From owner-freebsd-hackers@freebsd.org Mon Feb 6 13:05:25 2017 Return-Path: Delivered-To: freebsd-hackers@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id E4342CCFA9F for ; Mon, 6 Feb 2017 13:05:25 +0000 (UTC) (envelope-from slw@zxy.spb.ru) Received: from zxy.spb.ru (zxy.spb.ru [195.70.199.98]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id A720BC5A for ; Mon, 6 Feb 2017 13:05:25 +0000 (UTC) (envelope-from slw@zxy.spb.ru) Received: from slw by zxy.spb.ru with local (Exim 4.86 (FreeBSD)) (envelope-from ) id 1caiyv-000FGn-Et; Mon, 06 Feb 2017 16:05:21 +0300 Date: Mon, 6 Feb 2017 16:05:21 +0300 From: Slawa Olhovchenkov To: Fabian Keil Cc: freebsd-hackers@freebsd.org Subject: Re: Inactive memory Message-ID: <20170206130521.GE5366@zxy.spb.ru> References: <20170206102159.GD5366@zxy.spb.ru> <20170206131711.76fe19ed@fabiankeil.de> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20170206131711.76fe19ed@fabiankeil.de> User-Agent: Mutt/1.5.24 (2015-08-30) X-SA-Exim-Connect-IP: X-SA-Exim-Mail-From: slw@zxy.spb.ru X-SA-Exim-Scanned: No (on zxy.spb.ru); SAEximRunCond expanded to false X-BeenThere: freebsd-hackers@freebsd.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: Technical Discussions relating to FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 06 Feb 2017 13:05:26 -0000 On Mon, Feb 06, 2017 at 01:17:11PM +0100, Fabian Keil wrote: > Slawa Olhovchenkov wrote: > > > What is 'Inactive' memory? > > From sys/vm/vm_page.h: > /* > * Each pageable resident page falls into one of four lists: > * > * free > * Available for allocation now. > * > * cache > * Almost available for allocation. Still associated with > * an object, but clean and immediately freeable. > * > * The following lists are LRU sorted: > * > * inactive > * Low activity, candidates for reclamation. > * This is the list of pages that should be > * paged out next. > * > * active > * Pages that are "active" i.e. they have been > * recently referenced. > * > */ > > A case-insensitive grep for "inactive" in the directory > will result in a couple of additional relevant comments > and implementation details. > > > Can I let is definitly availably for allocation (after some time)? > > Under some load I am got about 30GB of Inactive memory and I am > > prefer migrate this memory to ARC cache. > > > > Can I have assurance about freeing this memory under pressure? > > While inactive memory is supposed to be freed under memory > pressure, in vanilla FreeBSD the ARC may shrink to relieve > memory pressure first. In extreme case the ARC may shrink > to its limit while the inactive memory continues to grow. > > In ElectroBSD, this patch is supposed to address this: > https://www.fabiankeil.de/sourcecode/electrobsd/ElectroBSD-r312620-6cfa243f1516/0239-sys-vm-Limit-the-inactive-pages-more-aggressively.diff > You could give it a try to see if it helps for your workload. I am afraid this patch. > If you haven't already, you also may want to look at: > https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=187594 > Among other things, it contains a bunch of (conflicting) > theories about how the system behaves under memory pressure. I am see some reports (in wild internet) about count as 'inactive' used memory in tmpfs. I am not sure about conversion this pages to free and ack this question. Thanks for URLs, I am try different patch for ARC.