Received: by 2002:ac0:aed5:0:0:0:0:0 with SMTP id t21csp707643imb; Fri, 1 Mar 2019 11:47:11 -0800 (PST) X-Google-Smtp-Source: AHgI3IZjVW7LXz1IjNACMKhtT4GOZmlfa5msEJU28Zsqf+y6Zl3gqACbckj0qsrE72JnfMerG3F2 X-Received: by 2002:aa7:9102:: with SMTP id 2mr7175609pfh.179.1551469631553; Fri, 01 Mar 2019 11:47:11 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1551469631; cv=none; d=google.com; s=arc-20160816; b=bfS95PMN5xu9cJLsdVzY5S1NdXF8njHqTEftD1ZqU+cm4TN5FokxyhJBoUTIru4Dvv c44Pw1agaub62rClDGz3/GzgQ9bF43JA9TbJZlkI0rtG86iovbeKYUwW0BS0l7k/oVVM xtE8vJ3UNbw2/hH8KQwln2w/R5SwywCYSXLWexFNUeYiKOwJ4qqN8uPkgjo661nPiWp5 E/DNz9zri5THlaq2Cgm+Uhh72ce3UzbbEEMatNQitEq0381G7NFk63s2GEuWOYUx6mID W5rxSnYge92ScIIPirpXD+FRQolzOJWK0z4WPmGWsIVsYea8PAicEX2nFmlCznpdWdRD dGvg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:sender:content-transfer-encoding :content-language:in-reply-to:mime-version:user-agent:date :message-id:from:references:cc:to:subject; bh=/9pYijVspOTO1OQlh3GdyD/R1MCNfqHdIZs0yu+GNAo=; b=VxqludDy+ko6wsADGJ+hBhFSwXxa9ItXMsts8LTiWj7Fax2SvpK4v1MPiPgDxGRuMr QR+a+9nQIM5gm9A+vx1GlY9aUAjZW06MyXt2F7xe5cMUWRmtv1SJFRSs22eEdOxP7kdJ ccavVzuTxpFm8nG/6+0H4XYVKTBRbnwT8OEupaLJGmb1RTnFp9xNwciTIeABn8R5iUql EYqnMFneNHJPwny6sb87HkX7fFOQuJLybLyG/Mw+5CrzrmxO+NdmduxON20phrCPEJAZ LERswTttKEyKPLNDaE1v2KWWULI77VpAcSgv91yhPfO4pTCa3JniU5S0kTxbd188C7yP p16Q== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=virtuozzo.com Return-Path: Received: from vger.kernel.org (vger.kernel.org. [209.132.180.67]) by mx.google.com with ESMTP id s139si13903190pfs.56.2019.03.01.11.46.55; Fri, 01 Mar 2019 11:47:11 -0800 (PST) Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67; Authentication-Results: mx.google.com; spf=pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=virtuozzo.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726002AbfCATqf (ORCPT + 99 others); Fri, 1 Mar 2019 14:46:35 -0500 Received: from relay.sw.ru ([185.231.240.75]:59556 "EHLO relay.sw.ru" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1725862AbfCATqf (ORCPT ); Fri, 1 Mar 2019 14:46:35 -0500 Received: from [172.16.25.12] by relay.sw.ru with esmtp (Exim 4.91) (envelope-from ) id 1gzo6q-0007Uo-RS; Fri, 01 Mar 2019 22:46:17 +0300 Subject: Re: [PATCH RFC] mm/vmscan: try to protect active working set of cgroup from reclaim. To: Johannes Weiner Cc: Andrew Morton , linux-mm@kvack.org, linux-kernel@vger.kernel.org, Michal Hocko , Vlastimil Babka , Rik van Riel , Mel Gorman , Roman Gushchin , Shakeel Butt References: <20190222175825.18657-1-aryabinin@virtuozzo.com> <20190222191552.GA15922@cmpxchg.org> <7c915942-6f52-e7a4-b879-e4c99dd65968@virtuozzo.com> <20190301174907.GA2375@cmpxchg.org> From: Andrey Ryabinin Message-ID: <51ac7aaa-6890-c674-854d-1e2d132b83f9@virtuozzo.com> Date: Fri, 1 Mar 2019 22:46:34 +0300 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.5.2 MIME-Version: 1.0 In-Reply-To: <20190301174907.GA2375@cmpxchg.org> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 3/1/19 8:49 PM, Johannes Weiner wrote: > Hello Andrey, > > On Fri, Mar 01, 2019 at 01:38:26PM +0300, Andrey Ryabinin wrote: >> On 2/26/19 3:50 PM, Andrey Ryabinin wrote: >>> On 2/22/19 10:15 PM, Johannes Weiner wrote: >>>> On Fri, Feb 22, 2019 at 08:58:25PM +0300, Andrey Ryabinin wrote: >>>>> In a presence of more than 1 memory cgroup in the system our reclaim >>>>> logic is just suck. When we hit memory limit (global or a limit on >>>>> cgroup with subgroups) we reclaim some memory from all cgroups. >>>>> This is sucks because, the cgroup that allocates more often always wins. >>>>> E.g. job that allocates a lot of clean rarely used page cache will push >>>>> out of memory other jobs with active relatively small all in memory >>>>> working set. >>>>> >>>>> To prevent such situations we have memcg controls like low/max, etc which >>>>> are supposed to protect jobs or limit them so they to not hurt others. >>>>> But memory cgroups are very hard to configure right because it requires >>>>> precise knowledge of the workload which may vary during the execution. >>>>> E.g. setting memory limit means that job won't be able to use all memory >>>>> in the system for page cache even if the rest the system is idle. >>>>> Basically our current scheme requires to configure every single cgroup >>>>> in the system. >>>>> >>>>> I think we can do better. The idea proposed by this patch is to reclaim >>>>> only inactive pages and only from cgroups that have big >>>>> (!inactive_is_low()) inactive list. And go back to shrinking active lists >>>>> only if all inactive lists are low. >>>> >>>> Yes, you are absolutely right. >>>> >>>> We shouldn't go after active pages as long as there are plenty of >>>> inactive pages around. That's the global reclaim policy, and we >>>> currently fail to translate that well to cgrouped systems. >>>> >>>> Setting group protections or limits would work around this problem, >>>> but they're kind of a red herring. We shouldn't ever allow use-once >>>> streams to push out hot workingsets, that's a bug. >>>> >>>>> @@ -2489,6 +2491,10 @@ static void get_scan_count(struct lruvec *lruvec, struct mem_cgroup *memcg, >>>>> >>>>> scan >>= sc->priority; >>>>> >>>>> + if (!sc->may_shrink_active && inactive_list_is_low(lruvec, >>>>> + file, memcg, sc, false)) >>>>> + scan = 0; >>>>> + >>>>> /* >>>>> * If the cgroup's already been deleted, make sure to >>>>> * scrape out the remaining cache. >>>>> @@ -2733,6 +2739,7 @@ static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc) >>>>> struct reclaim_state *reclaim_state = current->reclaim_state; >>>>> unsigned long nr_reclaimed, nr_scanned; >>>>> bool reclaimable = false; >>>>> + bool retry; >>>>> >>>>> do { >>>>> struct mem_cgroup *root = sc->target_mem_cgroup; >>>>> @@ -2742,6 +2749,8 @@ static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc) >>>>> }; >>>>> struct mem_cgroup *memcg; >>>>> >>>>> + retry = false; >>>>> + >>>>> memset(&sc->nr, 0, sizeof(sc->nr)); >>>>> >>>>> nr_reclaimed = sc->nr_reclaimed; >>>>> @@ -2813,6 +2822,13 @@ static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc) >>>>> } >>>>> } while ((memcg = mem_cgroup_iter(root, memcg, &reclaim))); >>>>> >>>>> + if ((sc->nr_scanned - nr_scanned) == 0 && >>>>> + !sc->may_shrink_active) { >>>>> + sc->may_shrink_active = 1; >>>>> + retry = true; >>>>> + continue; >>>>> + } >>>> >>>> Using !scanned as the gate could be a problem. There might be a cgroup >>>> that has inactive pages on the local level, but when viewed from the >>>> system level the total inactive pages in the system might still be low >>>> compared to active ones. In that case we should go after active pages. >>>> >>>> Basically, during global reclaim, the answer for whether active pages >>>> should be scanned or not should be the same regardless of whether the >>>> memory is all global or whether it's spread out between cgroups. >>>> >>>> The reason this isn't the case is because we're checking the ratio at >>>> the lruvec level - which is the highest level (and identical to the >>>> node counters) when memory is global, but it's at the lowest level >>>> when memory is cgrouped. >>>> >>>> So IMO what we should do is: >>>> >>>> - At the beginning of global reclaim, use node_page_state() to compare >>>> the INACTIVE_FILE:ACTIVE_FILE ratio and then decide whether reclaim >>>> can go after active pages or not. Regardless of what the ratio is in >>>> individual lruvecs. >>>> >>>> - And likewise at the beginning of cgroup limit reclaim, walk the >>>> subtree starting at sc->target_mem_cgroup, sum up the INACTIVE_FILE >>>> and ACTIVE_FILE counters, and make inactive_is_low() decision on >>>> those sums. >>>> >>> >>> Sounds reasonable. >>> >> >> On the second thought it seems to be better to keep the decision on lru level. >> There are couple reasons for this: >> >> 1) Using bare node_page_state() (or sc->targe_mem_cgroup's total_[in]active counters) would be wrong. >> Because some cgroups might have protection set (memory.low) and we must take it into account. Also different >> cgroups have different available swap space/memory.swappiness and it must be taken into account as well to. >> >> So it has to be yet another full memcg-tree iteration. > > It should be possible to take that into account on the first iteration > and adjust the inactive/active counters in proportion to how much of > the cgroup's total memory is exempt by memory.low or min, right? > Should be possible, more complexity though to this subtle code. >> 2) Let's consider simple case. Two cgroups, one with big 'active' set of pages the other allocates one-time used pages. >> So the total inactive is low, thus checking inactive ratio on higher level will result in reclaiming pages. >> While with check on lru-level only inactive will be reclaimed. > > It's the other way around. Let's say you have two cgroups, A and B: > > A: 500M inactive 10G active -> inactive is low > B: 10G inactive 500M active -> inactive is NOT low > ---------------------------------------------------------- > global: 10.5G inactive 10.5G active -> inactive is NOT low > > Checking locally will scan active pages from A. No, checking locally will not scan active from A. Initial state of sc->may_shrink_active = 0, so A group will be skipped completely, and will reclaim from B. Since overall reclaim was successful, sc->may_shrink_active remain 0 and A will be protected as long as B supply enough inactive pages. > Checking globally will > not, because there is plenty of use-once pages from B. > That is correct. So in this example global vs local check will not make a difference. > So if you check globally, without any protection, A and B compete > evenly during global reclaim. Under the same reclaim pressure, A has > managed to activate most of its pages whereas B has not. That means A > is hotter and B provides the better reclaim candidates. > > If you apply this decision locally, on the other hand, you are no > longer aging the groups at the same rate. And then the LRU orders > between groups will no longer be comparable, and you won't be > reclaiming the coldest memory in the system anymore. > I really don't see any how global check will make any difference in this example. In both cases, we reclaim only from B and don't touch A. And this what we actually want.