Can we add header as label nginx metric in prometheus

[medicharlachiranjeevi]

Can we add a custom header to the label of Prometheus status zone metrics in Nginx Ingress Plus?

ie:
location_zone_responses_codes{ location_zone=**/,code=200,my-header=value} 0

something like this.

[github-actions]

Hi @medicharlachiranjeevi thanks for reporting!

Be sure to check out the docs and the Contributing Guidelines while you wait for a human to take a look at this 🙂

Cheers!

[medicharlachiranjeevi]

We are requesting the ability to add a custom request header as a label to the NGINX Ingress Controller's Prometheus metrics, specifically for the status zone metrics (like location_zone_responses_codes).

Problem/Use Case

We are utilizing header-based routing (e.g., routing traffic to different backend applications based on a custom header like X-App-Name or X-Client-ID).

Currently, the default NGINX metrics give us overall HTTP response codes and location zones, but they lack the context of which specific application or client generated the traffic.

By adding a custom header value as a metric label, we can:

1. Differentiate Traffic: Accurately attribute traffic and response codes to the specific backend application or routing decision made by the header.
2. Granular Monitoring: Create granular Prometheus queries, dashboards, and alerts to monitor the health and performance of individual header-based routes.

Proposed Metric Format

We propose adding a configurable header name (e.g., my-header) as a label to the existing metrics, resulting in a format similar to this:

location_zone_responses_codes{location_zone="/", code="200", my_header="value"} 0

This functionality is critical for users leveraging header-based routing to effectively monitor and troubleshoot their deployed applications.

[haywoodsh]

I assume you are using VirtualServer with conditional match and you are looking for more information in the metrics about the requests? Could you share a sample VirtualServer configuration you are using, and the current metric output from Prometheus you see?

From what I can see from the source code, the upstream NGINX Prometheus Exporter does not currently support location zone variable labels, so changes to this repo would be required.

Additionally, if we were to consider this enhancement, it would make sense to design it more generically to support all VirtualServer match conditions (headers, cookies, arguments, variables) rather than just headers specifically, since users may want observability into any of these routing criteria.