Tutorial for using rules

The rules system works with the XL Deploy planning phase. You can use XML or Jython to specify the steps that belong in a deployment plan and how the steps are configured.

This tutorial describes the process of using rules to create an new XL Deploy plugin.

The plugin actions:

  • Waits a specific interval to start the deployment.
  • Deploys and undeploys an artifact.
  • Starts and stops a server.

Requirements for you to use this tutorial:

Tip: The code provided in this tutorial is available as a demo plugin in the samples directory of your XL Deploy installation.

Run the examples

To run the examples in this tutorial, no specific configuration or plugin is required.

Required files

To configure XL Deploy to use the examples in this tutorial, you must add or modify the following files in the ext folder of the XL Deploy server:

  • synthetic.xml, which contains the configuration item (CI) types that are defined.
  • xl-rules.xml, which contains the rules that are defined.

Place the additional scripts that you will define in the ext folder.

The structure of the ext folder after you finish this tutorial:

ext/
├── planning
│   └── start-stop-server.py
├── scripts
│   ├── deploy-artifact.bat.ftl
│   ├── deploy-artifact.sh.ftl
│   ├── undeploy-artifact.bat.ftl
│   ├── undeploy-artifact.sh.ftl
│   ├── start.bat.ftl
│   ├── start.sh.ftl
│   ├── stop.bat.ftl
│   └── stop.sh.ftl
├── synthetic.xml
└── xl-rules.xml

Restarting the server

After you change synthetic.xml, you must restart the XL Deploy server.

By default, you must also restart the XL Deploy server after you change xl-rules.xml and scripts in the ext folder. You can configure XL Deploy to periodically rescan xl-rules.xml and the ext folder and apply any changes that it finds. Use this when you are developing a plugin. For more information, see Define a rule.

Error handling

If you make a mistake in the definition of synthetic.xml or xl-rules.xml, the server will return an error and may fail to start. Mistakes in the definition of scripts or expressions usually appear in the server log when you execute a deployment. For more information about troubleshooting the rules configuration, refer to Best practices for rules.

Deploy an artifact

Start with an application that contains one artifact and deploy the artifact to a server.

This part of the plugin:

  • Uploads the artifact.
  • Runs a script that installs the artifact in the correct location.

Add type definitions

In synthetic.xml, add a type definition called example.ArtifactDeployed for the application and add a container type named example.Server:

<type type="example.Server" extends="udm.BaseContainer" description="Example server">
    <property name="host" kind="ci" referenced-type="overthere.Host" as-containment="true"/>
    <property name="home" description="Home directory for the server"/>
</type>

<type type="example.ArtifactDeployed" extends="udm.BaseDeployedArtifact" deployable-type="example.Artifact" container-type="example.Server" description="Artifact that can be deployed to an example server">
    <generate-deployable type="example.Artifact" extends="udm.BaseDeployableFileArtifact"/>
</type>

Notes:

  • example.Server extends from udm.BaseContainer and has a host property that refers to a CI of type overthere.Host.
  • The deployed example.ArtifactDeployed extends from udm.BaseDeployedArtifact, which contains a file property that the step uses.
  • The generated deployable example.Artifact extends from udm.BaseDeployableFileArtifact.

Define a rule for the artifact

To define an XML rule for the CI in xl-rules.xml:

<rule name="example.ArtifactDeployed.CREATE_MODIFY" scope="deployed">
    <conditions>
        <type>example.ArtifactDeployed</type>
        <operation>CREATE</operation>
        <operation>MODIFY</operation>
    </conditions>
    <steps>
        <os-script>
            <script>scripts/deploy-artifact</script>
        </os-script>
    </steps>
</rule>

Notes:

  • The name example.ArtifactDeployed.CREATE_MODIFY identifies the rule in the system. Use a descriptive name that includes the name of the plugin and the type and operation the rule responds to.
  • The scope is deployed because this rule must contribute a step for every instance of example.ArtifactDeployed in the deployment.
  • The rule matches on deltas with the operations CREATE and MODIFY. Matching on CREATE means that this rule will trigger when XL Deploy knows that the application must be created or deployed. Matching on MODIFY means that the rule will contribute the same step to the plan upon modification.
  • The rule will create a step of type os-script, which can upload a file and run a templated script. The script defines the path where the script template is located, relative to the plugin definition.

The following os-script parameters are defined automatically:

  • A description that includes the artifact name and the name of the server it will deploy to. You can optionally override the default description.
  • The order, which is automatically set to 70, the default step order for artifacts. You can optionally override the default order.
  • The target-host property receives a reference to the host of the container. The step will use this host to run the script.

