From owner-freebsd-current Thu Nov 19 21:02:37 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id VAA13641 for freebsd-current-outgoing; Thu, 19 Nov 1998 21:02:37 -0800 (PST) (envelope-from owner-freebsd-current@FreeBSD.ORG) Received: from phoenix (phoenix.aye.net [206.185.8.134]) by hub.freebsd.org (8.8.8/8.8.8) with SMTP id VAA13630 for ; Thu, 19 Nov 1998 21:02:35 -0800 (PST) (envelope-from terbart@aye.net) Received: (qmail 22261 invoked by uid 3185); 20 Nov 1998 05:00:28 -0000 Received: from localhost (sendmail-bs@127.0.0.1) by localhost with SMTP; 20 Nov 1998 05:00:28 -0000 Date: Fri, 20 Nov 1998 00:00:28 -0500 (EST) From: Barrett Richardson To: garman@earthling.net cc: current@FreeBSD.ORG Subject: Re: more dying daemons In-Reply-To: <199811190040.QAA12276@hub.freebsd.org> 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 Swap usage at 64% after only hour seems awfully high. I've seem similar behaviour in the world of SCO. Folks just under the threshold of what they can get out of their box with given swap/memory size with 3.2v4.2 are having no problems. They upgrade to Open Server 5, put the same load on it, and the box starts a slow downward spiral and finally becomes unusable. They reboot it, and its ok for a couple of hours. In Greg Lehey's book there is a hint that swap be around 2.5 times what the physical memory is _if_ you have a sizable amount of memory (looks like you have around 128 mb) _and_ expect to be paging. You can use vnconfig to test the beast without reslicing if you have ample disk space. dd if=/dev/zero of=bigswapfile bs=1024 count=262144 vnconfig /dev/vn0c bigswapfile swap swapon /dev/vn0c That will give you another 256 mb of swap (may have to trim to availability of disk space). If your swap usage stabilizes at some point and your box lives, then you've nailed the problem. If problems persist, entertain the idea that something you are using has a mongo memory leak. In the Mem: line in your message memory dedicated to Cache and Bufs looks really low. May want to shove some more RAM in it if the above test works out. It looks like a bonafide memory starvation problem to me. -- Barrett Richardson What is a magician but a practicing theorist? barrett@aye.net On Wed, 18 Nov 1998 garman@earthling.net wrote: > > I'm still having the problems I described earlier. > inetd, samba, and ssh exhibit symptoms where they wither away and die > after a certain amount of time, and this is easily and rapidly > reproducible (in the space of only an hour my freshly-started inetd will > begin to die each time it forks...) > > obviously, this is quite an annoying problem. I have a debugging > version of inetd compiled, along with samba. I'll repost my samba > backtraces if they'll help. inetd gives the "junk pointer: too high to > make sense" messages before it dies. I'm still figuring out how to > debug that problem, as it dies before I can attach to it with gdb. > > what can I do to help diagnose/fix these problems? > > This is on -current as of ~2 days ago. (my machine has been up since > then, and the problems only get worse the longer the machine has been > up) > > my swap usage is: > Device 1024-blocks Used Avail Capacity Type > /dev/wd0s1b 69632 49032 20472 71% Interleaved > /dev/wd2s1b 81920 47568 34224 58% Interleaved > Total 151296 96600 54696 64% > > Mem: 54M Active, 2596K Inact, 16M Wired, 1664K Cache, 2844K Buf, 18M Free > Swap: 148M Total, 94M Used, 53M Free, 64% Inuse > > and i do have one "suggest more swap" message in my logs: > Nov 16 21:54:32 jason /kernel: swap_pager: suggest more swap space: 190 MB > > along with an "interesting" message from sshd (1.2.26): > Nov 18 20:39:32 jason sshd[90364]: fatal: xmalloc: out of memory (allocating 107 > 3675515 bytes) > (oddly enough it isn't even 8pm yet, weird. unrelated problem? :)) > > and many instances of inetd dying on sigsegv's. > > please, i'm willing to give people accounts on this machine, do > debugging, test patches, whatever it takes. just tell me what > data/information you need or point me in the right direction. > > i had none of these problems when i ran 2.2-stable a while back, on the > same hardware. > > thanks, > -- > Jason Garman http://garman.dyn.ml.org/ > Student, University of Maryland garman@earthling.net > And now... did you know that: Whois: JAG145 > "If you fart consistently for 6 years and 9 months, enough gas is > produced to create the energy of an atomic bomb." -- 0xdeadbeef posting > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-current" in the body of the message > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message