Mark van Holsteijn 06 Apr, 2015
Share 
You can use fleetctl and script files to deploy your container applications to CoreOS. However, using XL Deploy for deployment automation is a great solution when you need to deploy and track versions of many applications. What does it take to create a XL Deploy plugin to deploy these container applications to your CoreOS clusters?
XL Deploy can be extended with custom plugins which add deployment capabilities. Using XL Rules custom plugins can be created quickly with limited effort. In this blog you can read how a plugin can be created in a matter of hours.
In a number of blog posts, Mark van Holsteijn explained how to create a high available Docker container platform using CoreOS and Consul. In these posts shell scripts (with fleetctl commands) are used deploy container applications. Based on these scripts I have built a XL Deploy plugin which deploys fleet unit configuration files to a CoreOS cluster.
Deploying these container applications using XL Deploy has a number of advantages:
- Docker containers can be deployed without creating, adjusting and maintaining scripts for individual applications.
- XL Deploy will track and report the applications deployed to the CoreOS clusters.
- Additional deployment scenarios can be added with limited effort.
- Deployments will be consistent and configuration is managed across environments.
- XL Deploy permissions can be used to control (direct) access to the CoreOS cluster(s).
Building an XL Deploy plugin is fast, since you can:
- Reuse existing XL Deploy capabilities, like the Overthere plugin.
- Utilize XL Deploy template processing to inject property values in rules and deploy scripts.
- Exploit the XL Deploy unified deployment model to get the deltas which drive the required fleetctl deployment commands for any type of deployment (new, update, undeploy and rollback deployments).
- Use xml and script based rules to build deployment tasks.
Getting started
- Install XL Deploy, you can download a free edition here. If you are not familiar with XL Deploy, read the getting started documentation.
- Next add the plugin resources to the ext directory of your XL Deploy installation. You can find the plugin resources in this Github repository. Add the synthetic.xml, xl-rules.xml file from the repository root. In addition, add the scripts directory and its contents. Restart XL Deploy.
- Next, setup a CoreOS cluster. This blog post explains how you can setup such a platform locally.
- Now you can connect to XL deploy using your browser. On the deployment tab you can import the sample application, located in the sample-app folder of the plugin resources Github repository
- You can now setup the target deployment container based on the Overthere.SshHost configuration item type. Verfiy that you can connect to your CoreOS cluster using this XL Deploy container.
- Next, you can setup a XL Deploy environment, which contains your target deployment container.
- Now you can use the deployment tab to deploy and undeploy your fleet configuration file applications.
Building the plugin
The plugin consists of two xml files and a number of script files. Below you find a description of the plugin implementation steps.
The CoreOS container application deployments are based on fleet unit configuration files. So, first we create a XL Deploy configuration Item type definition which represents such a file. This XL Deploy deployed type is defined in the XL Deploy synthetic.xml file. The snippet below shows the contents of this file. I have assigned the name “fleet.DeployedUnit”.
The definition contains a container-type attribute. The Overthere.SshHost container is referenced. The plugin can simple use the Overthere.SshHost container type to connect to the CoreOS cluster and to execute fleet commands.
Furthermore, I have added two properties. One property which can be used to specify the number of instances. Note that XL Deploy dictionaries can be utilized to define the number of instances for each environment separately. The second property is a flag which controls whether instances will be started (or only submitted and loaded).
If you want to deploy a fleet configuration file using fleetctl, you can issue the following three commands: submit, load and start. In the plugin, I have created a separate script file for each of these fleetctl commands. The caption below shows the script file to load a fleet configuration file. This load script uses the file name property and numberOfInstances property of the “fleet.DeployedUnit” configuration item.
Finally, the plugin can be completed with XML-based rules which create the deployment steps. The caption below shows the rule which adds steps to [1] submit the unit configuration and [2] load the unit when (a version of ) the application is deployed.
Using rules, you can easily define logic to add deployment steps. These steps can closely resemble the commands you perform when you are using fleetctl. For this plugin I have utized xml-based rules only. Using script rules, you can add more intelligence to your plugin. For example, the logic of the restart script can be converted to rules and more fine grained deployment steps.
More information
If you are interested in building your own XL Deploy plugin, the XL Deploy product documentation contains tutorials which will get you started.
If you want to know how you can create a High Available Docker Container Platform Using CoreOS And Consul, the following blogs are great starting point:
