Added documentation for Metrics Server

Author: piosz

_data/tasks.yml

@@ -80,6 +80,7 @@ toc:
 
 - title: Monitor, Log, and Debug
   section:
+  - docs/tasks/debug-application-cluster/core-metrics-pipeline.md
   - docs/tasks/debug-application-cluster/resource-usage-monitoring.md
   - docs/tasks/debug-application-cluster/get-shell-running-container.md
   - docs/tasks/debug-application-cluster/monitor-node-health.md

docs/tasks/debug-application-cluster/core-metrics-pipeline.md

@@ -0,0 +1,46 @@
+---
+approvers:
+- fgrzadkowski
+- piosz
+title: Core metrics pipeline
+---
+
+Resource usage metrics are important for:
+ * Kubernetes components to operate (e.g. autoscalers to automatically scale user applications)
+ * Kubernetes users to understand their applications behavior
+
+In Kubernetes they are available through Metrics API.
+
+## Metrics API
+
+Metrics API provides current resource usage metrics for pods and nodes. It doesn't offer historical metrics, so one needs to
+archive metrics themselves. In the future Historical API might be defined but it will not be implemented as a part of
+the core metrics pipeline.
+
+The API is discoverable through the same endpoint as the other Kubernetes APIs under `/apis/metrics.k8s.io/` path.
+It also offers the same security, scalability and reliability guarantees.
+
+The API is defined in [k8s.io/metrics](https://github.com/kubernetes/metrics/tree/master/pkg/apis/metrics) repository,
+where the detailed semantic is explained.
+
+### Supported endpoints
+
+The list of supported endpoints in the API:
+
+- `/nodes` - all node metrics; support for `labelSelector` parameter
+- `/nodes/{node}` - metrics for the specified node
+- `/namespaces/{namespace}/pods` - all pod metrics within the namespace; support for `labelSelector` parameter
+- `/namespaces/{namespace}/pods/{pod}` - metrics for the specified pods
+
+## Metrics Server
+
+[Metrics Server](https://github.com/kubernetes-incubator/metrics-server) is a cluster-wide aggregator of monitoring data.
+Starting from Kubernetes 1.8 it's work on all Kubernetes setups
+and runs as a pod in a user space (on one of the nodes), similar to how any Kubernetes application would run.
+
+It discovers all nodes in the cluster and queries usage information from Summary API exposed by [Kubelet](/docs/admin/kubelet/)s.
+
+Metrics Server use the same authorization/authentication mechanism as the main API server. It's registered there through
+[Kubernetes aggregator](https://kubernetes.io/docs/concepts/api-extension/apiserver-aggregation/).
+
+Learn more about the metrics server in [the design doc](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/instrumentation/metrics-server.md).