Re: [PATCH 1/1] mm: vmscan: Reduce throttling due to a failure to make progress

From: Alexey Avramov <hakavlad@inbox.lv>
To: Mel Gorman <mgorman@techsingularity.net>
Cc: Andrew Morton <akpm@linux-foundation.org>,
	Michal Hocko <mhocko@suse.com>, Vlastimil Babka <vbabka@suse.cz>,
	Rik van Riel <riel@surriel.com>, Mike Galbraith <efault@gmx.de>,
	Darrick Wong <djwong@kernel.org>,
	regressions@lists.linux.dev,
	Linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	Linux-MM <linux-mm@kvack.org>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 1/1] mm: vmscan: Reduce throttling due to a failure to make progress
Date: Wed, 1 Dec 2021 01:03:48 +0900	[thread overview]
Message-ID: <20211201010348.31e99637@mail.inbox.lv> (raw)
In-Reply-To: <20211129150117.GO3366@techsingularity.net>

I tested this [1] patch on top of 5.16-rc2. It's the same test with 10 tails.

- with noswap

Summary:

2021-11-30 23:32:36,890: Stall times for the last 548.6s:
2021-11-30 23:32:36,890: -----------
2021-11-30 23:32:36,891: some cpu     3.7s, avg 0.7%
2021-11-30 23:32:36,891: -----------
2021-11-30 23:32:36,891: some io      187.6s, avg 34.2%
2021-11-30 23:32:36,891: full io      178.3s, avg 32.5%
2021-11-30 23:32:36,891: -----------
2021-11-30 23:32:36,892: some memory  392.2s, avg 71.5%
2021-11-30 23:32:36,892: full memory  390.7s, avg 71.2%

full psi:
https://raw.githubusercontent.com/hakavlad/cache-tests/main/516-reclaim-throttle/516-rc2/patch5/noswap/psi

mem:
https://raw.githubusercontent.com/hakavlad/cache-tests/main/516-reclaim-throttle/516-rc2/patch5/noswap/mem

- with swappiness=0

Summary:

2021-11-30 23:51:48,969: Stall times for the last 919.4s:
2021-11-30 23:51:48,969: -----------
2021-11-30 23:51:48,969: some cpu     5.5s, avg 0.6%
2021-11-30 23:51:48,970: -----------
2021-11-30 23:51:48,970: some io      240.4s, avg 26.2%
2021-11-30 23:51:48,970: full io      230.6s, avg 25.1%
2021-11-30 23:51:48,970: -----------
2021-11-30 23:51:48,970: some memory  806.1s, avg 87.7%
2021-11-30 23:51:48,971: full memory  800.5s, avg 87.1%

psi log:
https://raw.githubusercontent.com/hakavlad/cache-tests/main/516-reclaim-throttle/516-rc2/patch5/swappiness0/psi

mem log:
https://raw.githubusercontent.com/hakavlad/cache-tests/main/516-reclaim-throttle/516-rc2/patch5/swappiness0/mem

In some cases stalls was very short, but in many cases stalls was long. 
The result is still not good enough.

offtop
======

The same test with the patch [1] on top of 5.16-rc2 + le9 patch [2] 
with vm.clean_min_kbytes=99000.

- with noswap

Summary:

2021-11-30 23:59:32,209: Stall times for the last 73.1s:
2021-11-30 23:59:32,209: -----------
2021-11-30 23:59:32,209: some cpu     0.4s, avg 0.5%
2021-11-30 23:59:32,209: -----------
2021-11-30 23:59:32,210: some io      5.8s, avg 8.0%
2021-11-30 23:59:32,210: full io      5.3s, avg 7.3%
2021-11-30 23:59:32,210: -----------
2021-11-30 23:59:32,210: some memory  3.3s, avg 4.5%
2021-11-30 23:59:32,210: full memory  3.1s, avg 4.2%

This is just an example of what a result close to the expected 
result might be (especially note io pressure values).

full psi:
https://raw.githubusercontent.com/hakavlad/cache-tests/main/516-reclaim-throttle/516-rc2/patch5/noswap_le9_min99k/psi

mem:
https://raw.githubusercontent.com/hakavlad/cache-tests/main/516-reclaim-throttle/516-rc2/patch5/noswap_le9_min99k/mem

[1] https://lore.kernel.org/lkml/20211129150117.GO3366@techsingularity.net/
[2] https://lore.kernel.org/all/20211130201652.2218636d@mail.inbox.lv/