The gocd-pipeline-builder lets you create a GoCD pipeline from a git repository.
The actions of the gocd-pipeline-builder are driven by a json file containing a list of operations to perform. See below and the pattern.json files under src/texttest.
This json file can use Jinja2 templates to pass in custom values etc, so that the .json-files can be generic templates.
The values to pass to the Jinja2 template can either be passed to the tool on the command file, and/or come from a .yaml file.
As of today, there is no builtin githook support, but it's no rocket science to call the tool on post-receive.
The gocdpb script configures new pipelines in a GoCD server.
usage: gocdpb [-h] [-j JSON_SETTINGS | -y YAML_SETTINGS] [-p PLUGIN]
[-D DEFINE] [--dump-test-config DUMP_TEST_CONFIG] [-d DUMP]
[-v] [-c CONFIG] [-C CONFIG_PARAM] [-P PASSWORD_PROMPT]
[--set-test-config SET_TEST_CONFIG]
Add pipeline to Go CD server.
optional arguments:
-h, --help show this help message and exit
-j JSON_SETTINGS, --json-settings JSON_SETTINGS
Read json file / url with settings for GoCD pipeline.
-y YAML_SETTINGS, --yaml-settings YAML_SETTINGS
Read yaml files with parameters for GoCD pipeline.
-p PLUGIN, --plugin PLUGIN
Plugin module for custom functions.
-D DEFINE, --define DEFINE
Define setting parameter on command line.
--dump-test-config DUMP_TEST_CONFIG
Copy of some sections of new GoCD configuration XML
file.
-d DUMP, --dump DUMP Copy of new GoCD configuration XML file.
-v, --verbose Write status of created pipeline.
-c CONFIG, --config CONFIG
Yaml file with configuration.
-C CONFIG_PARAM, --config-param CONFIG_PARAM
Define config parameter on command line.
-P PASSWORD_PROMPT, --password-prompt PASSWORD_PROMPT
Prompt for config parameter without echo.
--set-test-config SET_TEST_CONFIG
Set some sections in config first. (For test setup.)
-c and -C/-P are two ways of providing the same information: From file or on command line. -P allows us to secretly provide passwords.
-j together with -D can provide the same information on the commandline as -y would provide in a file. See below.
Assuming that you have an appropriate settings file available at http:/server/settings.json
with pattern parameters {{ repo_url }}, {{ group }} and {{ repo_name }}, and you
stand in a git checkout and want a new pipeline named as the checkout directory, in
pipeline group X at http://mygoserver:8153, where you log in as charlie, you
can type this:
gocdpb -C url=http://mygoserver:8153 -C username=charlie -P password -D group=X -j http:/server/settings.json
You will be prompted for password.
If you want the pipeline to be called Y instead of the directory name,
simply add -D repo_name=Y to the command line.
The -p | --plugin flag is new in version 7. See section on plugins below.
See http://www.go.cd/documentation/user/current/configuration/pipeline_templates.html and http://www.go.cd/documentation/user/current/configuration/admin_use_parameters_in_configuration.html
GoCD has the concepts of pipeline templates and parameters. If we use this, there is not so much that we need to define outside the normal GoCD configuration.
To create a GoCD pipeline with the pipeline builder, you use a json file which follows this pattern:
[
{
"environment": "name of environment",
"unpause": "true or false"
"<action>": <data>
}
]
Use gocdpb -j to pass the json settings file to the builder.
See *.json files under src/texttest for more examples of intended usage.
Both environment and unpause are optional.
<action> can be one of:
- "create-a-pipeline": This means that we plan
to call the REST API for creation of a new pipeline.
See https://api.go.cd/current/#create-a-pipeline .
<data>should be a json pipeline configuration as described by this API. For this action, the "environment" field will indicate that the new pipeline should belong to a Go server environment with the given name. Newly created pipelines are typically paused, but setting "unpause" to "true" will make it an immediate candidate for getting built. - "add-downstream-dependencies": This action is
used to add the newly created pipeline as a
dependency to a downstream pipeline.
<data>is an object with the following members:- "name" of the downstream pipeline.
- "material" (optional) specifies the dependency material, so that defaults can be overridden. This object should follow the structure of dependency material in the REST API. If missing, a dependency material object with the name of the newly created pipeline with default values will be used.
