From owner-freebsd-stable Tue Aug 21 12: 7:18 2001 Delivered-To: freebsd-stable@freebsd.org Received: from earth.backplane.com (earth-nat-cw.backplane.com [208.161.114.67]) by hub.freebsd.org (Postfix) with ESMTP id 42D8737B405 for ; Tue, 21 Aug 2001 12:07:15 -0700 (PDT) (envelope-from dillon@earth.backplane.com) Received: (from dillon@localhost) by earth.backplane.com (8.11.4/8.11.2) id f7LJ7DK65902; Tue, 21 Aug 2001 12:07:13 -0700 (PDT) (envelope-from dillon) Date: Tue, 21 Aug 2001 12:07:13 -0700 (PDT) From: Matt Dillon Message-Id: <200108211907.f7LJ7DK65902@earth.backplane.com> To: Steve Shorter Cc: freebsd-stable@FreeBSD.ORG Subject: Re: VM exhaustion and kernel hangs References: <20010821115724.A2562@nomad.lets.net> <200108211820.f7LIKZf65278@earth.backplane.com> <20010821144406.A2733@nomad.lets.net> Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG :> :> You might also want to try to figure out what is eating up the memory :> (or, more specifically, dirtying memory since clean file-backed pages :> would not lead to this situation). :> : : Can this situation be created if a web server is serving pages :that are being modified by another process, if they are both accessing the :fs over NFS? : : thanx - steve No. What creates this situation tends to be malloc()'d data (which is not backed by a file), or MAP_PRIVATE file memory mappings which are modified. If you are running a large number of processes the situation can also occur simply due to the modified data/bss areas of the programs in question, and do to the modified pages of 'glue' used to tie the shared libraries together. In a turnkey environment the shared library glue can be gotten rid of by compiling the programs -static, leaving only data/bss and malloc() space left. What you need to do is determine exactly what is causing the system to run out of memory. For example, if you are running apache you may need to limit the number of daemons it forks off or limit the number of requests each one takes before Apache respawns it. It is simply not possible to give any on-the-mark recommendations until you know exactly what is eating the memory. Also, just to throw this out... swap over NFS is not actually all that bad a thing to do, especially with FreeBSD-4.x and modern switched 100BaseTX ethernet links. -Matt To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message