From owner-freebsd-questions Thu Apr 6 13:47: 4 2000 Delivered-To: freebsd-questions@freebsd.org Received: from cerberus.techfuel.com (irvine.techfuel.com [209.80.51.55]) by hub.freebsd.org (Postfix) with ESMTP id 117B137BA41 for ; Thu, 6 Apr 2000 13:47:02 -0700 (PDT) (envelope-from kehlet@techfuel.com) Received: from basilisk.techfuel.com (basilisk.techfuel.com [172.16.1.2]) by cerberus.techfuel.com (8.9.3/8.9.3) with ESMTP id NAA83847 for ; Thu, 6 Apr 2000 13:46:58 -0700 (PDT) Received: from phoenix.techfuel.com (phoenix.techfuel.com [172.16.1.19]) by basilisk.techfuel.com (8.9.3/8.9.3) with ESMTP id NAA96166 for ; Thu, 6 Apr 2000 13:46:57 -0700 (PDT) Received: from localhost (kehlet@localhost) by phoenix.techfuel.com (8.9.3/8.9.3) with ESMTP id NAA06219 for ; Thu, 6 Apr 2000 13:46:57 -0700 X-Authentication-Warning: phoenix.techfuel.com: kehlet owned process doing -bs Date: Thu, 6 Apr 2000 13:46:57 -0700 (PDT) From: Steven Kehlet To: freebsd-questions@freebsd.org Subject: Problem SOLVED, uw-imapd and NFS (was: weird hangs: networking hosed (except for ping), can switch consoles but can't type) Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG I posted this message last December, and have been wrangling with the problem ever since--until now, I think :-). I'm almost certain what the problem is, so this message is for the archives and any luckless soul who may have the same problem :-). In summary: my 3.3 box was routinely hanging (every 1-2wks) for no apparant reason. You could connect to ports, but nothing would happen. The console would let you switch screens, but not type anything. The only solution was to reboot. Well, the problem ended up being NFS and some (mis)interaction with uw-imapd, and I only finally figured this out by going off a hunch. I've been running uw-imapd with the INBOX "snarfing" feature--basically this automatically moves mail for users from /var/mail/user to ~user/INBOX. Well, my users' home directories are NFS mounted off a NetApp... (Yes, I know the cardinal rule of NFS is never do mail over it, but I figured this was mostly for locking problems, and I figured since all mail access comes only from this one machine, file locking shouldn't be a problem). Also I really had no choice since I'm pretty much locked into uw-imapd these days, and there is no way to get uw-imapd to do its snarfing thing (which I need for another (non-freebsd) problem) to anywhere but the user's home directory. Out of ideas, I finally tried hacking imapd (by wrapping getpwnam()) to have it snarf into a local directory instead of NFS. Et voila, I've been up a month with no hangs! :-) I hear there are a ton of NFS fixes in 4.x which might fix this hanging problem, but I'll probably stick with my current solution anyway. I hope this might be helpful to anyone out there :-). Thanks to everyone who responded to my initial post! Steve > Hi :-), > > I'm using a FreeBSD 3.3 box at work as a mail/web/login server. It serves > about 30-40 people (mostly Windows users getting mail) but never really > pushes much of a load. > > However recently, it's been having weird hangs about once every week > to two weeks. All the sudden all network services appear to be hosed. > You can ping the box. You can telnet to it, at which point the connect > succeeds, but the banner and login prompt never appear. Mail and web > services behave similarly--you can connect to the port but the service > doesn't respond. You step over to the console, and you can Alt-F? to > switch consoles. However nothing else you type gets registered. > > I know this is bizarre and a total shot in the dark but I was hoping > anyone might have some suggestions as to what the problem might be. I > don't think it's some kind of packet attack since the box is behind a > well-configured firewall (unless, of course, it's somebody inside). > > Here's some information about the box: > > FreeBSD basilisk.techfuel.com 3.3-RELEASE FreeBSD 3.3-RELEASE #1: Thu Dec 2 14:37:16 PST 1999 kehlet@basilisk.techfuel.com:/usr/src/sys/compile/BASILISK i386 > > I have maxusers set to 256 and NMBCLUSTERS=2048. I don't think this is > the problem, though. > > basilisk:~-> uptime > 9:45AM up 5 days, 21:45, 7 users, load averages: 0.22, 0.05, 0.02 > basilisk:~-> netstat -m > 190/1376 mbufs in use: > 129 mbufs allocated to data > 61 mbufs allocated to packet headers > 128/496/2048 mbuf clusters in use (current/peak/max) > 1164 Kbytes allocated to network (24% in use) > 0 requests for memory denied > 0 requests for memory delayed > 0 calls to protocol drain routines > > > I have SOFTUPDATES enabled. It's also running a lot of other services > like Zope, MySQL, NOCOL, Samba, nfsd, amd, but I wouldn't think any of > these (except maybe Zope) could manage to hose the machine. > > Thanks in advance! Any suggestions are appreciated. My kernel config > file is attached. > > Steve > > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message