Document how to run e2e tests locally

vertical-pod-autoscaler/README.md

@@ -7,6 +7,7 @@
 - [Getting started](#getting-started)
 - [Components and Architecture](#components-and-architecture)
 - [Features and Known limitations](#features-and-known-limitations)
+- [Development and testing](#development-and-testing)
 - [Related links](#related-links)
 
 ## Intro
@@ -51,6 +52,10 @@ You can also read about the [features](./docs/features.md) and [known limitation
 > * When the admission-controller tries to set a container-level limit higher than the pod-level limit, the operation is prohibited, and the pod will not be created.
 > * When the admission-controller tries to set the newly recommended container-level requests, their total value cannot exceed the pod-level request or limit. For example, if the pod-level memory request is 100Mi, but the newly recommended total of container-level requests is 250Mi, the pod will fail to be created.
 
+## Development and testing
+
+Have a look at the [Development and testing](./docs/development-and-testing.md) page for more information about developing on the VPA.
+
 ## Related links
 
 - [Design proposal](https://github.com/kubernetes/design-proposals-archive/blob/main/autoscaling/vertical-pod-autoscaler.md)

vertical-pod-autoscaler/docs/development-and-testing.md

@@ -0,0 +1,43 @@
+# Development and testing
+
+## Contents
+
+- [Development and testing](#development-and-testing)
+  - [Introduction](#introduction)
+  - [Running e2e tests](#running-e2e-tests)
+    - [Feature gates](#feature-gates)
+    - [Parallelism](#parallelism)
+    - [External Metrics Tests](#external-metrics-tests)
+
+## Introduction
+
+This project contains various scripts and tools to aid in the development of the three VPA components.
+
+## Running e2e tests
+
+The VPA contains some e2e tests that test how each component interacts with Pods and VPA resources inside a real Kubernetes cluster.
+
+They can be run using the `./hack/run-e2e-locally.sh` helper script. Please note that this script will delete any existing [kind](https://kind.sigs.k8s.io) cluster running on the local machine before creating a fresh cluster for executing the tests.
+
+### Feature gates
+
+By default, the e2e test suite only runs feature-gated tests for features that are enabled by default (typically beta and GA). Alpha features, which are disabled by default, are not tested.
+
+Setting the environment variable `ENABLE_ALL_FEATURE_GATES=true` will enable all feature gates and run all feature-gated tests.
+
+### Parallelism
+
+By default, the e2e tests create 4 worker processes, each one running its own test. This can be changed by setting the `NUMPROC=<workers>` variable.
+
+### External Metrics Tests
+
+The external metrics tests (`recommender-externalmetrics`, available in `run-e2e-locally.sh` and `deploy-for-e2e-locally.sh`)
+use a stack of 4 additional programs to support testing:
+
+1. `hack/emit-metrics.py` to generate random CPU and RAM metrics for every pod in the local cluster.
+2. Prometheus Pushgateway to accept metrics from `hack/emit-metrics`.
+3. Prometheus to store the metrics accepted by the Pushgateway.
+4. Prometheus Adapter to provide an External Metrics interface to Prometheus.
+
+The External Metrics tests run by configuring a `recommender` to use the External Metrics interface
+from the Prometheus Adapter.  With that configuration, it runs the standard `recommender` test suite.

vertical-pod-autoscaler/hack/local-cluster.md

@@ -1,40 +1,3 @@
-# Running Integration Tests locally
-
-Included in parallel with `run-e2e.sh` and `deploy-for-e2e.sh` are two alternate versions
-with `-locally` as part of their names.  They use Kubernetes in Docker (`kind`) to run a local
-cluster in Docker.  Using them will require `docker` and `kind` in your `PATH`.
-
-## External Metrics Tests
-
-The external metrics tests (`recommender-externalmetrics`, available on the `-locally` variants)
-use a stack of 4 additional programs to support testing:
-
-1. `hack/emit-metrics.py` to generate random CPU and RAM metrics for every pod in the local cluster.
-2. Prometheus Pushgateway to accept metrics from `hack/emit-metrics`.
-3. Prometheus to store the metrics accepted by the Pushgateway.
-4. Prometheus Adapter to provide an External Metrics interface to Prometheus.
-
-The External Metrics tests run by configuring a `recommender` to use the External Metrics interface
-from the Prometheus Adapter.  With that configuration, it runs the standard `recommender` test suite. 
-
-## Non-recommender tests
-
-The `recommender` and `recommender-externalmetrics` test work locally, but none of the others do;
-they require more Makefile work.
-
-# Configuration Notes
-
-To support the regular `recommender` tests locally, we've added the stock Kubernetes Metrics Server.
-Unfortunately, it doesn't work with TLS turned on.  The metrics server is being run in insecure mode
-to work around this.  This only runs in the local `kind` case, not in a real cluster.
-
-# RBAC Changes
-
-The local test cases support running the `recommender` with external metrics.  This requires
-additional permissions we don't want to automatically enable for all customers via the 
-configuration given in `deploy/vpa-rbac.yaml`.  The scripts use a context diff `hack/e2e/vpa-rbac.diff`
-to enable those permission when running locally.
-
 # Quick Integration Tests
 
 `run-integration-locally.sh` is a quicker way to integration test compared to `run-e2e-locally.sh`. Only used for simple tests.