SUMMARY: nfs server broken: svcktcp_reply: mbuf_send returned 32

From: Why not ? <gege_at_cal.enst.fr>
Date: Wed, 11 Feb 1998 22:52:52 +0100 (MET)

        Hi Managers,

        This is the lame excuse of the century... Well, I did not past any
summary since none of the answers proved to do something. We were running
solaris 2.4 on NFS clients. UPgrading ( a few months ago) to solaris 2.5-1
seems to have cleared the problem. We cannot tell for sure since this problem
was only happening one day in a month - the one I wanted to save for my family -

        Best regards,
- gg

> Dear managers,
>
> I am suddenly stuck with a problem with NFS.
> suddenly, all clients to the NFS server on DU cannot access their disks,
> and the /var/adm/message fills up with messages such as the following...
>
> >Dec 2 15:33:52 beethoven vmunix: svcktcp_reply: mbuf_send returned 32
> >Dec 2 15:33:52 beethoven vmunix: rfs_dispatch: sendreply failed
> >Dec 2 15:33:52 beethoven vmunix: svcktcp_reply: mbuf_send returned 32
> >Dec 2 15:33:52 beethoven vmunix: rfs_dispatch: sendreply failed
> >Dec 2 15:33:52 beethoven vmunix: svcktcp_reply: mbuf_send returned 32
> >Dec 2 15:33:53 beethoven vmunix: rfs_dispatch: sendreply failed
> >Dec 2 15:33:53 beethoven vmunix: svcktcp_reply: mbuf_send returned 32
> >Dec 2 15:33:53 beethoven vmunix: rfs_dispatch: sendreply failed
> >Dec 2 15:33:53 beethoven vmunix: svcktcp_reply: mbuf_send returned 32
>
> and then i have to reboot the machine to make nfs work again
> (kill -9 nfs leaves the process in the U state)
>
> Does this ring any bell ?

        Harald Lundberg wrote the following, but it did not cure my case.

>Sure. I have the same problem on an Ultrix machine, although I just get
>the 'sendreply failed' message and not the 'mbuf_send' message. But yes,
>one of my NFSDs is in 'U' state and I havbe to reboot the machine.
>
>This was worse earlier, before I fixed my Solaris clients. Solaris
>'fixed' an old NFS bug that caused this behaviour on hosts that didn't
>have this bug fixed. What I did to my Solaris clients was add the line
>set nfs:nfs_shrinkreaddir=1
>in "/etc/system" and reboot (the solaris clients).
>
>If you come up with something else, please tell me too. This happens to me
>1-2 times per month *still*, and usually during low-traffic hours. (last
>time it happened at 10 pm).
Received on Wed Feb 11 1998 - 22:53:30 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:37 NZDT