A juju-deployer based test runner for bundles and charms.
pip2 install bundletester
This is designed around the fail fast principle. Each bundle is composed of charms, each of those charms should have tests ranging from unit tests to integration tests. The bundle itself will have integration tests. Those tests should run from least expensive to most expensive in terms of time, failing as soon as possible. The theory is that if tests fail in the charms, the bundle cannot function. Bundletester will therefore pull all the charms and look for any tests it can find and run those prior running its own integration tests.
This test runner uses the following pattern. A bundle has a URL. The runner can be run directly from within a bundle checkout or will pull the branch to which the bundle url refers. From there it will attempt to look for a 'tests' directory off the project root. Using the rules described below it will find and execute each test within that directory and produce a report.
This also includes a bundlewatcher script which can be used as a Jenkins script trigger.
From within the top level of a bundle or charm:
bundletester -r json -o result.json
Using a specific bundle or charm directory:
bundletester -t ~/charms/trusty/mycharm -l DEBUG
Passing -l/--log-level DEBUG will give additional insight into what steps bundletester is taking.
bundletester
can fetch and run tests from remote locations:
bundletester -t cs:trusty/wordpress
The -t
option accepts a variety of URL types:
-t cs:wordpress
-t cs:precise/wordpress
-t bundle:mediawiki-single
-t bundle:mediawiki-single-8
-t bundle:~charmers/mediawiki-single
-t bundle:~charmers/mediawiki-single-8
-t lp:~charmers/charms/precise/ghost/trunk
-t launchpad:~charmers/charms/precise/ghost/trunk
-t https://launchpad.net/~charmers/charms/precise/ghost/trunk
# Add '@revision' to any Launchpad URL to test a specific revision
-t lp:~charmers/charms/precise/ghost/trunk@4
# Test charm or bundle merge proposal
-t lp:~davidpbritton/charms/precise/apache2/avoid-regen-cert/+merge/221102
-t gh:charms/apache2
-t github:charms/apache2
-t https://github.com/charms/apache2
# Add '@revision' to any Github URL to test a specific revision
-t https://github.com/charms/apache2@52e73d
# For Bitbucket, repos that don't end in '.git' are assumed to be Hg
-t bb:battlemidget/juju-apache-gunicorn-django.git
-t bitbucket:battlemidget/juju-apache-gunicorn-django.git
-t https://bitbucket.org/battlemidget/juju-apache-gunicorn-django.git
# Add '@revision' to any Bitbucket URL to test a specific revision
-t bb:battlemidget/juju-apache-gunicorn-django.git@daff5d9
The driver file tests/tests.yaml
is used (by default) to control the overall
flow of how tests work. All values in this file (and indeed the file itself) are
optional. When not provided, defaults will be used.
You can override the tests/tests.yaml
file in a charm or bundle by
using the -y
argument to bundletester, e.g.:
bundletester -t cs:meteor -y /path/to/custom/tests.yaml
A sample tests.yaml
file::
bootstrap: false
reset: false
virtualenv: true
virtualenv_python: python3
tests: "[0-9]*"
excludes:
- `charm-name`
sources:
- ppa:ubuntu-lxc/lxd-stable
packages:
- lxd
python_packages:
- bzr
- juju-deployer
- amulet
- requests
requirements:
- test-requirements.txt
- requirements.txt
makefile:
- lint
- test
setup: `filename`
teardown: `filename`
bundle_deploy: false
deployment_timeout: 2700
Explanation of keys:
bootstrap: Bootstrap the environment if necessary (default: true).
reset: Use juju-deployer to reset the model between each test file execution (default: true).
reset_timeout: Max time (in seconds) allowed for each of two routines in a model reset: machine termination, and application removal. Total wait time can be 2X this value. This option has no effect if reset
has any value other than true
(default: 180).
virtualenv: Create and activate a virtualenv in which all tests are run (default: false).
virtualenv_python: The version of python with which to create the
virtualenv (if virtualenv
is `true'). Examples: python, python2.7,
python3.5 (default: python).
tests: A glob pattern of executable files in the tests/
directory to treat as tests (default: "*"). Only files that match this pattern will be executed.
excludes: List of charm names for which tests should be skipped. Useful if executing against a bundle.
sources: List of apt package sources to add before installing packages.
packages: List of packages to install with apt before running tests.
python_packages: List of python packages to install with pip install -U
before running tests. If virtualenv
is true
, the packages will be installed in the virtualenv.
requirements: List of pip requirements file names (relative to the
charm or bundle root dir), which will be
passed to pip install -r
. If virtualenv
is true, the packages will
be installed in the virtualenv, otherwise they will be installed at the
system level. (default: [])
makefile: List of make targets to execute (default: [lint, test]).
setup: Optional name of a script in the tests/
directory to run before each test.
teardown: Optional name of script in the tests/
directory to run after each test.
bundle_deploy: Only applies when testing a bundle. Acceptable values are true
, false
, or a file name. If true
(the default), the bundle will be deployed by juju-deployer before any tests are run. If false
, the deployment step will be skipped. If a filename is given, it should be an executable file in the tests/
directory (likely an amulet script). The file will be executed as the deployment step. Useful if you want to deploy the bundle, but need to modify it first. Note that if the bundle_deploy
filename matches the tests
glob pattern, it will be executed twice - once in the deploy step, and once as a test. To prevent this, use a tests
glob pattern that doesn't match the bundle_deploy
file name.
deployment_timeout: Max time (in seconds) allowed for the initial bundle
deploy (before any tests are run). This option has no effect if bundle_deploy
has any value other than true
(default: 2700).
When tests.yaml's test pattern is executed it will yield a test which should be run. Each
of these test files can optionally have a control file of the same name but with a .yaml
extension. If present, this file has a similar layout to tests.yaml.
A sample 01-test.yaml
file::
bundle: bundle.yaml
setup: script
teardown: script
Explanation of keys:
bundle: A bundle file name, relative to the tests directory. If bundle isn't specified <BUNDLE_ROOT>/bundle.yaml will be used by default.
setup: An optional script to be run before this test. This is called after the tests.yaml setup, if present.
teardown: An optional script to be run after this test. This is called before the tests.yaml teardown, if present.
If these scripts fail with a non-zero exit code, the test will be recorded as a failure with that exit code.
Each test should be executable. If virtualenv
is true, each test will be
executed with that environment activated.
Each test will have its stdout and stderr captured.
Each test's exit code will be captured. A non-zero return indicates failure.
Any test failures will result in a non-zero exit code.
Using -r json a JSON string will be printed on stdout. This string will contain, at minimum, the following structure (additional keys are possible):
[{test result}, ...]
each test result is a dict with at least:
{'test': test file, result: exit_code,
'exit': 'script name which returned exit code',
'returncode': exitcode of process,
'duration': timedelta in seconds}
exit
will not be included if result is sucess (0)
- Finish Tests, started as TDD and then hulk-smashed the end
- Better runtime streaming