Re: [PATCH] mm: vmscan: do not share cgroup iteration between reclaimers

[Michal Hocko]

On Mon 12-08-19 15:23:16, Johannes Weiner wrote:
> One of our services observed a high rate of cgroup OOM kills in the
> presence of large amounts of clean cache. Debugging showed that the
> culprit is the shared cgroup iteration in page reclaim.
> 
> Under high allocation concurrency, multiple threads enter reclaim at
> the same time. Fearing overreclaim when we first switched from the
> single global LRU to cgrouped LRU lists, we introduced a shared
> iteration state for reclaim invocations - whether 1 or 20 reclaimers
> are active concurrently, we only walk the cgroup tree once: the 1st
> reclaimer reclaims the first cgroup, the second the second one etc.
> With more reclaimers than cgroups, we start another walk from the top.
> 
> This sounded reasonable at the time, but the problem is that reclaim
> concurrency doesn't scale with allocation concurrency. As reclaim
> concurrency increases, the amount of memory individual reclaimers get
> to scan gets smaller and smaller. Individual reclaimers may only see
> one cgroup per cycle, and that may not have much reclaimable
> memory. We see individual reclaimers declare OOM when there is plenty
> of reclaimable memory available in cgroups they didn't visit.
> 
> This patch does away with the shared iterator, and every reclaimer is
> allowed to scan the full cgroup tree and see all of reclaimable
> memory, just like it would on a non-cgrouped system. This way, when
> OOM is declared, we know that the reclaimer actually had a chance.
> 
> To still maintain fairness in reclaim pressure, disallow cgroup
> reclaim from bailing out of the tree walk early. Kswapd and regular
> direct reclaim already don't bail, so it's not clear why limit reclaim
> would have to, especially since it only walks subtrees to begin with.

The code does bail out on any direct reclaim - be it limit or page
allocator triggered. Check the !current_is_kswapd part of the condition.

> This change completely eliminates the OOM kills on our service, while
> showing no signs of overreclaim - no increased scan rates, %sys time,
> or abrupt free memory spikes. I tested across 100 machines that have
> 64G of RAM and host about 300 cgroups each.

What is the usual direct reclaim involvement on those machines?

> [ It's possible overreclaim never was a *practical* issue to begin
>   with - it was simply a concern we had on the mailing lists at the
>   time, with no real data to back it up. But we have also added more
>   bail-out conditions deeper inside reclaim (e.g. the proportional
>   exit in shrink_node_memcg) since. Regardless, now we have data that
>   suggests full walks are more reliable and scale just fine. ]

I do not see how shrink_node_memcg bail out helps here. We do scan up-to
SWAP_CLUSTER_MAX pages for each LRU at least once. So we are getting to
nr_memcgs_with_pages multiplier with the patch applied in the worst case.

How much that matters is another question and it depends on the
number of cgroups and the rate the direct reclaim happens. I do not
remember exact numbers but even walking a very large memcg tree was
noticeable.

For the over reclaim part SWAP_CLUSTER_MAX is a relatively small number
so even hundreds of memcgs on a "reasonably" sized system shouldn't be
really observable (we are talking about 7MB per reclaim per reclaimer on
1k memcgs with pages). This would get worse with many reclaimers. Maybe
we will need something like the regular direct reclaim throttling of
mmemcg limit reclaim as well in the future.

That being said, I do agree that the oom side of the coin is causing
real troubles and it is a real problem to be addressed first. Especially with
cgroup v2 where we have likely more memcgs without any pages because
inner nodes do not have any tasks and direct charges which makes some
reclaimers hit memcgs without pages more likely.

Let's see whether we see regression due to over-reclaim. 

> Signed-off-by: Johannes Weiner <hannes@xxxxxxxxxxx>

With the direct reclaim bail out reference fixed - unless I am wrong
there of course

Acked-by: Michal Hocko <mhocko@xxxxxxxx>

It is sad to see this piece of fun not being used after that many years
of bugs here and there and all the lockless fun but this is the life
;)

> ---
>  mm/vmscan.c | 22 ++--------------------
>  1 file changed, 2 insertions(+), 20 deletions(-)
> 
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index dbdc46a84f63..b2f10fa49c88 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -2667,10 +2667,6 @@ static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc)
>  
>  	do {
>  		struct mem_cgroup *root = sc->target_mem_cgroup;
> -		struct mem_cgroup_reclaim_cookie reclaim = {
> -			.pgdat = pgdat,
> -			.priority = sc->priority,
> -		};
>  		unsigned long node_lru_pages = 0;
>  		struct mem_cgroup *memcg;
>  
> @@ -2679,7 +2675,7 @@ static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc)
>  		nr_reclaimed = sc->nr_reclaimed;
>  		nr_scanned = sc->nr_scanned;
>  
> -		memcg = mem_cgroup_iter(root, NULL, &reclaim);
> +		memcg = mem_cgroup_iter(root, NULL, NULL);
>  		do {
>  			unsigned long lru_pages;
>  			unsigned long reclaimed;
> @@ -2724,21 +2720,7 @@ static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc)
>  				   sc->nr_scanned - scanned,
>  				   sc->nr_reclaimed - reclaimed);
>  
> -			/*
> -			 * Kswapd have to scan all memory cgroups to fulfill
> -			 * the overall scan target for the node.
> -			 *
> -			 * Limit reclaim, on the other hand, only cares about
> -			 * nr_to_reclaim pages to be reclaimed and it will
> -			 * retry with decreasing priority if one round over the
> -			 * whole hierarchy is not sufficient.
> -			 */
> -			if (!current_is_kswapd() &&
> -					sc->nr_reclaimed >= sc->nr_to_reclaim) {
> -				mem_cgroup_iter_break(root, memcg);
> -				break;
> -			}
> -		} while ((memcg = mem_cgroup_iter(root, memcg, &reclaim)));
> +		} while ((memcg = mem_cgroup_iter(root, memcg, NULL)));
>  
>  		if (reclaim_state) {
>  			sc->nr_reclaimed += reclaim_state->reclaimed_slab;
> -- 
> 2.22.0

-- 
Michal Hocko
SUSE Labs