Telegraf plugin help needed

Help
1

There is a PR on GitHub for a Telegraf plugin for Netdata - this would be a brilliant addition to Telegraf.

Is there anyone here who could help development? It is not one of my skill sets but I’d love to see it so I could use it :slight_smile:

2

Hi @borpin

Netdata Agent exposes metrics in prometheus format.

I think it is possible to use telegraf prometheus input to gather them. Looks like it will do.

Why do you think we need a dedicated netdata input? What am i missing?

1 Like
3

I had no idea about prometheus! I’ll have a look.

[edit]

My argument against Prometheus and why a native plugin is needed, is that Prometheus is yet another install and adds an additional (u and potentially unnecessary) step NetData → Prometheus → Telegraf → InfluxDB. It also yet another system to learn and maintain.

With a Telegraf plugin, the Telegraf agent can run on a remote server (in my case the InfluxDB server), and get the data directly.

Seems a much neater solution. As most of the work has already been done, it seems a good idea :slight_smile:

4

If i understand it correctly you need no Prometheus.

Telegraf has prometheus input, it collects data from any http endpoint that exposes data in prometheus format - Netdata is able to do it.

So it is:

Server <-[collects server metrics] Netdata ← [pulls metrics via http] Telegraf → InfluxDB

5

I think that is not correct.

From the page you directed me to, it wants me to install the Prometheus client Using Netdata with Prometheus | Learn Netdata

Telegraf can connect to a Prometheus client and scrape the data, but that is still an intermediate step rather than a native Netdata plugin for Telegraf that would then be a connector in it’s own right.

6

Ok, by bad, i wasn’t clear, i meant to highlight the following:

Using Netdata with Prometheus | Learn Netdata

Querying Metrics
Fetch with your web browser this URL:

http://your.netdata.ip:19999/api/v1/allmetrics?format=prometheus&help=yes

So it is native. Telegraf scrapes Netdata Agent directly.

I am trying to understand why we need something else.

1 Like
7

OK thanks.

For anyone finding this by search the configuartion that will work…

[[inputs.prometheus]]
  urls = ["http://<IP>:19999/api/v1/allmetrics?format=prometheus"]
  name_override = "netdata-pve"

Where the name-override is the measurment you want the metrics to go to.

2 Likes
8

@ilyam8, Telegraf is complaining

[inputs.prometheus] Use of deprecated configuration: 'metric_version = 1'; please update to 'metric_version = 2'

There is no mention if the Netdata output is V1 or V2. As the URL is V1, I’m thinking it is the V1 version (it certainly works).

9

That is telegraf configuration option - metric_version

Default is != 2. It complains about prometheus input configuration, try to set metric_version to 2.

2 Likes
10

OK thanks. I don’t know what the difference between the metric_versions is. I wondered if the format the data was presented differently.