Table of Contents
gem install modulesync msync --help
ModuleSync was written as a simple script with ERB templates to help the
Puppet Labs module engineers manage the zoo of Puppet modules on GitHub, and
has now been restructured and generalized to be used within other
organizations. Puppet modules within an organization tend to have a number of
meta-files that are identical or very similar between modules, such as the
spec_helper.rb. If a file needs to
change in one module, it likely needs to change in the same way in every other
module that the organization manages.
One approach to this problem is to use sed in a bash for loop on the modules to make a single change across every module. This approach falls short if there is a single file that is purposefully different than the others, but still needs to be managed. Moreover, this approach does not help if two files are structured differently but need to be changed with the same meaning; for instance, if the .travis.yml of one module uses a list of environments to include, and another uses a matrix of includes with a list of excludes, adding a test environment to both modules will require entirely different approaches.
ModuleSync provides the advantage of defining a standard template for each file to follow, so it becomes clear what a file is supposed to look like. Two files with the same semantics should also have the same syntax. A difference between two files should have clear reasons, and old cruft should not be left in files of one module as the files of another module march forward.
Another advantage of ModuleSync is the ability to run in no-op mode, which makes local changes and shows the diffs, but does not make permanent changes in the remote repository.
How It Works
ModuleSync is a gem that uses the GitHub workflow to clone, update, and push module
repositories. It expects to be activated from a directory containing
configuration for modulesync and the modules, or you can pass it the location
of this configuration directory. The configuration for the Puppet Labs
modules, can be used as an
example for your own configuration. The configuration directory contains a
directory called moduleroot which mirrors the structure of a module. The files
in the moduleroot are ERB templates, and MUST be named after the target file,
.erb. appended. The templates are
rendered using values from a file called
config_defaults.yml in the root (not
moduleroot) of the configuration directory. The default values can be
overridden or extended by adding a file called
.sync.yml to the module itself.
This allows us to, for example, have a set of "required" gems that are added
to all Gemfiles, and a set of "optional" gems that a single module might add.
Within the templates, values can be accessed in the
@configs hash, which is
merged from the values under the keys
:global and the target file name (no
The list of modules to manage is in
managed_modules.yml in the configuration
directory. This lists just the names of the modules to be managed.
ModuleSync can be called from the command line with parameters to change the
branch you're working on or the remote to clone from and push to. You can also
define these parameters in a file named
modulesync.yml in the configuration
gem install modulesync
gem build modulesync.gemspec gem install modulesync-*.gem
With no additional arguments, ModuleSync clones modules from the puppetlabs github organization and pushes to the master branch.
Make changes to a file in the moduleroot. For sanity's sake you should commit and push these changes, but in this mode the update will be rendered from the state of the files locally.
Do a dry-run to see what files will be changed, added and removed. This clones
the modules to
modules/<namespace>-<modulename> in the current working
directory, or if the modules are already cloned, does an effective
git fetch origin; git checkout master; git reset --hard origin/master on the modules.
Don't run modulesync if the current working directory contains a modules/
directory with changes you want to keep. The dry-run makes local changes there,
but does not commit or push changes. It is still destructive in that it
overwrites local changes.
msync update --noop
The --offline flag was added to allow a user to disable git support within msync. One reason for this is because the user wants to control git commands external to msync. Note, when using this command, msync assumes you have create the folder structure and git repositories correctly. If not, msync will fail to update correctly.
msync update --offline
Make changes for real and push them back to master. This operates on the pre-cloned modules from the dry-run or clones them fresh if the modules aren't found.
msync update -m "Commit message"
Amend the commit if changes are needed.
msync update --amend
For most workflows you will need to force-push an amended commit. Not required for gerrit.
msync update --amend --force
You can install a pre-push git hook to automatically clone, update, and push modules upon pushing changes to the configuration directory. This does not include a noop mode.
msync hook activate
If you have activated the hook but want to make changes to the configuration directory (such as changes to managed_modules.yml or modulesync.yml) without touching the modules, you can deactivate the hook.
msync hook deactivate
Submitting PRs/MRs to GitHub or GitLab
You can have modulesync submit Pull Requests on GitHub or Merge Requests on
GitLab automatically with the
--pr CLI option.
msync update --pr
In order for GitHub PRs or GitLab MRs to work, you must either provide
GITLAB_TOKEN environment variables,
or set them per repository in
managed_modules.yml, using the
gitlab keys respectively.
For GitHub Enterprise and self-hosted GitLab instances you also need to set the
GITLAB_BASE_URL environment variables, or use the
base_url parameter in
--- repo1: github: token: 'EXAMPLE_TOKEN' base_url: 'https://api.github.com/' repo2: gitlab: token: 'EXAMPLE_TOKEN' base_url: 'https://git.example.com/api/v4'
- Set the PR/MR title with
- Assign labels to the PR/MR with
pr_labelsshould be a list. When using the
--pr-labelsCLI option, you should use a comma separated list.
