I was wondering whether Performance Co-Pilot (PCP) be set up in the way that it has all the features of NMS application, ie. Nagios or Zabbix. I was using it to extract metrics from a host and plot values on a dashboard, but I would like to know how to use it to monitor several hundreds of servers and detect faults. PCP Documentation describes all the modules and deamons, but I could not find much about how to put them to work together. There is Performance Metrics Inference Engine (PMIE) which can evaluate logical expressions and raise alerts or PCP Manager (pmmgr) that can be used to collect data from multiple hosts. So if I want to use PMIE for fault detection and alerting is it supposed to run on every monitored host or on the central NMS server? Does it only read metrics from PCP logs or can it connect to monitored hosts through WebAPI? If I want to store metrics in InfluxDB or ElasticSearch should I have the exporters (pcp2influxdb or pcp2elasticsearch) running on monitored hosts or collect PCP logs in one place first and export from there? And if pmmgr collecting metrics from multiple hosts needs to have one instance of pmlogger per target host can it work efficiently on one server in large environments, ie. 1000 monitored servers?
That is a lot of questions in one post, but it all boils down to one: can PCP be NMS or not?