Dropped metrics#

This page lists the series the stack drops at scrape time. A dropped series never reaches Prometheus, Thanos, or Grafana.

Dropped histogram buckets#

The ServiceMonitors for the Kubernetes API server and the kubelet both drop these metric names with a metricRelabelings rule of action drop.

Regular expression

Matches

apiserver_request_duration_seconds_bucket

API server request latency buckets

apiserver_request_sli_duration_seconds_bucket

API server SLI latency buckets

apiserver_request_body_size_bytes_bucket

API server request body size buckets

apiserver_response_sizes_bucket

API server response size buckets

etcd_request_duration_seconds_bucket

etcd request latency buckets as seen by the API server

apiserver_watch_(cache_read_wait_seconds|list_duration_seconds|events_sizes)_bucket

API server watch cache buckets

workqueue_(queue|work)_duration_seconds_bucket

Controller work queue latency buckets

scheduler_plugin_execution_duration_seconds_bucket

Scheduler plugin latency buckets

The matching _sum and _count series stay. Average latencies and request rates remain available; quantiles computed from buckets do not.

Disabled scrape endpoints#

Setting

Effect

kubelet.serviceMonitor.probes: false

The kubelet /metrics/probes endpoint is not scraped, so no prober_probe_duration_seconds series exist.

kubeProxy.enabled: false

No kube-proxy target.

kubeEtcd.enabled: false

No etcd target.

kubeControllerManager.enabled: false

No kube-controller-manager target.

kubeScheduler.enabled: false

No kube-scheduler target.

See also#