we are also facing same issue with federate Prometheus what you face earlier if you provide a solution to me it really helpful to me
I try to scrape from the federation endpoint, which resides inside an open shift cluster. The openshift-prometheus-instance has two replicas. I am getting the double amount of metrics, labeled with "prometheus_replica=replica_0" and "prometheus_replica=replica_1".
job_name: prometheus_openshift_prod
honor_labels: true
metrics_path: '/federate'
params:
'match[]':
-'{job="kube-state-metrics"}'
'{job="kubelet"}'
scheme: https
static_configs:
-targets:
- endpoint
Thanks,
Veera