From owner-freebsd-questions Wed Sep 25 12:29:46 1996 Return-Path: owner-questions Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id MAA09682 for questions-outgoing; Wed, 25 Sep 1996 12:29:46 -0700 (PDT) Received: from root.com (implode.root.com [198.145.90.17]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id MAA09655 for ; Wed, 25 Sep 1996 12:29:41 -0700 (PDT) Received: from localhost (localhost [127.0.0.1]) by root.com (8.7.5/8.6.5) with SMTP id MAA06910; Wed, 25 Sep 1996 12:30:39 -0700 (PDT) Message-Id: <199609251930.MAA06910@root.com> X-Authentication-Warning: implode.root.com: Host localhost [127.0.0.1] didn't use HELO protocol To: Sonja Jo Krenz-Bush cc: questions@freebsd.org Subject: Re: help! system crashing 2-4 times a day with vinvalbuf - flush failed In-reply-to: Your message of "Wed, 25 Sep 1996 12:03:20 PDT." From: David Greenman Reply-To: dg@root.com Date: Wed, 25 Sep 1996 12:30:39 -0700 Sender: owner-questions@freebsd.org X-Loop: FreeBSD.org Precedence: bulk >Our server, which had been running fine for the last 7 months started >crashing 2-4 times a day this week. Today, I finally managed to see the >error message echoed to the console (or part of it). > >It said: > >vinvalbuf - flush failed > >We're running FreeBSD 2.1-STABLE. This system serves as web server and >shell access server. Usually around 6-10 people logged on, most of them >reading mail and/or news. > >Things I have noticed - we've had some unkillable processes that were >mail related - ie, /usr/bin/mail, pine, RMAIL, etc... Crashes seem to >happen when something like this is going on. In the past, this had >happened occascionally but never crashed the system. > >Things that have changed: our /usr/spool/mail is NFS mounted from the mail >server. Earlier this week, the drive on the mail server started giving us >problems so we replaced it. The mail server itself isn't having any >problems since we replaced the disk and is running FreeBSD 2.1.0-RELEASE. It's almost a sure bet that NFS is the cause, although I don't know precisely what would trigger something like this. NFS doesn't do any vnode locking, so it's easy to imagine how new buffers could get attached during a flush (causing the panic). >Is this something that is fixed in 2.1.5R? Let me know if I need to give >additional information - I'm not the person who primarily does this but >wanted to get started on figuring this out right away. If the cause is what I think it is, not likely. -DG David Greenman Core-team/Principal Architect, The FreeBSD Project