From owner-freebsd-hackers Sat Oct 11 05:15:50 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id FAA17455 for hackers-outgoing; Sat, 11 Oct 1997 05:15:50 -0700 (PDT) (envelope-from owner-freebsd-hackers) Received: from zibbi.mikom.csir.co.za (zibbi.mikom.csir.co.za [146.64.24.58]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id FAA17446 for ; Sat, 11 Oct 1997 05:15:38 -0700 (PDT) (envelope-from jhay@zibbi.mikom.csir.co.za) Received: (from jhay@localhost) by zibbi.mikom.csir.co.za (8.8.7/8.8.7) id OAA20692; Sat, 11 Oct 1997 14:15:14 +0200 (SAT) From: John Hay Message-Id: <199710111215.OAA20692@zibbi.mikom.csir.co.za> Subject: Re: vm bogons in 2.2? In-Reply-To: from Marc Slemko at "Oct 11, 97 02:30:06 am" To: marcs@znep.com (Marc Slemko) Date: Sat, 11 Oct 1997 14:15:14 +0200 (SAT) Cc: hackers@FreeBSD.ORG X-Mailer: ELM [version 2.4ME+ PL31 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-hackers@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk I have also seen this. It got so bad that I had to restart innd a few times a day. After I have upgraded to inn 1.6b3 and disabled MMAP it hasn't happened again. I haven't switched MMAP on with 1.6b3, so I'm not sure which of the two changes it was that fixed it. John -- John Hay -- John.Hay@mikom.csir.co.za > This is 2.2-stable as of Aug 18. > > I have posted on similar things before (running out of swap when there > should be no reason), but haven't really got any ideas. > > What I'm seeing now is swap usage growing and growing without being able > to find anything using it. > > last pid: 9885; load averages: 0.79, 0.72, 0.74 01:59:02 > 52 processes: 2 running, 48 sleeping, 2 stopped > CPU states: 9.8% user, 0.0% nice, 17.3% system, 1.1% interrupt, 71.8% idle > Mem: 43M Active, 11M Inact, 19M Wired, 18M Cache, 8350K Buf, 2112K Free > Swap: 320M Total, 71M Used, 249M Free, 22% Inuse > > PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND > 17925 news -14 0 29740K 23424K ufslk2 237:27 8.16% 8.16% innd > 3059 news 2 0 9040K 8988K select 27:33 3.17% 3.17% innfee > 15153 root 2 0 676K 596K select 0:23 0.72% 0.72% sshd > 9883 marcs 31 0 632K 768K RUN 0:00 2.74% 0.38% top > 9885 news 33 0 252K 740K RUN 0:00 4.69% 0.23% expire > 8476 news 2 0 940K 12504K sbwait 0:17 0.08% 0.08% nnrpd > > > Then I kill pid 3059. > > last pid: 10457; load averages: 0.95, 0.85, 0.79 02:05:10 > 48 processes: 2 running, 44 sleeping, 2 stopped > CPU states: 17.5% user, 0.0% nice, 14.1% system, 3.0% interrupt, 65.4% idle > Mem: 52M Active, 6920K Inact, 19M Wired, 14M Cache, 8351K Buf, 1060K Free > Swap: 320M Total, 36M Used, 284M Free, 11% Inuse > > Why is a 10 meg process freeing 35 megs of swap? I am quite sure that > killing that process is what caused the problem. > > The process allocate and deallocates a reasonably large amount of > memory over time (say... a total of a couple of gigs over the couple > of days it had been running; not much more than ~10 at a time > though). > > I haven't seen any changes in 2.2-stable recently that would fix > this, but I will upgrade it when I get around to it to see... >