Skip to content

Latest commit

 

History

History
303 lines (214 loc) · 12.9 KB

BUILDING.md

File metadata and controls

303 lines (214 loc) · 12.9 KB

Build containerd from source

This guide is useful if you intend to contribute on containerd. Thanks for your effort. Every contribution is very appreciated.

This doc includes:

Getting started with GitHub Codespaces

To get started, create a codespace for this repository by clicking this 👇

Open in GitHub Codespaces

A codespace will open in a web-based version of Visual Studio Code. The dev container is fully configured with software needed for this project and the containerd built. If you use a codespace, then you can directly skip to the testing section of this document.

Note: Dev containers is an open spec which is supported by GitHub Codespaces and other tools.

Build requirements

To build the containerd daemon, and the ctr simple test client, the following build system dependencies are required:

  • Go 1.22.x or above
  • Protoc 3.x compiler and headers (download at the Google protobuf releases page)
  • Btrfs headers and libraries for your distribution. Note that building the btrfs driver can be disabled via the build tag no_btrfs, removing this dependency.

Note: On macOS, you need a third party runtime to run containers on containerd

Build the development environment

First you need to setup your Go development environment. You can follow this guideline How to write go code and at the end you have go command in your PATH.

You need git to checkout the source code:

git clone https://github.com/containerd/containerd

For proper results, install the protoc release into /usr/local on your build system. When generating source code from .proto files, containerd may rely on some external protocol buffer files. These external dependencies should be added to the /usr/local/include directory. To install the appropriate version of protoc and download any necessary external protocol buffer files on a Linux host, run the install script located at script/setup/install-protobuf.

To enable optional Btrfs snapshotter, you should have the headers from the Linux kernel 4.12 or later. The dependency on the kernel headers only affects users building containerd from source. Users on older kernels may opt to not compile the btrfs support (see BUILDTAGS=no_btrfs below), or to provide headers from a newer kernel.

Note The dependency on the Linux kernel headers 4.12 was introduced in containerd 1.7.0-beta.4.

containerd 1.6 has different set of dependencies for enabling btrfs. containerd 1.6 users should refer to https://github.com/containerd/containerd/blob/release/1.6/BUILDING.md#build-the-development-environment

At this point you are ready to build containerd yourself!

Runc

Runc is the default container runtime used by containerd and is required to run containerd. While it is okay to download a runc binary and install that on the system, sometimes it is necessary to build runc directly when working with container runtime development. Make sure to follow the guidelines for versioning in RUNC.md for the best results.

Note: Runc only supports Linux

Build containerd

containerd uses make to create a repeatable build flow. It means that you can run:

cd containerd
make

This is going to build all the project binaries in the ./bin/ directory.

You can move them in your global path, /usr/local/bin with:

sudo make install

The install prefix can be changed by passing the PREFIX variable (defaults to /usr/local).

Note: if you set one of these vars, set them to the same values on all make stages (build as well as install).

If you want to prepend an additional prefix on actual installation (eg. packaging or chroot install), you can pass it via DESTDIR variable:

sudo make install DESTDIR=/tmp/install-x973234/

The above command installs the containerd binary to /tmp/install-x973234/usr/local/bin/containerd

The current DESTDIR convention is supported since containerd v1.6. Older releases was using DESTDIR for a different purpose that is similar to PREFIX.

When making any changes to the gRPC API, you can use the installed protoc compiler to regenerate the API generated code packages with:

make generate

Note: Several build tags are currently available:

  • no_cri: A build tag disables building Kubernetes CRI support into containerd. See here for build tags of CRI plugin.
  • snapshotters (alphabetical order)
    • no_aufs: A build tag disables building the aufs snapshot driver. (Ignored since containerd v2.0, as the aufs snapshot driver is no longer supported)
    • no_btrfs: A build tag disables building the Btrfs snapshot driver.
    • no_devmapper: A build tag disables building the device mapper snapshot driver.
    • no_zfs: A build tag disables building the ZFS snapshot driver.
  • platform
    • no_systemd: disables any systemd specific code

For example, adding BUILDTAGS=no_btrfs to your environment before calling the binaries Makefile target will disable the btrfs driver within the containerd Go build.

Vendoring of external imports uses the Go Modules. You need to use go mod command to modify the dependencies. After modifition, you should run go mod tidy and go mod vendor to ensure the go.mod, go.sum files and vendor directory are up to date. Changes to these files should become a single commit for a PR which relies on vendored updates.

Please refer to RUNC.md for the currently supported version of runc that is used by containerd.

Note: On macOS, the containerd daemon can be built and run natively. However, as stated above, runc only supports linux.

Static binaries

You can build static binaries by providing a few variables to make:

make STATIC=1

Note:

  • static build is discouraged
  • static containerd binary does not support loading shared object plugins (*.so)
  • static build binaries are not position-independent

Via Docker container

The following instructions assume you are at the parent directory of containerd source directory.

