Docker – Analyze memory and cpu peaks of docker container

Well, first you misspelled cgroup :)

After poking around, I found what you were looking for:

# cat /sys/fs/cgroup/memory/system.slice/docker-88018f8043d00669bbf865855ebc8a6ccc93a04ce588111e01d4e63739250340.scope/memory.stat 
cache 5251072
rss 380928
rss_huge 0
mapped_file 2662400
writeback 0
swap 0
pgpgin 1570
pgpgout 195
pgfault 344
pgmajfault 40
inactive_anon 278528
active_anon 102400
inactive_file 5083136
active_file 167936
unevictable 0
hierarchical_memory_limit 18446744073709551615
hierarchical_memsw_limit 18446744073709551615
total_cache 5251072
total_rss 380928
total_rss_huge 0
total_mapped_file 2662400
total_writeback 0
total_swap 0
total_pgpgin 1570
total_pgpgout 195
total_pgfault 344
total_pgmajfault 40
total_inactive_anon 278528
total_active_anon 102400
total_inactive_file 5083136
total_active_file 167936
total_unevictable 0
recent_rotated_anon 248
recent_rotated_file 41
recent_scanned_anon 248
recent_scanned_file 1495

There is no Linux running inside the container, as Linux is the kernel and you share it with the host. Your apps (like free and top) read system info from /proc and thus get details about the host as lxc does not fake the limited resources by default (in contrast to eg OpenVZ). This is ok unless your app behaves differently depending on the ram/swap/cpu numbers. If your app tries to allocate more memory than available to the container it will be OOM-killed as any other app in a non-container environment.

If you want to see the limited resources inside the container, do the following on the host:

apt-get install lxcfs

Afterwards restart the container. LXCFS will emulate a few files in /proc of the container and apps will see the limited resources (cpu, ram, swap) properly.

More info on the lxcfs homepage: https://linuxcontainers.org/lxcfs/