howto monitor performance metrics from thousands of windows clients?

[grum...@gmail.com]

Hello everyone,

I like Prometheus and the concept of pulling/scraping for datacenter and cloud workloads really much.

We haven't really setup a prometheus, so my knowledge of itself for now is more of documentation (and theoretically).

What i would like to know is, if one could also gather performance stats from windows desktop clients?

As far as i know on client side i could use telegraf or the wmi_exporter (or probably also sonar).

But it would stilll be the concept of prometheus pulling/scraping the metrics.

I bet i would get into some trouble if there are a few thousands of clients.

Also i would probably loose data when the clients are working offline.

Has someone already tried something similar?

I think the concept for dektops really would need to switch into pushing metrics to central

and also caching it if desktops are offline.

How do you think about this?

Best regards,

Johannes

[Brian Brazil]

On 24 January 2018 at 08:44, <grum...@gmail.com> wrote:

Hello everyone,

I like Prometheus and the concept of pulling/scraping for datacenter and cloud workloads really much.

We haven't really setup a prometheus, so my knowledge of itself for now is more of documentation (and theoretically).

What i would like to know is, if one could also gather performance stats from windows desktop clients?

As far as i know on client side i could use telegraf or the wmi_exporter (or probably also sonar).

But it would stilll be the concept of prometheus pulling/scraping the metrics.

I bet i would get into some trouble if there are a few thousands of clients.

Also i would probably loose data when the clients are working offline.

If you care about offline clients, then Prometheus is probably not what you want. Thousands of exporters is not a problem for Prometheus.

Brian

 

Has someone already tried something similar?

I think the concept for dektops really would need to switch into pushing metrics to central

and also caching it if desktops are offline.

How do you think about this?

Best regards,

Johannes

--
You received this message because you are subscribed to the Google Groups "Prometheus Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to prometheus-users+unsubscribe@googlegroups.com.
To post to this group, send email to prometheus-users@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/5f0f7320-498e-4ce0-b016-3a8ce18a0755%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Brian Brazil

www.robustperception.io

[Ben Kochie]

In addition to what Brian said, Prometheus is also designed to be distributed in nature.  If you are talking about office desktops, you will want to avoid polling over WAN links.  Having a "Prometheus per building" would be a good idea.  Each building is then an independent entity that can send you alerts.  You can also use that Prometheus to forward data via remote write streaming to something like Cortex[0].

[0]: https://github.com/weaveworks/cortex

--