In regards to troubleshoot a misbehaving systems sar is still the first tool of choice, but it records data only at system level. With cgroup v2 around for several years and in some distros already (or very soon) default, we have a load of system data already collected by the kernel. Thanks to systemd all services and user processes are neatly grouped into a cgroup hierarchy. This allows allows a more granular view on resource consumption.

This project provides a small collector logging all of these data to disk and a tool to filter those logs to provide a CSV for further analysis or graph printing.

https://github.com/scmschmidt/cgar