- "stage" is used for fetchartifact tasks to tell where to add it. If missing, the first stage will be used.
- "job" is used for fetchartifact tasks to tell where to add it. If missing, the first job will be used.
- "task" should be a fetch task as described in the REST API which fetches some build artifact from the newly created pipeline.
- Custom value: See section on Plugins below.
The pipeline builder has a concept called patterns.
We could have called it templates, but since GoCD has
something else which is called pipeline templates,
we call these things patterns.
A pattern is a Jinja2 template, see http://jinja.pocoo.org/
With a pipeline builder pattern, we can avoid a lot
of repetition. For instance, if we have several pipelines
like the one above, that only differ on name, we can have
yaml parameter file like this:
path: ./simple_pipeline_pattern.json
parameters:
name: gocd
url: https://github.com/magnus-lycka/gocd.git
Use gocdpb -y to pass the yaml parameter file to the builder.
The pattern file ./simple_pipeline_pattern.json is a
json file as described in the previous section, with
jinja2 templating. See the test suite for examples.
See also Jinja2 docs at http://jinja.pocoo.org/
See *.yaml files under src/texttest for more examples of intended usage.
Whether you run goplbld -y or goplbld -j, you can
use builtin parameters to substitute values via the Jinja2
template mechanism. The following builtin parameters exist:
| Parameter | Default value |
|---|---|
| repo_url | URL for git repo from ./.git/config |
| repo_name | Name of current working directory |
If you need to maintain a released version of your software in parallel with
development of new features, you might want to tag and/or branch all source
code repos involved in creating a release. The gocdrepos and gocdtagrepos
tools help prepare that, while the clone-pipelines action for gocdpb will
set up the actual pipelines you need for your release, with release pipelines
in a separate pipeline-group using branched source code repos as material.
(This might not be the ideal way of working with Continuous Deployment, but its a reality for many teams that strive towards Continuous Deployment.)
Features not supported by gocdpb can be added through plugins.
You need to provide a Python module with a module level dictionary
called action_plugins with the action names as keys, and each
corresponding handler as values. A handler could be any callable
Python object, i.e. a class with an __init__, an instance with
a __call__, or a function or method.
The callable should only accept keyword arguments, and accept any keyword argument to be able to deal with future expansions without breaking. The caller of the callable doesn't expect any return value. Throw an explanatory exception if the expectations of calling can't be met.
Something like this:
# my_plugin_module.py
def beer_pipeline_handler(go=None, operation=None, **kwargs):
...
if brand not in inventory:
raise ValueError('Unknown brand')
...
action_plugins = {'beer-pipeline': beer_pipeline_handler}
By calling gocdpb with -p my_plugin_module and using beer-pipeline
as action in the GoCD Pipeline Builder Setting File, as described above,
we'll make gocdpb run the beer_pipeline_handler.
The parameter go is an instance of goserver_adapter.Goserver. The general
idea is that Goserver has methods corresponding to the GoCD REST API calls
which take the needed parameters (url and login is already taken care of)
and returns either json_data as the result of json.load (with OrderedDict)
or a tuple of etag, json_data.
The parameter operation in the value in the Json settings file corresponding
to the key in the action_plugins dictionary. So, for the example above, if
we have the following in our settings file...
...
'beer-pipeline': {'brand': 'Carnegie Porter', 'flow': '600l/h', 'alc': '6.2%'},
...
...we'll call beer_pipeline_handler with operation set to
{'brand': 'Carnegie Porter', 'amount': '600', 'alc': '6.2%'}.
The gocdrepos tool lists all the source code repositories used directly or
indirectly to build a certain instance of a pipeline. The -f flag is used to
decide whether to use json format or a flat semicolon separated format for the
output. All other flags work just as for gocdpb.
The pipeline_instance parameter has to be provided on the form
/, so if you want to have a json listing
of the source code repositories used to create build 12 or the hello
pipeline, you might type this and provide the password when prompted:
gocdrepos -C url=http://go -C username=gouser -P password -f json hello/12
usage: gocdrepos [-h] [-f {semicolon,json}] [-v] [-c CONFIG] [-C CONFIG_PARAM]
[-P PASSWORD_PROMPT] [--set-test-config SET_TEST_CONFIG]
pipeline_instance
Recursively fetch all source code revisions used in a pipeline build.
positional arguments:
pipeline_instance pipeline/instance to start at.
optional arguments:
-h, --help show this help message and exit
-f {semicolon,json}, --format {semicolon,json}
Format for output.
