I am deploying prometheus using stable/prometheus-operator chart. It is installed in monitoring namespace. In the default namespace I have a pod running named my-pod with three replicas. This pod spits out metrics on port 9009 (I have verified this by doing k port-forward and validating the metrics show up in localhost:9009). I would like prometheus-operator to scrape these metrics. So I added the configuration below to values.yaml
prometheus:
prometheusSpec:
additionalScrapeConfigs:
- job_name: 'my-pod-job'
scrape_interval: 15s
kubernetes_sd_configs:
- role: pod
namespaces:
names:
- default
relabel_configs:
- source_labels: [__meta_kubernetes_pod_name]
action: keep
regex: 'my-pod' I then install prometheus using the command below:
helm upgrade --install prometheus stable/prometheus-operator \
--set kubeEtcd.enabled=false \
--set kubeControllerManager.enabled=false \
--set kubeScheduler.enabled=false \
--set prometheusOperator.createCustomResource=true \
--set grafana.smtp.existingSecret=smtp-secret \
--set kubelet.serviceMonitor.https=true \
--set kubelet.enabled=true \
-f values.yaml --namespace monitoring
However, when I go to /service-discover I see
my-pod-job (0/40 active targets)Question
How can I configure prometheus such that it scrapes metrics from pods running in default namespace and spitting out metrics on port 9009?
To tell prometheus to scrape pods, add these annotations:
...
template:
metadata:
annotations:
prometheus.io/scrape: 'true'
prometheus.io/port: '9009'As already mentioned by @gears, if you want to scrape metrics from a specific pod or service, then it is mandatory to apply the prometheus scrape annotations to it. For example:
...
spec:
replicas: 1
template:
metadata:
annotations:
prometheus.io/path: <path_to_scrape>
prometheus.io/port: "80"
prometheus.io/scrape: "true"
...However, as already stated:
The prometheus operator does not support annotation-based discovery of services, using the serviceMonitor CRD in its place as it provides far more configuration options.
Now, if I understand you correctly the ServiceMonitor object should be created in the same namespace as the application lives in. You need to make sure that the serviceMonitorNamespaceSelector selects that namespace and that the Prometheus server has the appropriate permissions to access Service/Endpoints/Pod objects in that namespace.
Please let me know if that helped.