have two ethernet cards in my box. One connected to a local net
(net1) and the other connected to a different net (net2). All nets are
thinwire. (I had them both on the same hardware net for awhile with
different static routes, but that made no difference in what I am

On net1, I had mounted user home directories (from a SUN running Solaris
2.3 to the linux box). The nets would run for awhile after booting, but
intermittently would stop. The local net where the NSF mounts were
would stop functioning. The other worked sometimes when this happened
and othertimes it too would cease. Rebooting made them work for

I dropped the NFS mounts and the problem seems to have disappeared.
I would like to get the NFS working again, but I don't know how.

Thanks for any suggestions.


| Sheldon E. Newhouse | mail: |
| Mathematics Department | www: |
| Michigan State University | telephone: 517-355-9684 |
| E. Lansing, MI 48824-1027 USA | FAX: 517-432-1562 |