A continuous delivery demo using Jenkins on DC/OS.
This demo is a Python script that, when run with the install
command, will:
- Installs Jenkins if it isn't already available.
Running the pipeline
command will:
-
Sets up a Jenkins "workflow" pipeline job and the necessary credentials to demonstrate a basic continuous delivery pipeline. Jenkins will:
- Spin up a new Jenkins agent using the Mesos plugin. This agent runs inside a Docker container on one of our DC/OS agents.
- Clone the git repository
- Build a Docker container based off the Jekyll Docker image that includes the content stored in /site and push it to DockerHub.
- Run the newly created container and a Linkchecker container that runs a basic integration test against the container, checking that the webserver comes up correctly and that all links being served are valid (i.e. no 404s).
- Manually trigger a Marathon deployment of the newly created container to the DC/OS base Marathon instance. If the application already exists, Marathon will simply upgrade it.
- Make the application available on a public agent at port 80 using Marathon-lb.
When run with the dynamic-agents
command, it will:
- Creates 50 build jobs that take a random amount of time between 1 and 2 minutes. These jobs will randomly fail.
- The Mesos plugin will spin up build agents on demand for these jobs, using as much capacity as your cluster has available.
- When these jobs are finished, the Jenkins tasks will terminate and the resources will be relinquished back to other users of your cluster.
When run with the uninstall
command, it will:
- Remove any persisted credentials, build job and view configurations.
- Uninstall Jenkins.
bin/demo.py --help
will show you full help text and usage information.
- GitHub account
- Python 3
pip3 install -r requirements.txt
Tip: If Python 3 is not the default on your system, you may need to specifically instruct pip to install the Python 3 versions of the requirements.
-
Clone this repository!
git clone https://github.com/mesosphere/cd-demo.git
-
Check out the latest stable tag, depending on DC/OS cluster version you're running against:
DC/OS 1.10+:
git checkout v1.8.7-2.7.2
DC/OS 1.9 and below:
git checkout v1.8.6-2.7.2
-
Create a branch from the latest stable tag, this is mandatory:
git checkout -b my-demo-branch git push origin my-demo-branch
-
Ensure you have a DC/OS cluster available. 1 node will work but more than 1 node is preferable to demonstrate build parallelism.
-
Export the demo Docker Hub password (NOT the DC/OS cluster password) to an environment variable. You will need to replace the password here with the password for the
cddemo
user with permission to push tomesosphere/cd-demo-app
(or your own repo, if you override the--org
and--username
flags later):export PASSWORD=mypass123
-
Run the install command. This is mainly a wrapper for the
dcos package install
command but will also check to see if you're authenticated.python3 bin/demo.py install http://my.dcos.cluster/
NOTE: You must use the domain name for your cluster; the IP address will fail.
-
You can now run either the pipeline demo or the dynamic agents demo. To run the pipeline demo, you will also need to specify the ELB address (
Public Agent
):python3 bin/demo.py pipeline --password=$PASSWORD http://my.elb/ http://my.dcos.cluster/
-
The script will first install Marathon-lb if it looks like it isn't available. It will also update the
marathon.json
in the branch you specified to include the ELB hostname so that Marathon-lb can route to it. -
The script will then use the Jenkins HTTP API to install jobs and necessary credentials. It will automatically trigger the initial build before finishing.
-
Navigate to the Jenkins UI to see the builds in progress. After a few seconds, you should see a build executor spinning up on Mesos. If you navigate to the job, you'll see the pipeline in progress.
-
The deploy will happen almost instantaneously. After a few seconds, you should be able to load the application by navigating to the ELB hostname you provided earlier in your browser.
-
Now let's run the dynamic agents demo. It will create 50 jobs that will randomly fail.
python3 bin/demo.py dynamic-agents http://my.dcos.cluster/
-
Navigate back to the Jenkins and/or DC/OS UI to show build agents spinning up manually.
