On 2018/12/3 PM 7:57, Michal Hocko wrote: > On Mon 03-12-18 16:01:19, Xunlei Pang wrote: >> When memcgs get reclaimed after its usage exceeds min, some >> usages below the min may also be reclaimed in the current >> implementation, the amount is considerably large during kswapd >> reclaim according to my ftrace results. > > And here again. Describe the setup and the behavior please? > step 1 mkdir -p /sys/fs/cgroup/memory/online cd /sys/fs/cgroup/memory/online echo 512M > memory.max echo 409600000 > memory.min echo $$ > tasks dd if=/dev/sda of=/dev/null while true; do sleep 1; cat memory.current ; cat memory.min; done step 2 create global memory pressure by allocating annoymous and cached pages to constantly trigger kswap: dd if=/dev/sdb of=/dev/null step 3 Then observe "online" groups, hundreds of kbytes a little over memory.min can cause tens of MiB to be reclaimed by kswapd. Here is one of test results I got: cat memory.current; cat memory.min; echo; 409485312 // current 409600000 // min 385052672 // See current got over reclaimed for 23MB 409600000 // min Its corresponding ftrace output I monitored: kswapd_0-281 [000] .... 304.706632: shrink_node_memcg: min_excess=24, nr_reclaimed=6013, sc->nr_to_reclaim=1499997, exceeds 5989pages
