puppetlabs/ezbake⁠

The Dockerfile for this image is available in the ezbake repository here⁠.

This container will let you build custom packages for projects using ezbake, including PuppetServer and PuppetDB.

Configuration

The following environment variables are supported:

• EZBAKE_ALLOW_UNREPRODUCIBLE_BUILDS

  Allow ezbake to use locally-hosted SNAPSHOT builds rather than requiring they come from a configured snapshots repository. Defaults to true

• EZBAKE_NODEPLOY

  Do not deploy SNAPSHOT builds to a configured snapshots repository. Defaults to true

• GEM_SOURCE

  The rubygems host you want to point to. Defaults to https://rubygems.org

• LEIN_PROFILES

  The lein profiles to use to build your package. Use a comma-separated list for multiple profiles. Defaults to ezbake

• EZBAKE_REPO

  The URL for the repo of your custom ezbake. i.e., https://github.com/puppetlabs/ezbake. Unset by default

• EZBAKE_BRANCH

  The branch you want to build your custom ezbake from. Unset by default. You should only set one of EZBAKE_BRANCH and EZBAKE_REF

• EZBAKE_REF

  The git ref you want to build your custom ezbake from. Unset by default. You should only set one of EZBAKE_BRANCH and EZBAKE_REF

• PROJECT_REPO

  The URL for the repo you're building. i.e., https://github.com/underscorgan/puppetdb. Unset by default

• PROJECT_BRANCH

  The branch you want to build your packages from. Unset by default. You should only set one of PROJECT_BRANCH and PROJECT_REF

• PROJECT_REF

  The git ref you want to build your packages from. Unset by default. You should only set one of PROJECT_BRANCH and PROJECT_REF

• MOCK

  NOTEMOCK and COW have some weird requirements tied to assumptions in the packaging repo⁠. There's a plan to make this better at some point.

  The rpm platforms you want to build. Individual entries must match the pattern pl-<os>-<version>-<arch>. For multiple OSes, use a space separated list. Defaults to pl-el-6-i386 pl-el-7-x86_64 pl-el-8-x86_64 pl-redhatfips-7-x86_64 pl-sles-12-x86_64

• COW

  NOTEMOCK and COW have some weird requirements tied to assumptions in the packaging repo⁠. There's a plan to make this better at some point.

  The deb platforms you want to build. Individual entries must match the pattern base-<codename>-<arch>.cow. For multiple OSes, use a space separated list. Defaults to base-jessie-i386.cow base-xenial-i386.cow base-stretch-i386.cow base-bionic-i386.cow

• UPDATE_EZBAKE_VERSION

  Whether or not to update the ezbake version in the project you're building. This is mostly intended to let you test ezbake changes without having to commit code with a -SNAPSHOT version in project.clj. Defaults to false

• EZBAKE_VERSION

  The version of ezbake to use for your build. Only does anything if UPDATE_EZBAKE_VERSION is set to true. Defaults to unset

Running

There are a few different ways I've found for running builds on this container. I'll document the pros and cons here.

This container assumes that the code you're building will be in /workspace. The packages will be copied into /output before the container terminates, and if you're building a custom ezbake that will be cloned into /ezbake.

1. Run from a remotely accessible repo (github, gitlab, etc)

This is by far the fastest build option, but does require the extra step of committing and pushing your code rather than using the code in your working dir.

docker run --rm --volume $(PWD)/output:/output --env PROJECT_REPO=https://github.com/underscorgan/puppetdb --env PROJECT_BRANCH=my_feature_branch puppet/ezbake

1. Mount in your current working directory

   • Please note: due to the amount of I/O used in compiling the artifacts it will be incredibly slow if you mount in your entire current working directory. The options below will speed it up significantly (I haven't actually managed to get a build to finish and have killed it after ~30 minutes when I've mounted in my whole working directory), but it still takes significantly more than the builds from remote repo. The options are:

   a. Mount in full working directory and use anonymous volumes for the I/O heavy directories

   This is the slower of the two options, but also the simplest. The volumes in the example should be the same for all ezbake projects.

   docker run --rm --volume $(PWD)/output:/output --volume /workspace/test --volume /workspace/test-resources --volume /workspace/target --volume /workspace/tmp puppet/ezbake
   

   Copy

   b. Only mount in needed files/directories

   This is faster than mounting in everything and using anonymous volumes, but it'll take more effort to get the correct things mounted in. This will vary some between projects, but you can get a pretty good idea of what is needed by looking at the .gitignore or .dockerignore files.

   NOTE you will always need to mount in .git so the packaging can figure out what version it's supposed to be (uses git describe).

   docker run --rm --volume $(PWD)/output:/output --volume $(PWD)/src:/workspace/src --volume $(PWD)/project.clj:/workspace/project.clj --volume $(PWD)/resources:/workspace/resources --volume $(PWD)/.git:/workspace/.git puppet/ezbake
   

   Copy