Might be a little off-topic, but some general advice, please.
I want to monitor some 60 servers. I want to put infulxdb (collecting from collectd) on each server, rather than having one central instance.
Pros:
resilient - if it goes down, then only stats from that server are lost.
predictable - being a TSD, it’s easier to know exactly how big the DB will grow for each machine (a minor point, I know)
Performance: only queries and results go across the wire, rather than data (this might even be more data), although 60 machines could generate a lot of traffic?
Cons:
Multiple datasources in grafana
…
Any advice, please - I instinctively dislike my solution - it feels “wrong”, but I’m growing to like it.
Hi,
Doable yes, but I agree feels kind of “wrong”. Another con is the hassle of having influxdb running on each host adding load on the actual machine.
In general I think that 60 machines is quite small
I would vote for going for central instance and add hardware as you find needed. You can also use influx db clustering if you need HA, but I think you’ll need to pay for that.
Would recommend you to ask the same question in the influx community as well: https://community.influxdata.com/
Marcus
Thanks for that. You hit the nail on head with the paying thing. I’ll probably experiment with both setups.
PS permissions was the issue, (which I saw immediately once I twigged that there was a separate grafana.log, on the plugins thing.)
Thanks
Graham
Okay ![:slight_smile: :slight_smile:](https://emoji.discourse-cdn.com/twitter/slight_smile.png?v=5)
Don’t want to recommend something over something else, but i’ve personally had great experience using prometheus for monitoring servers.
Marcus
Thanks, others are pointing me in that direction, too, so I’ll take a look.
Regards
Graham