[Jedd]

Server infrastructure is mostly a solved problem - hardware (snmp/ipmi etc) and OS layer.

I think it'd be very hard at this point to come up with compelling alternatives to the incumbents in this space.

I'd certainly not want a non-free, non-battle-tested, potentially incompatible & lock-in agent that wouldn't align with the agents I currently utilise (all free in the good sense).

Push vs pull is an age-old conundrum - at dayjob we're pull - Prometheus scraping Telegraf - for OS metrics.

Though traces, front-end, RUM, SaaS metrics, logs, etc, are obviously more complex.

Whether to pull or push often comes down to how static your fleet is, but mostly whether you've got a decent CMDB that you can rely on to tell you what the state of all your endpoints - registering and decom'ing endpoints, as well as coping with scheduled outages.