Support histogram metrics in prometheus

[gianrubio]

Ingress should support histogram metrics

[databus23]

What kind of metrics do you have in mind? response times? requests sizes?

One problem I see is that having them labeled with servername/vhost is ok but it rarely is enough. Normally you need those metrics by different api endpoints. But you can't just add the request path as a label because that might contain ids creating a unbounded amount of time series. At the ingress layer there is no way to give a proper label for different parts of a backend api. Only the backend app itself can do that.

We are using https://github.com/knyar/nginx-lua-prometheus in our nginx ingress controller (with a custom config template) to get detailed request counts by status code with success. Unfortunately this library also comes with a caveat that limits the amount if time series you can create which is not optimal in the case of histograms.

[gianrubio]

What kind of metrics do you have in mind? response times? requests sizes?

Exactly, do you have another metrics in mind?

... Normally you need those metrics by different api endpoints ...

Good point on monitoring endpoints. I was not thinking about it but we can achieve this changing the vts filter vhost_traffic_status_filter_by_set_key from $geoip_country_code country::* to $uri $host;. Look above the metrics output.

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

Prevent issues from auto-closing with an /lifecycle frozen comment.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or @fejta.
/lifecycle stale

[gianrubio]

/lifecycle frozen