Build containerd in a container

You can build containerd via a Linux-based Docker container. You can build an image from this Dockerfile:

FROM golang

Let's suppose that you built an image called containerd/build. From the containerd source root directory you can run the following command:

docker run -it \
    -v ${PWD}/containerd:/go/src/github.com/containerd/containerd \
    -e GOPATH=/go \
    -w /go/src/github.com/containerd/containerd containerd/build sh

This mounts containerd repository

You are now ready to build:

make && make install

Build containerd and runc in a container

To have complete core container runtime, you will need both containerd and runc. It is possible to build both of these via Docker container.

You can use git to checkout runc:

git clone https://github.com/opencontainers/runc

We can build an image from this Dockerfile:

FROM golang

RUN apt-get update && \
    apt-get install -y libseccomp-dev

In our Docker container we will build runc build, which includes seccomp, SELinux, and AppArmor support. Seccomp support in runc requires libseccomp-dev as a dependency (AppArmor and SELinux support do not require external libraries at build time). Refer to RUNC.md in the docs directory to for details about building runc, and to learn about supported versions of runc as used by containerd.

Let's suppose you build an image called containerd/build from the above Dockerfile. You can run the following command:

docker run -it --privileged \
    -v /var/lib/containerd \
    -v ${PWD}/runc:/go/src/github.com/opencontainers/runc \
    -v ${PWD}/containerd:/go/src/github.com/containerd/containerd \
    -e GOPATH=/go \
    -w /go/src/github.com/containerd/containerd containerd/build sh

This mounts both runc and containerd repositories in our Docker container.

From within our Docker container let's build containerd:

cd /go/src/github.com/containerd/containerd
make && make install

These binaries can be found in the ./bin directory in your host. make install will move the binaries in your $PATH.

Next, let's build runc:

cd /go/src/github.com/opencontainers/runc
make && make install

For further details about building runc, refer to RUNC.md in the docs directory.

When working with ctr, the simple test client we just built, don't forget to start the daemon!

containerd --config config.toml

Testing containerd

During the automated CI the unit tests and integration tests are run as part of the PR validation. As a developer you can run these tests locally by using any of the following Makefile targets:

  • make test: run all non-integration tests that do not require root privileges
  • make root-test: run all non-integration tests which require root
  • make integration: run all tests, including integration tests and those which require root. TESTFLAGS_PARALLEL can be used to control parallelism. For example, TESTFLAGS_PARALLEL=1 make integration will lead a non-parallel execution. The default value of TESTFLAGS_PARALLEL is 8.
  • make cri-integration: CRI Integration Tests run cri integration tests

To execute a specific test or set of tests you can use the go test capabilities without using the Makefile targets. The following examples show how to specify a test name and also how to use the flag directly against go test to run root-requiring tests.

# run the test <TEST_NAME>:
go test	-v -run "<TEST_NAME>" .
# enable the root-requiring tests:
go test -v -run . -test.root

Example output from directly running go test to execute the TestContainerList test:

sudo go test -v -run "TestContainerList" . -test.root
INFO[0000] running tests against containerd revision=f2ae8a020a985a8d9862c9eb5ab66902c2888361 version=v1.0.0-beta.2-49-gf2ae8a0
=== RUN   TestContainerList
--- PASS: TestContainerList (0.00s)
PASS
ok  	github.com/containerd/containerd	4.778s

Note: in order to run sudo go you need to

  • either keep user PATH environment variable. ex: sudo "PATH=$PATH" env go test <args>
  • or use go test -exec ex: go test -exec sudo -v -run "TestTarWithXattr" ./archive/ -test.root

Additional tools

containerd-stress

In addition to go test-based testing executed via the Makefile targets, the containerd-stress tool is available and built with the all or binaries targets and installed during make install.

With this tool you can stress a running containerd daemon for a specified period of time, selecting a concurrency level to generate stress against the daemon. The following command is an example of having five workers running for two hours against a default containerd gRPC socket address:

containerd-stress -c 5 -d 120m

For more information on this tool's options please run containerd-stress --help.

bucketbench

Bucketbench is an external tool which can be used to drive load against a container runtime, specifying a particular set of lifecycle operations to run with a specified amount of concurrency. Bucketbench is more focused on generating performance details than simply inducing load against containerd.

Bucketbench differs from the containerd-stress tool in a few ways:

  • Bucketbench has support for testing the Docker engine, the runc binary, and containerd 0.2.x (via ctr) and 1.0 (via the client library) branches.
  • Bucketbench is driven via configuration file that allows specifying a list of lifecycle operations to execute. This can be used to generate detailed statistics per-command (e.g. start, stop, pause, delete).
  • Bucketbench generates detailed reports and timing data at the end of the configured test run.

More details on how to install and run bucketbench are available at the GitHub project page.