-v, --verbose Write status of created pipeline.
-c CONFIG, --config CONFIG
Yaml file with configuration.
-C CONFIG_PARAM, --config-param CONFIG_PARAM
Define config parameter on command line.
-P PASSWORD_PROMPT, --password-prompt PASSWORD_PROMPT
Prompt for config parameter without echo.
--set-test-config SET_TEST_CONFIG
Set some sections in config first. (For test setup.)
The gocdtagrepos tool uses a json-file of the format created by gocdrepos. For each repo in the json-file, it will make a clone under DIRECTORY, tag it with TAG_NAME, and also branch it with the TAG_NAME if the pipeline used to create it was listed in BRANCH_LIST. The -p flag is used to automatically push the repository, and the -c flag to remove the clone when it's done.
Currently, this tool only supports git.
usage: gocdtagrepos [-h] [-d DIRECTORY] -t TAG_NAME [-b BRANCH_LIST] [-p] [-c]
[jsonfile]
Tag and/or branch a set of Git repositories as provided by json data.
positional arguments:
jsonfile Json file as produced by gocdrepos. Read from stdin if
no filename given.
optional arguments:
-h, --help show this help message and exit
-d DIRECTORY, --directory DIRECTORY
Parent directory of repository clones clones. Default
to /tmp.
-t TAG_NAME, --tag-name TAG_NAME
Name of tag / branch to create.
-b BRANCH_LIST, --branch-list BRANCH_LIST
Comma-separated list of pipeline names. Create
branches for these.
-p, --push Push changes to remote repo.
-c, --clean Remove cloned repo.
- Make git hooks to run gocdpb automatically.
This software is mainly developed through an ATDD approach, with the bulk of tests in texttest. Unit tests exist are mainly complements to the the functional tests.
Run ./test.sh in the root folder for unit tests, and follow the instructions below for functional tests:
Install a go-server for functional tests. See https://www.go.cd/documentation/user/current/installation/installing_go_server.html
The test will wipe most of the configuration before each test, so don't use a go-server that's used for anything else for these tests.
Create a user in the go-server with credentials corresponding to src/texttest/goplbld.yaml.
See https://www.go.cd/documentation/user/current/configuration/dev_authentication.html
hint:
sudo htpasswd -bcs /etc/go/htpasswd gouser verysecret
If you manage to lock yourself out of the go-server when you do this, you can fix the /etc/go/cruise-config.xml in an editor and restart the service. It might be good to save a backup of cruise-config.xml before you start changing stuff.
Install TextTest and CaptureMock: [http://texttest.org]. Also install all the python libraries listed in 'requirements.txt'.
hint:
sudo pip install TextTest
sudo pip install CaptureMock
sudo pip install -r requirements.txt
In your $TEXTTEST_HOME folder create a softlink pointing out where the test code is. Given an environment variable $CLONE_LOCATION which points to the directory where you have cloned this repo:
cd $TEXTTEST_HOME
ln -s $CLONE_LOCATION/gocd-pipeline-builder/src/texttest gocdpb
In your texttest personal config file (defaults to ~/.texttest/config, create it if it doesn't exist), add the following to your checkout locations:
(You should replace $CLONE_LOCATION with an actual path or environment variable that is set in all shells)
[checkout_location]
gocd-pipeline-builder:$CLONE_LOCATION/gocd-pipeline-builder
[end]
start texttest:
texttest -a gocdpb
Since the tests change the state of the Go server, it's important to run them sequentially unless capturemock is in replay mode.
TODO: Document how to set up docker for the repo_checks tests.
Make sure that all tests pass and that the version has been updated in setup.py. Commit and push changes to the repository before doing the following.
$ python setup.py sdist
$ twine upload dist/gocdpb-<new version>.tar.gz
$ sudo pip install --upgrade gocdpb
Verify that the new version is downloaded, and works as intended.