prometheus-operator: [Helm] Separated Prometheus doesnt have Alertmanager endpoints

I wanted to deploy a separated Prometheus (to create a CRD Prometheus and let Operator generates configs ). I already deployed kube-prometheus before and it’s working FINE. I used this command to deploy one more Prometheus (And only Prometheus, no Alertmanager, no Grafana, No rules, No Servicemonitors):

cd helm/prometheus
helm install -n prometheus-app -f prometheus-app-values.yaml

And I got 2 Prometheus CRD

kk get prometheus
NAME                         AGE
kube-prometheus-prometheus   9d
prometheus-app               16m

I specified Alertmanager endpoints in prometheus-app-values.yaml like this (I wanted to send the alert to previous deployed Alertmanager):

alertingEndpoints:
  - name: "kube-prometheus-alertmanager"
     namespace: "monitoring"
     port: 9093
     scheme: http

And my “kube-prometheus-alertmanager” Endpoint object already created before like this:

kubectl get endpoints
kube-prometheus-alertmanager          100.110.210.106:9093,100.110.210.118:9093

I expected the newly created prometheus’s alertmanager should be like this (this is the Prometheus of kube-prometheus, which is already deployed before and is working fine): screen shot 2018-05-18 at 11 48 53 am However, this Alertmanager section of the separated Prometheus is BLANK.

So how should I deploy a separated Prometheus properly ?

About this issue

Original URL
State: closed
Created 6 years ago
Comments: 33 (33 by maintainers)

Commits related to this issue

Change example alerting endpoint port to http See #1356 — committed to ball-hayden/prometheus-operator by ball-hayden 6 years ago
Helm - Prometheus: Change example alerting endpoint port to http The Helm Prometheus chart expects alertingEndpoints to have a port of http, not a port number. I'm unsure of the reasoning behind this... — committed to ball-hayden/prometheus-operator by ball-hayden 6 years ago

Most upvoted comments

Wait. Gotcha. That default config has “port: http”.

The following Helm values worked for me @vielktus - maybe you’ve seen the same thing:

alertingEndpoints:
  - name: alertmanager
    namespace: monitoring
    port: http
    scheme: http

Quite why the port should be “http”, not “9093” nor “web” as suggested above, I don’t know…

ball-hayden on May 31, 2018

If that is the case, I’d suggest the bug here is the comments in the default Values.yaml (specifically line 7):

https://github.com/coreos/prometheus-operator/blob/82761eff2e9a923c256604e82a99d3c95b2d7faa/helm/prometheus/values.yaml#L4-L8

ball-hayden on May 31, 2018

Yes

brancz on May 30, 2018