Building Kubernetes

Building Kubernetes is easy if you take advantage of the containerized build environment. This document will help guide you through understanding this build process.

Requirements

1. Docker, using one of the following configurations:

• macOS You can either use Docker for Mac or docker-machine. See installation instructions here. Note: You will want to set the Docker VM to have at least 4.5GB of initial memory or building will likely fail. (See: #11852).
• Linux with local Docker Install Docker according to the instructions for your OS.
• Windows with Docker Desktop WSL2 backend Install Docker according to the instructions. Be sure to store your sources in the local Linux file system, not the Windows remote mount at /mnt/c.
• Remote Docker engine Use a big machine in the cloud to build faster. This is a little trickier so look at the section later on.

2. Optional Google Cloud SDK

You must install and configure Google Cloud SDK if you want to upload your release to Google Cloud Storage and may safely omit this otherwise.

Overview

While it is possible to build Kubernetes using a local golang installation, we have a build process that runs in a Docker container. This simplifies initial set up and provides for a very consistent build and test environment.

Key scripts

The following scripts are found in the build/ directory. Note that all scripts must be run from the Kubernetes root directory.

• build/run.sh: Run a command in a build docker container. Common invocations:
  • build/run.sh make: Build just linux binaries in the container. Pass options and packages as necessary.
  • build/run.sh make cross: Build all binaries for all platforms. To build only a specific platform, add KUBE_BUILD_PLATFORMS=<os>/<arch>
  • build/run.sh make kubectl KUBE_BUILD_PLATFORMS=darwin/amd64: Build the specific binary for the specific platform (kubectl and darwin/amd64 respectively in this example)
  • build/run.sh make test: Run all unit tests
  • build/run.sh make test-integration: Run integration test
  • build/run.sh make test-cmd: Run CLI tests
• build/copy-output.sh: This will copy the contents of _output/dockerized/bin from the Docker container to the local _output/dockerized/bin. It will also copy out specific file patterns that are generated as part of the build process. This is run automatically as part of build/run.sh.
• build/make-clean.sh: Clean out the contents of _output, remove any locally built container images and remove the data container.
• build/shell.sh: Drop into a bash shell in a build container with a snapshot of the current repo code.

Basic Flow

The scripts directly under build/ are used to build and test. They will ensure that the kube-build Docker image is built (based on build/build-image/Dockerfile and after base image's KUBE_BUILD_IMAGE_CROSS_TAG from Dockerfile is replaced with one of those actual tags of the base image, like v1.13.9-2) and then execute the appropriate command in that container. These scripts will both ensure that the right data is cached from run to run for incremental builds and will copy the results back out of the container.

The kube-build container image is built by first creating a "context" directory in _output/images/build-image. It is done there instead of at the root of the Kubernetes repo to minimize the amount of data we need to package up when building the image.

There are 3 different containers instances that are run from this image. The first is a "data" container to store all data that needs to persist across to support incremental builds. Next there is an "rsync" container that is used to transfer data in and out to the data container. Lastly there is a "build" container that is used for actually doing build actions. The data container persists across runs while the rsync and build containers are deleted after each use.

rsync is used transparently behind the scenes to efficiently move data in and out of the container. This will use an ephemeral port picked by Docker. You can modify this by setting the KUBE_RSYNC_PORT env variable.

All Docker names are suffixed with a hash derived from the file path (to allow concurrent usage on things like CI machines) and a version number. When the version number changes all state is cleared and clean build is started. This allows the build infrastructure to be changed and signal to CI systems that old artifacts need to be deleted.

Proxy Settings

If you are behind a proxy and you are letting these scripts use docker-machine to set up your local VM for you on macOS, you need to export proxy settings for Kubernetes build, the following environment variables should be defined.

export KUBERNETES_HTTP_PROXY=http://username:password@proxyaddr:proxyport
export KUBERNETES_HTTPS_PROXY=https://username:password@proxyaddr:proxyport

Optionally, you can specify addresses of no proxy for Kubernetes build, for example

export KUBERNETES_NO_PROXY=127.0.0.1

If you are using sudo to make Kubernetes build for example make quick-release, you need run sudo -E make quick-release to pass the environment variables.

Really Remote Docker Engine

It is possible to use a Docker Engine that is running remotely (under your desk or in the cloud). Docker must be configured to connect to that machine and the local rsync port must be forwarded (via SSH or nc) from localhost to the remote machine.

To do this easily with GCE and docker-machine, do something like this:

# Create the remote docker machine on GCE.  This is a pretty beefy machine with SSD disk.
KUBE_BUILD_VM=k8s-build
KUBE_BUILD_GCE_PROJECT=<project>
docker-machine create \
  --driver=google \
  --google-project=${KUBE_BUILD_GCE_PROJECT} \
  --google-zone=us-west1-a \
  --google-machine-type=n1-standard-8 \
  --google-disk-size=50 \
  --google-disk-type=pd-ssd \
  ${KUBE_BUILD_VM}

# Set up local docker to talk to that machine
eval $(docker-machine env ${KUBE_BUILD_VM})

# Pin down the port that rsync will be exposed on the remote machine
export KUBE_RSYNC_PORT=8730

# forward local 8730 to that machine so that rsync works
docker-machine ssh ${KUBE_BUILD_VM} -L ${KUBE_RSYNC_PORT}:localhost:${KUBE_RSYNC_PORT} -N &

Look at docker-machine stop, docker-machine start and docker-machine rm to manage this VM.

Releasing

The build/release.sh script will build a release. It will build binaries, run tests, (optionally) build runtime Docker images.

The main output is a tar file: kubernetes.tar.gz. This includes:

• Cross compiled client utilities.
• Script (kubectl) for picking and running the right client binary based on platform.
• Examples
• Cluster deployment scripts for various clouds
• Tar file containing all server binaries

In addition, there are some other tar files that are created:

• kubernetes-client-*.tar.gz Client binaries for a specific platform.
• kubernetes-server-*.tar.gz Server binaries for a specific platform.

When building final release tars, they are first staged into _output/release-stage before being tar'd up and put into _output/release-tars.

Reproducibility

make release, its variant make quick-release, and Bazel all provide a hermetic build environment which should provide some level of reproducibility for builds. make itself is not hermetic.

The Kubernetes build environment supports the SOURCE_DATE_EPOCH environment variable specified by the Reproducible Builds project, which can be set to a UNIX epoch timestamp. This will be used for the build timestamps embedded in compiled Go binaries, and maybe someday also Docker images.

One reasonable setting for this variable is to use the commit timestamp from the tip of the tree being built; this is what the Kubernetes CI system uses. For example, you could use the following one-liner:

SOURCE_DATE_EPOCH=$(git show -s --format=format:%ct HEAD)