External Monitoring via SNMP of systemd services

[maxmeyer]

Hi,
I'm using systemd in my containers and on my servers to run/watch services. To monitor the whole infrastructure I'm using SNMP with some central server.

1. Question: Is there some effort to expose information of running/not running services via SNMP?
2. Would be helpful to have a small daemon exposing that information either via SNMP or via AgentX-protocol?

I found that paper which suggests to use the so called AgentX protocol for that. They implemented a small proof-of-concept daemon in python, which uses systemctl instead of dbus to query state information and uses AgentX to register itself to the snmpd from NET-SNMP:

systemd <--systemctl--> python-script <--agentx--> snmpd <--snmp--> monitoring solution

Maybe a stable solution could look like this:

systemd <--dbus--> systemd-monitord <--agentx--> snmpd <--snmp--> monitoring solution

Here's the solution of libvirt which also uses AgentX to expose information about domains to "the outer world". And here you can find the RFC 2741 for the "Agent Extensibility (AgentX) Protocol".

[maxmeyer]

@poettering What do you think abou that?

[dvdhrm]

systemctl uses DBus, so both your examples are effectively the same.

Anyway, we're not interested in shipping this upstream. There's no direct benefit in including such a monitor in the systemd repository. It can live in its own repository just fine, and provide a DBus-to-snmp bridge.

[poettering]

Yeah, I agree, this might be useful stuff, but should really live outside of systemd. We provide APIs so that people can write this, but that's really how it should be: we provide the APIs and the SNMP hookup should live outside of systemd.

[maxmeyer]

Ok. Thanks anyway.