The docker_cpu and docker_memory munin plugins monitor container CPU and memory usage over time, respectively.

These plugins read the metrics exposed by control groups (cgroups) on a pseudo-filesystem. We can find the memory metrics for a Docker container under the /sys/fs/cgroup/memory/docker/container_id/ directory and the cpu metrics under the /sys/fs/cgroup/cpuacct/docker/container_id/ directory.

This plugin uses the pseudo-file memory.usage_in_bytes to get the memory metrics. The munin data type must be defined as "gauge".

This plugin uses the pseudo-file cpuacct.usage and cpuacct.usage_percpu to get the CPU metrics. The pseudo-file cpuacct.usage reports the total CPU time (in nanoseconds) consumed by a container and the pseudo-file cpuacct.usage_percpu reports the CPU time (in nanoseconds) consumed on each CPU. This plugin gets the total number of cores available to scale the munin graph.

The munin data type must be defined as "derive" because the CPU usage is a cumulative metric. A derivative shows the difference between data points, and allows the monitoring of CPU usage during each period. The munin sample interval is fixed at 300 seconds.

1. Install the munin-node in the docker server. For more information check the following URL: munin guide.
2. Copy the docker_cpu and docker_memory plugins to the munin plugin folder: /usr/share/munin/plugins.
3. Set plugin permissions: chmod 755 docker_*
4. Create symbolic links.

cd /etc/munin/plugins/
ln -s /usr/share/munin/plugins/docker_memory docker_memory
ln -s /usr/share/munin/plugins/docker_cpu docker_cpu

5. Give root privilege to the docker plugin. Root privilege is required for munin to execute the docker command.
6. Create a new file named "docker" inside the folder /etc/munin/plugin-conf.d/
7. Insert the following content in the new docker file:

[docker_*]
user root

6. Restart the munin-node service.