From owner-freebsd-current Wed Nov 18 13:47:52 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id NAA12950 for freebsd-current-outgoing; Wed, 18 Nov 1998 13:47:52 -0800 (PST) (envelope-from owner-freebsd-current@FreeBSD.ORG) Received: from helmholtz.salk.edu (helmholtz.salk.edu [198.202.70.34]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id NAA12942 for ; Wed, 18 Nov 1998 13:47:50 -0800 (PST) (envelope-from bartol@salk.edu) Received: from eccles.salk.edu (eccles [198.202.70.120]) by helmholtz.salk.edu (8.7.5/8.7.3) with SMTP id NAA08246; Wed, 18 Nov 1998 13:47:22 -0800 (PST) Date: Wed, 18 Nov 1998 13:47:22 -0800 (PST) From: Tom Bartol To: Doug Rabson cc: freebsd-current@FreeBSD.ORG Subject: Re: NFS problems in -current In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Wed, 18 Nov 1998, Doug Rabson wrote: > On Tue, 17 Nov 1998, Tom Bartol wrote: > > > > > Hi all, > > > > I've been seeing the following problem for a while now (since > > 3.0-19981006-BETA) but was too busy at the time to report it or help > > diagnose it further. Now I've got some time on my hands... > > > > I can fairly consistently wedge NFS writes to our server with the > > following: > > > > dd if=/dev/zero of=ick bs=8192 count=1000 > > > > which tries to write an 8MB file to our file server which is an Auspex. I > > have mounted Auspex from my 3.0-current machine (cvsupped and built this > > morning) in NFSv3 mode. This command will wedge one or more of the nfsiod > > processes. "ps -auwx | grep nfsiod" reports the following: > > > > root 137 0.0 0.0 216 64 ?? I 9:12AM 0:02.31 nfsiod -n 4 > > root 138 0.0 0.0 216 64 ?? I 9:12AM 0:01.74 nfsiod -n 4 > > root 139 0.0 0.0 216 64 ?? D 9:12AM 0:00.27 nfsiod -n 4 > > root 140 0.0 0.0 216 64 ?? D 9:12AM 0:00.24 nfsiod -n 4 > > > > The two processes in state "D" are the wedged ones. A separate machine > > running -current from July 11 works fine. > > > > Eventually all the nfsiod's get wedged and I'm forced to reboot. NFS > > reads work fine. > > > > Is anyone else seeing similar behaviour? What can I do to try to fine > > tune a diagnosis and help solve the problem? > > > > Thanks for your help, > > The best place to start is to get a tcpdump of the set of transactions > which wedged the iod. > O.K. I've got a tcpdump of nfs traffic during a successful dd of 800KB and during a dd of 1600KB which resulted in a wedged nfsiod process. I'm not sure what to look for in the output. I don't see any obvious problems or major differences between the success and the failure, but then I'm not an expert at reading tcpdump output. Here is the command I used to capture the data: tcpdump -s 192 -vv >& nfs_traffic Which should see everything on my segment. We have switched 100BT to the desktop here and I have my own private port on the switch so my wire is pretty quite (except for the occasional broadcast). The traffic I captured contains only a few non-nfs tidbits. So, should I post my output for some helpful expert to take a look at? I'd sure like to see this problem fixed in -current... Tom To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message