I used to be a huge fan of munin for capturing system utilization metrics and seeing what was going on when things weren't working well. While munin was easy and automatic to set up, it captured data infrequently and put quite a bit of stress on the host.
I've since move to an infrastructure using InfluxDB, Telegraf, and Grafana. That system replaced collectd, which worked and produced high-resolution data, but also put a high load on the hosts.
But I just learned about netdata and installed it on my dev and staging environments and it looks really good. It requires basically no setup to capture useful data, generates great graphs, and looks like it could expand greatly (putting data into Influx or Prometheus, etc). The default graphs are great, much easier to get into than the other system I set up.
I've since move to an infrastructure using InfluxDB, Telegraf, and Grafana. That system replaced collectd, which worked and produced high-resolution data, but also put a high load on the hosts.
But I just learned about netdata and installed it on my dev and staging environments and it looks really good. It requires basically no setup to capture useful data, generates great graphs, and looks like it could expand greatly (putting data into Influx or Prometheus, etc). The default graphs are great, much easier to get into than the other system I set up.