I'm looking for an open source server monitoring tool for my servers:
- All servers running Linux, mostly current Debian Buster, but partly CentOS
- Other OS (Windows, *BSD etc.) don't matter
- Physical hardware servers and virtual servers (KVM/Qemu)
- Non-server devices like switches (SNMP) actually don't matter
I want to monitor at least standard parameters like load, free disk space etc. In case of the hardware servers, it was "nice to have" if the tool could access lm-sensors, IPMI or MegaCLI.
I'm looking for a tool (infrastructure) with a certain architecture:
- There's a central monitoring server listening to HTTP(S) (e.g. REST endpoint). (No binary (RMI, CORBA etc.) or proprietary mystery protocols allowed.) If there's no TLS support, that's no problem. I could place an Apache httpd in front of it for TLS termination.
- On all servers, there's an agent that sends operating parameters to the central monitoring server (PUSH) on a regular basis (by daemon or cron job).
- The agents must not listen to any other interfaces than 127.0.0.1. I don't want the central monitoring server to be able to connect to the agents in any way (pull).
- The central monitoring server must not be able run commands on the monitored nodes / servers.
- The central monitoring server has NO credentials / SSH / shell access to the monitored nodes / servers.
- If the monitoring server was infected or hacked, I don't want it to compromise the entire network and each and every server.
- I want the monitoring server to be only a passive data sink. (But it wasn't wrong if it could generate alarm emails when configured conditions occur. But that's no requirement.)
I thought about programming a very simple solution on my own - using shell scripts and curl as "agents" and ElasticSearch for the central REST server.
But I thought, maybe there's something like that ready to use so I could get around that effort??? :)
