This repository contains a set of resources that ultimately results in OpenTelemetry Collector distributions. This document contains information needed to start contributing to this repository, including how to add new distributions.
Each distribution has its own directory at the root of this repository, such as opentelemetry-collector
. Within each one of those, you'll find at least two files:
Dockerfile
, determining how to build the container image for this distributionmanifest.yaml
, which is used with ocb to generate the sources for the distribution.
Within each distribution, you are expected to be able to build it using the builder, like:
ocb --config manifest.yaml
You can build all distributions by running:
make build
If you only interested in generating the sources for the distributions, use:
make generate
Due to an incompatibility between goreleaser
and how this directory is structured, the default configuration files to be included in the container images should be placed under ./configs instead of within the distribution's main directory.
The main Makefile
is mostly a wrapper around scripts under the ./scripts directory.
goreleaser plays a big role in producing the final artifacts. Given that the final configuration file for this tool would be huge and would cause relatively big changes for each new distribution, a Makefile
target exists to generate the .goreleaser.yaml
for the repository. The Makefile
contains a list of distributions (DISTRIBUTIONS
) that is passed down to the script, which will generate snippets based on the templates from ./scripts/goreleaser-templates/
. Adding a new distribution is then only a matter of adding the distribution's directory, plus adding it to the Makefile. Adding a new snippet is only a matter of adding a new template.
Once there's a change either to the templates or to the list of distributions, a new .goreleaser.yaml
file can be generated with:
make generate-goreleaser
After that, you can test the goreleaser build process with:
make goreleaser-verify
goreleaser will build Docker images for x86_64, 386, arm, arm64 and ppc64le processors. The build process involves executing RUN
steps on the target architecture, which means the system you run it on needs support for emulating foreign architectures.
This is accomplished by installing qemu, and then enabling support for qemu within Docker:
apt-get install qemu binfmt-support qemu-user-static
docker run --rm --privileged multiarch/qemu-user-static --reset -p yes
When a new collector distribution image or binary is needed in a different platform or architecture, the following should be considered:
- Add the new platform or architecture to the Continuous Integration test matrix for both the core and contrib repositories, to ensure they can be compiled with the new combination. Failing to do so will eventually cause the release to fail due to compilation failures on those uncovered platforms, resulting in them being removed from the release matrix.
- In the
goreleaser/configure.go
file, add the new platform or architecture - Regenerate the
.goreleaser
(see goreleaser above) - In the
.github/workflows/ci-goreleaser.yaml
file, under the "Setup QEMU" action, add the new platform and architecture - In the
.github/workflows/release.yaml
file, under the "Setup QEMU" action, add the new platform and architecture