- Set the target branch with
More details for GitHub:
Using Forks and Non-master branches
The default functionality is to run ModuleSync on the puppetlabs modules, but you can use this on your own organization's modules. This functionality also applies if you want to work on a fork of the puppetlabs modules or work on a non-master branch of any organization's modules. ModuleSync does not support cloning from one remote and pushing to another, you are expected to fork manually. It does not support automating pull requests.
If you dry-run before doing the live update, you need to specify what namespace to clone from because the live update will not re-clone if the modules are already cloned. The namespace should be your fork, not the upstream module (if working on a fork).
msync update -n puppetlabs --noop
You don't technically need to specify the namespace if the modules are already cloned from the dry-run, but it doesn't hurt. You do need to specify the namespace if the modules are not pre-cloned. You need to specify a branch to push to if you are not pushing to master.
msync update -n puppetlabs -b sync_branch -m "Commit message"
Configuring ModuleSync defaults
If you're not using the puppetlabs modules or only ever pushing to a fork of them, then specifying the namespace and branch every time you use ModuleSync probably seems excessive. You can create a file called modulesync.yml in the configuration directory that provides these arguments automatically. This file has a form such as:
--- namespace: mygithubusername branch: modulesyncbranch
Then you can run ModuleSync without extra arguments:
msync update --noop msync update -m "Commit message"
Available parameters for modulesync.yml
git_base: The default URL to git clone from (Default: 'email@example.com:')
namespace: Namespace of the projects to manage (Default: 'puppetlabs'). This value can be overridden in the module name (e.g. 'namespace/mod') or by using the
namespacekey for the module in
branch: Branch to push to (Default: 'master')
remote_branch: Remote branch to push to (Default: Same value as branch)
message: Commit message to apply to updated modules.
pre_commit_script: A script to be run before commiting (e.g. 'contrib/myfooscript.sh')
pr_title: The title to use when submitting PRs/MRs to GitHub or GitLab.
pr_labels: A list of labels to assign PRs/MRs created on GitHub or GitLab.
--- namespace: MySuperOrganization branch: modulesyncbranch pr_title: "Updates to module template files via modulesync" pr_labels: - TOOLING - MAINTENANCE - MODULESYNC
--- git_base: 'firstname.lastname@example.org:' namespace: MySuperOrganization branch: modulesyncbranch
--- namespace: stackforge git_base: ssh://email@example.com:29418/ branch: msync_foo remote_branch: refs/publish/master/msync_foo pre_commit_script: openstack-commit-msg-hook.sh
If you only want to sync some of the repositories in your managed_modules.yml, use the
-f flag to filter by a regex:
msync update -f augeas -m "Commit message" # acts only on the augeas module msync update -f puppet-a..o "Commit message"
If you want to skip syncing some of the repositories in your managed_modules.yml, use the
-x flag to filter by a regex:
msync update -x augeas -m "Commit message" # acts on all modules except the augeas module msync update -x puppet-a..o "Commit message"
-f is specified, all repository are processed, if no
-x is specified no repository is skipped. If a repository matches both
-x it is skipped.
Pushing to a different remote branch
If you want to push the modified branch to a different remote branch, you can use the -r flag:
msync update -r master_new -m "Commit message"
If you install a git hook, you need to tell it what remote and branch to push to. This may not work properly if you already have the modules cloned from a different remote. The hook will also look in modulesync.yml for default arguments.
msync hook activate -n puppetlabs -b sync_branch
Modulesync can optionally bump the minor version in
metadata.json for each
modified modules if you add the
--bump flag to the command line:
msync update -m "Commit message" --bump
If you wish to tag the modified repositories with the newly bumped version,
you can do so by using the
msync update -m "Commit message" --bump --tag
Setting the tag pattern
You can also set the format of the tag to be used (
by setting the
msync update -m "Commit message" --bump --tag --tag_pattern 'v%s'
The default for the tag pattern is
Updating the CHANGELOG
When bumping the version in
metadata.json, modulesync can let you
CHANGELOG.md in each modified repository.
This is one by using the
msync update -m "Commit message" --bump --changelog
This flag will cause the
CHANGELOG.md file to be updated with the
current date, bumped (minor) version, and commit message.
CHANGELOG.md is absent in the repository, nothing will happen.
Working with templates
As mentioned, files in the moduleroot directory must be ERB templates (they must have an .erb extension, or they will be ignored). These files have direct access to @configs hash, which gets values from config_defaults.yml file and from the module being processed:
<%= @configs[:git_base] %> <%= @configs[:namespace] %> <%= @configs[:puppet_module] %>
Alternatively some meta data is passed to the template. This will allow you to add custom Ruby extensions inside the template, reading other files from the module, to make the template system more adaptive.
module: <%= @metadata[:module_name] %> target: <%= @metadata[:target_file] %> workdir: <%= @metadata[:workdir] %>
Will result in something like:
module: puppet-test target: modules/github-org/puppet-test/test workdir: modules/github-org/puppet-test