search

Kong / charts

Helm chart for Kong
Apache License 2.0
239 stars 473 forks source link

Service monitor scraping both status port of ingress pod #1005

Closed GDegrove closed 4 months ago

GDegrove commented 5 months ago

Hello,

since the version 2.35.0 of the helm chart, we see that the service monitor tries to scrape 2 ports per pod.

The release of the version 2.35.0 renamed the port 10254 of the ingressController status. The service monitor looks for that port name to scrape metrics, making the prometheus scrape job effectively scrape 2 metrics port.

However, the ingress-controller does not expose any metrics on the port 10254 making half of the job target unavailable.

A simple proposition would be to rename the status port of the ingress-controller to status-ingress so the service monitor does not discover that port and does not try to scrape metrics that are unavailable.

GDegrove commented 5 months ago

The PR introducing the bug: https://github.com/Kong/charts/pull/989

sgrzemski commented 5 months ago

Yup, after this MR the container has 2 ports named status and the ServiceMonitor tries to scrape both. Maybe it should be renamed to something else.

PauloFerreira25 commented 4 months ago

@czeslavo Could you please revert it?

PauloFerreira25 commented 4 months ago

The ingressController status port responds 404, if it is up. it responds with an error, what is its function?

czeslavo commented 4 months ago

@GDegrove @PauloFerreira25 Thanks for raising this issue. @sgrzemski's https://github.com/Kong/charts/pull/1008 was merged and it has been released in the kong/kong 2.37.1 patch release.