Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932427AbdIFOKj (ORCPT ); Wed, 6 Sep 2017 10:10:39 -0400 Received: from mx2.suse.de ([195.135.220.15]:56728 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S932306AbdIFOKi (ORCPT ); Wed, 6 Sep 2017 10:10:38 -0400 Date: Wed, 6 Sep 2017 16:10:34 +0200 From: Michal Hocko To: Roman Gushchin Cc: linux-mm@kvack.org, Vladimir Davydov , Johannes Weiner , Tetsuo Handa , David Rientjes , Andrew Morton , Tejun Heo , kernel-team@fb.com, cgroups@vger.kernel.org, linux-doc@vger.kernel.org, linux-kernel@vger.kernel.org Subject: Re: [v7 2/5] mm, oom: cgroup-aware OOM killer Message-ID: <20170906141034.aw2nzl577m5tt6om@dhcp22.suse.cz> References: <20170904142108.7165-1-guro@fb.com> <20170904142108.7165-3-guro@fb.com> <20170905145700.fd7jjd37xf4tb55h@dhcp22.suse.cz> <20170905202357.GA10535@castle.DHCP.thefacebook.com> <20170906083158.gvqx6pekrsy2ya47@dhcp22.suse.cz> <20170906125750.GB12904@castle> <20170906132249.c2llo5zyrzgviqzc@dhcp22.suse.cz> <20170906134142.GA15796@castle.DHCP.thefacebook.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20170906134142.GA15796@castle.DHCP.thefacebook.com> User-Agent: NeoMutt/20170609 (1.8.3) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 760 Lines: 18 On Wed 06-09-17 14:41:42, Roman Gushchin wrote: [...] > Although, I don't think the whole thing is useful without any way > to adjust the memcg selection, so we can't postpone if for too long. > Anyway, if you think it's a way to go forward, let's do it. I am not really sure we are in a rush here. The whole oom_score_adj fiasco has showed that most users tend to only care "to never kill this and that". A better fine tuned oom control sounds useful at first but apart from very special usecases turns out very impractical to set up. At least that is my experience. There are special cases of course but we should target general use first. Kill the whole memcg is a really useful feature on its own for proper container cleanup. -- Michal Hocko SUSE Labs