juju and monitoring systems

[Vasiliy Tolstov]

2015-11-13 17:33 GMT+03:00 Charles Butler <charles.butler at canonical.com>:
> Vasilily,
>
> I've been taking a deep look at prometheus lately - namely due to the claim
> of their top end benchmark of 360 thousand metric ingests per second on a
> beefy modern host. This intrigued me as to the use cases for prometheus. And
> seeing it being baked into the k8s apiserver for metric exports gave it
> additional credibility for being investigated.
>
> Rick's answer pretty much falls in line with what my analysis would be, that
> we dont have it on any road maps today, however an ingest for prometheus
> should be possible with some effort on both sides - the prometheus charm
> maintainer (tbd) and then the metric scrapers for juju (tbd).
>
> If this is something you are interested in, I think this conversation is a
> good starting point to look at what really needs to be done, and can be
> prototyped outside of juju core as a POC. If it gains traction, that would
> be a good identifier for it to be prioritized in a future iteration.
>
> As a bit of a curiosity, what metrics are you looking for?


Thanks for detailed answer. I'm looking for data, that can be used to
automatic scaling.
Now i'm looking for system la, cpu utilization, cpu steal time, memory
used, disk free space and utilization, network utilization...
May be the best look at https://github.com/prometheus/node_exporter
for by default enabled metrics.

-- 
Vasiliy Tolstov,
e-mail: v.tolstov at selfip.ru