-
Simply run the uninstall command to remove any persisted configuration and to uninstall the DC/OS service itself. This will allow you to run multiple demos on the same cluster but you should recycle clusters if the version of the Jenkins package has changed (to ensure plugins are upgraded):
python3 bin/demo.py uninstall http://my.dcos.cluster/
Alternatively, run the cleanup command instead to just remove jobs and to avoid having to re-install Jenkins.
By default, this script assumes you will be pushing to the mesosphere/cd-demo-app repository on Docker Hub.
-
Create a public repo called
cd-demo-app
under your example organisation. -
Create a Docker Hub user that has credentials to push to this repository.
-
Run the pipeline demo, passing in the credentials:
python3 bin/demo.py pipeline --org=myorg --username=myuser --password=$PASSWORD http://my.elb/ http://my.dcos.cluster/
-
Run the demo to completion. The pipeline will continue to monitor your branch after the script finishes:
python3 bin/demo.py pipeline --password=$PASSWORD http://my.elb/ http://my.dcos.cluster/
-
Create a new blog post with today's date, open it up in your text editor and make whatever changes you'd like to:
cp site/_posts/2016-02-25-welcome-to-cd-demo.markdown site/_posts/$(date +%Y-%m-%d)-my-test-post.markdown nano site/_posts/$(date +%Y-%m-%d)-my-test-post.markdown
-
Commit your changes and push them up to GitHub:
git add site/_posts/$(date +%Y-%m-%d)-my-test-post.markdown git commit -m "Demo change" git push origin my-demo-branch
-
Jenkins will pick up the change within a minute and kick off the pipeline. If you want to fail the build, simply insert a broken link into your post.
To demonstrate how you can install multiple Jenkins instances side by side on DC/OS, simply give your Jenkins instances unique names using the --name
argument and run the demo as follows. Note that if you only have one public agent, you will not be able to deploy applications from multiple pipelines (each application requires port 80).
-
Create one instance:
python3 bin/demo.py install --name=jenkins-1 http://my.dcos.cluster/
-
Open a new terminal tab and create a second instance and so on:
python3 bin/demo.py install --name=jenkins-2 http://my.dcos.cluster/
-
You can uninstall these in the same way:
python3 bin/demo.py uninstall --name=jenkins-1 http://my.dcos.cluster/ python3 bin/demo.py uninstall --name=jenkins-2 http://my.dcos.cluster/
The script will check to see if your current machine has a valid dcos_acs_token
set. If it doesn't:
-
it attempts to authenticate using the default username and password for an Enterprise DC/OS cluster. You can override these using the
--dcos-username
and--dcos-password
arguments. -
if this fails, it will attempt to use the
--dcos-oauth-token
arguments to authenticate against an Open DC/OS cluster.
If you would prefer to have an all-in-one demo experience, we now have a stage_cd_demo.sh script that will follow the default demo flow.
To run, the syntax is:
bash ./stage_cd_demo.sh <url-to-master> <url-of-public-elb>
This script will prompt you to walk through the demo, in the following order:
- Stage your local environment (install python dependencies, configure dcos cli, prompt you for the cd-demo docker password, install jenkins)
- Kick off the pipeline demo (install and configure marathon-lb, kick off initial pipeline demo build)
- Create a new post, commit to your branch, and trigger a pipeline rebuild by pushing the change to githhub
- Kick off dynamic load demonstration (launch 50 nodes in parallel, including a jenkins config change to speed up the time to launch new build executors)
- Clean up your github branch and cluster after the fact so you're ready for another runthrough
At each stage above, the script will pause and you will need to press ENTER to continue.
If you would prefer not to get prompted for the demo docker password, you can create a password.txt
file in the repo root directory (e.g. ~/repos/cd-demo/password.txt) with the password, which will override the manual prompt.
The password file will NOT be committed to your branch.
- This script is currently untested on Windows.