Zabbix-agent that is running inside the container (LXC/Docker/Podman) with resource limits (using cgroups) reports memory/cpuload values of the host system, not of the container.
Users may want to choose which values they want to monitor: host's values or container's.
Currently user is forced to use custom UserParameters to monitor container's metrics like memory, cpu load...
Container's metrics (that respect its limits) can be fetched from cgroup sysfs.
As different containerization systems are using different cgroups versions (v1/v2), user should be able to accordingly choose item keys for its version.
New item keys can be implemented to fetch container's metrics using cgroups' sysfs.
- caused by
ZBX-12164 zabbix-agent get wrong memory values in Linux Containers