Please read memory usage data from cgroups

[baryluk]

Similar to #1 but memory related stats:

# cd /sys/fs/cgroup/memory/system.slice/system-postgresql.slice
/sys/fs/cgroup/memory/system.slice/system-postgresql.slice# grep . *
cgroup.clone_children:0
grep: cgroup.event_control: Invalid argument
memory.failcnt:0
grep: memory.force_empty: Invalid argument
memory.kmem.failcnt:0
memory.kmem.limit_in_bytes:9223372036854771712
memory.kmem.max_usage_in_bytes:52396032
memory.kmem.tcp.failcnt:0
memory.kmem.tcp.limit_in_bytes:9223372036854771712
memory.kmem.tcp.max_usage_in_bytes:0
memory.kmem.tcp.usage_in_bytes:0
memory.kmem.usage_in_bytes:17018880
memory.limit_in_bytes:9223372036854771712
memory.max_usage_in_bytes:17903652864
memory.move_charge_at_immigrate:0
memory.numa_stat:total=0 N0=0
memory.numa_stat:file=0 N0=0
memory.numa_stat:anon=0 N0=0
memory.numa_stat:unevictable=0 N0=0
memory.numa_stat:hierarchical_total=226104 N0=226104
memory.numa_stat:hierarchical_file=18420 N0=18420
memory.numa_stat:hierarchical_anon=207684 N0=207684
memory.numa_stat:hierarchical_unevictable=0 N0=0
memory.oom_control:oom_kill_disable 0
memory.oom_control:under_oom 0
memory.oom_control:oom_kill 0
grep: memory.pressure_level: Invalid argument
memory.soft_limit_in_bytes:9223372036854771712
memory.stat:cache 0
memory.stat:rss 0
memory.stat:rss_huge 0
memory.stat:shmem 0
memory.stat:mapped_file 0
memory.stat:dirty 0
memory.stat:writeback 0
memory.stat:pgpgin 0
memory.stat:pgpgout 0
memory.stat:pgfault 0
memory.stat:pgmajfault 0
memory.stat:inactive_anon 0
memory.stat:active_anon 0
memory.stat:inactive_file 0
memory.stat:active_file 0
memory.stat:unevictable 0
memory.stat:hierarchical_memory_limit 9223372036854771712
memory.stat:total_cache 921415680
memory.stat:total_rss 7901184
memory.stat:total_rss_huge 0
memory.stat:total_shmem 846168064
memory.stat:total_mapped_file 135168
memory.stat:total_dirty 811008
memory.stat:total_writeback 0
memory.stat:total_pgpgin 18059580
memory.stat:total_pgpgout 17833660
memory.stat:total_pgfault 17232468
memory.stat:total_pgmajfault 0
memory.stat:total_inactive_anon 27570176
memory.stat:total_active_anon 823103488
memory.stat:total_inactive_file 67342336
memory.stat:total_active_file 8105984
memory.stat:total_unevictable 0
memory.swappiness:60
memory.usage_in_bytes:943161344
memory.use_hierarchy:1
notify_on_release:0
grep: [email protected]: Is a directory
# 

memory.total_rss and memory.usage_in_bytes are probably of biggest interest, but would be great to have almost all of that (especially total_* ones) as metrics.

[povilasv]

Hey, the plan is to figure out what we want from cgroups and what from procfs, but this is looks definitely right way to go.

PRs are welcome!