From owner-freebsd-hackers@FreeBSD.ORG Sat Sep 6 22:21:56 2008 Return-Path: Delivered-To: freebsd-hackers@FreeBSD.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 0B1C91065673; Sat, 6 Sep 2008 22:21:56 +0000 (UTC) (envelope-from lolo@agneau.org) Received: from bergerie.agneau.org (bergerie.agneau.org [88.173.248.15]) by mx1.freebsd.org (Postfix) with ESMTP id BF8478FC14; Sat, 6 Sep 2008 22:21:55 +0000 (UTC) (envelope-from lolo@agneau.org) Received: by bergerie.agneau.org (Postfix, from userid 500) id 90C761707CE; Sun, 7 Sep 2008 00:04:48 +0200 (CEST) Date: Sun, 7 Sep 2008 00:04:48 +0200 From: Laurent Frigault To: freebsd-stable@freebsd.org, freebsd-hackers@FreeBSD.org Message-ID: <20080906220448.GA78044@obelix.bergerie.agneau.org> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Disposition: inline User-Agent: Mutt/1.4.2.3i X-Powered-By: UUCP Cc: Subject: how to compute vm.pmap.pv_entry_count from procstat ? X-BeenThere: freebsd-hackers@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Technical Discussions relating to FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 06 Sep 2008 22:21:56 -0000 Hi, On a 7.0-STABLE i386 web server, I got kernel messages reporting problems with PV entries: Approaching the limit on PV entries, consider increasing either the vm.pmap.shpgperproc or the vm.pmap.pv_entry_max tunable. This server (DELL poweredge R200 quad core 4G RAM) is having random hang every few weeks (need power off/on cycle, IPMI not responding any more with those hangs) , but I don't know if the hangs are related to a PV entries problem. Before increasing blindly vm.pmap.shpgperproc , I would like to evaluate a good value for vm.pmap.shpgperproc (default 200) because I remember reading somewhere that increasing too much shpgperproc could result in panic at boot time or later. If I guess correctly (I did not find any understandable by me documentation on vm.pmap.*) vm.pmap.pv_entry_count is the value of currently used PV entries . My idea was to use procstat -av output to compute some statistics about the number of PV entries needed by various kind of process (apache, ...) It should be possible to compute a good evaluation of vm.pmap.pv_entry_count by adding some combination of RES,PRES, SHD columns. When I add RES,PRES, SHD values, the result is bigger than pv_entry_count. What is the formula to retrieve pv_entry_count from procstat output or in other words, how to compute the number of PV entries used by a process ? Regards, -- Laurent Frigault |