On Mon, Apr 18, 2011 at 11:42 AM, Michal Hocko wrote: > On Mon 18-04-11 10:01:20, Ying Han wrote: > > On Mon, Apr 18, 2011 at 2:13 AM, Michal Hocko wrote: > [...] > > > I see. I am just concerned whether 3rd level of reclaim is a good idea. > > > We would need to do background reclaim anyway (and to preserve the > > > original semantic it has to be somehow watermark controlled). I am just > > > wondering why we have to implement it separately from kswapd. Cannot we > > > just simply trigger global kswapd which would reclaim all cgroups that > > > are under watermarks? [I am sorry for my ignorance if that is what is > > > implemented in the series - I haven't got to the patches yes] > > > > > > > They are different on per-zone reclaim vs per-memcg reclaim. The first > > one is triggered if the zone is under memory pressure and we need > > to free pages to serve further page allocations. The second one is > > triggered if the memcg is under memory pressure and we need to free > > pages to leave room (limit - usage) for the memcg to grow. > > OK, I see. > > > > > > Both of them are needed and that is how it is implemented on the direct > > reclaim path. The kswapd batches only try to > > smooth out the system and memcg performance by reclaiming pages > proactively. > > It doesn't affecting the functionality. > > I am still wondering, isn't this just a nice to have feature rather > than must to have in order to get rid of the global LRU? The per-memcg kswapd is a must-have, and it is less related to the effort of "get rid of global LRU" than the next patch I am looking at "enhance the soft_limit reclaim". So this is the structure we will end up with background reclaim: 1. per-memcg : this patch 2. global: targeting reclaim by replacing the per-zone to soft_limit reclaim direct reclaim: 1. per-memcg: no change from today 2. global: targeting reclaim by replacing the per-zone to soft_limit reclaim. > Doesn't it make transition more complicated. I have noticed many if-else in > kswapd path to > distinguish per-cgroup from the traditional global background reclaim. > > > [...] > > > > > > > Step1: Create a cgroup with 500M memory_limit. > > > > > > $ mkdir /dev/cgroup/memory/A > > > > > > $ echo 500m >/dev/cgroup/memory/A/memory.limit_in_bytes > > > > > > $ echo $$ >/dev/cgroup/memory/A/tasks > > > > > > > > > > > > Step2: Test and set the wmarks. > > > > > > $ cat /dev/cgroup/memory/A/memory.low_wmark_distance > > > > > > 0 > > > > > > $ cat /dev/cgroup/memory/A/memory.high_wmark_distance > > > > > > 0 > > > > > > > > > > > > > > They are used to tune the high/low_marks based on the hard_limit. We > > > might > > > > need to export that configuration to user admin especially on > machines > > > where > > > > they over-commit by hard_limit. > > > > > > I remember there was some resistance against tuning watermarks > > > separately. > > > > > > > This API is based on KAMEZAWA's request. :) > > This was just as FYI. Watermarks were considered internal thing. So I > wouldn't be surprised if this got somehow controversial. > We went back and forth on how to set the high/low wmarks for different configurations (over-commit or not). So far, by giving the user ability to set the wmarks seems the most feasible way of fullfilling the requriment. > > > > > > > > > > > > $ cat /dev/cgroup/memory/A/memory.reclaim_wmarks > > > > > > low_wmark 524288000 > > > > > > high_wmark 524288000 > > > > > > > > > > > > $ echo 50m >/dev/cgroup/memory/A/memory.high_wmark_distance > > > > > > $ echo 40m >/dev/cgroup/memory/A/memory.low_wmark_distance > > > > > > > > > > > > $ cat /dev/cgroup/memory/A/memory.reclaim_wmarks > > > > > > low_wmark 482344960 > > > > > > high_wmark 471859200 > > > > > > > > > > low_wmark is higher than high_wmark? > > > > > > > > > > > > > hah, it is confusing. I have them documented. Basically, low_wmark > > > > triggers reclaim and high_wmark stop the reclaim. And we have > > > > > > > > high_wmark < usage < low_wmark. > > OK, I see how you calculate those watermarks now but it is really > confusing for those who are used to traditional watermark semantic. > that is true. I adopt the initial comment from Mel where we keep the same logic of triggering and stopping kswapd with low/high_wmarks and also comparing the usage_in_bytes to the wmarks. Either way is confusing and guess we just need to document it well. --Ying -- > Michal Hocko > SUSE Labs > SUSE LINUX s.r.o. > Lihovarska 1060/12 > 190 00 Praha 9 > Czech Republic >