3.0.1 - steady increase of host's RAM consumption

lxc host - ubuntu bionic 4.17.4 (being a VPS on a KVM 7_4.1 based hypervisor)
unpriviliged lxc containers - ubuntu, gentoo, archlinux

Noticed that with lxc containers running a steady increase in the amount of the host’s RAM consumption (rate ~ 1 MB per minute)
Thus went to monitoring the RAM consumption from within each container but those figures kept hovering steady respectively, thus it seems unlikely that being the cause.
Also tested with different single containers and combination of various containers. Only with no container running the lxc-host’s RAM consumption is not increasing.

Could there be perhaps a correlation to the observed steady increase of MINFLT (Total number of minor faults the task has made per second, those which have not required loading a memory page from disk.) for process /usr/lib/systemd/systemd-journald and /sbin/init belonging to CGROUP /lxc/container?

It seems that the more containers are running the faster the increase of RAM consumption over the same amout of time compared to only a single container running.

The way it is the host runs out of RAM eventually.


container config
lxc.mount.auto = proc:mixed sys:mixed

I suspect lxcfs being the culprit somehow. The degradation rate of available memory with ~25 MB/hrs is alarming.

output from head -2 /proc/slabinfo; cat /proc/slabinfo | egrep dentry\|inode is indicating a steady increase of active_objs and num_objs for

proc_inode_cache
inode_cache
dentry 

Running sync && echo 3 > /proc/sys/vm/drop_caches provides relief but not a solution.