Hello All,

We were under the assumption that Cgroups memory controller, limits memory based on just the RSS size of the process.

However, i see that the process is getting killed even when the RSS size is very low compared to the limit in bytes.

For example:

cat memory.stat

cache 85880266752 rss 2441216 rss_huge 0 mapped_file 8192 swap 0 pgpgin 29265847812 pgpgout 29244880354 pgfault 48530374728 pgmajfault 20 inactive_anon 0 active_anon 2441216 inactive_file 85879496704 active_file 770048 unevictable 0 hierarchical_memory_limit 85899345920 hierarchical_memsw_limit 85899345920 total_cache 85880266752 total_rss 2441216 total_rss_huge 0 total_mapped_file 8192 total_swap 0 total_pgpgin 29265847812 total_pgpgout 29244880354 total_pgfault 48530374728 total_pgmajfault 20 total_inactive_anon 0 total_active_anon 2441216 total_inactive_file 85879496704 total_active_file 770048 total_unevictable 0

cat memory.usage_in_bytes 85885800448

cat memory.limit_in_bytes 85899345920

What is confusing is that , why is memory.usage_in_bytes being shown as 85885800448, when the RSS is just 2441216?

Is it true that memory.usage_in_bytes/memory.limit_in_bytes takes files cache into consideration when calculating the usage?

Regards, Lohit