Script to deploy the artifact

The FreeMarker variable for the deployed object is automatically added to the freemarker-context. The script can refer to properties of the deployed object such as file location.

The script parameter refers to scripts for Unix (deploy-artifact.sh.ftl) and Windows (deploy-artifact.bat.ftl). The step will select the correct script for the operating system that XL Deploy runs on. The scripts are actually script templates processed by FreeMarker. The template can access the variables passed in by the freemarker-context parameter of the step.

The Unix script deploy-artifact.sh.ftl contains:

echo "Deploying file on Unix"
mkdir -p ${deployed.container.home + "/context"}
cp ${deployed.file.path} ${deployed.container.home + "/context"}
echo "Done"

The script accesses the variable deployed and uses it to find the location of the server installation and copy the file to the context folder. The script also prints progress information in the step log.

Add a wait step

You can improve the plan with an additional step that waits a specific number of seconds before the actual deployment starts.

  • While preparing the deployment, you can set the number of seconds to wait in the deployment properties.
  • If you do not set a number, XL Deploy will not add a wait step to the plan.

Add a property to type definition

You must store the wait time in the deployment properties by adding the following property to udm.DeployedApplication in synthetic.xml:

<type-modification type="udm.DeployedApplication">
     <property name="waitTime" kind="integer" label="Time in seconds to wait for starting the deployment" required="false"/>
</type-modification>

Define a rule to contribute a wait step

Define a rule in xl-rules.xml to contribute the wait step to the plan:

<rule name="example.DeployedApplication.wait" scope="pre-plan">
    <conditions>
        <expression>specification.deployedOrPreviousApplication.waitTime is not None</expression>
    </conditions>
    <steps>
        <wait>
            <order>10</order>
            <description expression="true">"Waiting %i seconds before starting the deployment" % specification.deployedOrPreviousApplication.waitTime</description>
            <seconds expression="true">specification.deployedOrPreviousApplication.waitTime</seconds>
        </wait>
    </steps>
</rule>

Notes:

  1. The scope is pre-plan. This means that:

    • The rule will only trigger once per deployment.
    • The step that the rule contributes is added to the pre-plan, which is a sub-plan that XL Deploy prepends to the deployment plan.
  2. Only contribute a step to the plan when the user supplies a value for the wait time. There is a condition that checks if the waitTime property is not None. The expression must be defined in Jython.
  3. If the condition holds, XL Deploy creates the step that is defined in the steps section and adds it to the plan. The step takes arguments that you specify in the rule definition:

    • The order is set to 10 to ensure that the rule will appear early in the plan. In this case, this will be the only step in the pre-plan, so the order value can be ignored. You must provide this required value for the wait step. The type of order is integer, so if it has a value that is not an integer, planning will fail.

      • description is a dynamically constructed string that describes what the step will do. Providing a description is optional. If you do not provide one, XL Deploy will use a default description.

        • expression="true" means that the definition will be evaluated by Jython and the resulting value will be passed to the step. This is required because the definition contains a dynamically constructed string.
    • The waitTime value is retrieved from the DeployedApplication and passed to the step. You can access the DeployedApplication through the specification and deployedOrPreviousApplication. This automatically selects the correct deployed, which means that this step will work for a CREATE or DESTROY operation.

For more information about the wait step, see Steps Reference.

Test the deployment rules

To test the rules that you created:

  1. Install XL Deploy and the scripts as described in How to run the examples.
  2. Under Applications in the left pane, create an application that contains a deployable of type example.Artifact. Upload a dummy file when creating the deployable CI.
  3. Under Infrastructure, create a host of type overthere.LocalHost and a container of type example.Server. Set the home directory of example.Server to a temporary location.
  4. Under Environments, create an environment that contains the example.Server container.
  5. Start a new deployment of the application to the environment. When preparing the deployment, click Deployment Properties and enter a wait time. If you do not provide a value, the wait step will not appear in the plan.

    Deployment properties

  6. Click Modify Plan or Deploy. XL Deploy will create the following deployment plan:

    First deployment plan

  7. Execute the plan. Check that the steps are succesful.
  8. Verify that there is a context folder in the directory that you set as the home directory of example.Server, and verify that the artifact was copied to it.

The folder structure should be similar to:

$ tree /tmp/srv/
/tmp/srv/
└── context
    └── your-file.txt

Undeploy an artifact

When you create rules to deploy packages, you should also define rules to undeploy them. Forthis plugin, undeployment removes the artifact that was deployed. The rule will use the state of the deployment to determine which files must be deleted.

Define an undeploy rule

The rule definition in xl-rules.xml is:

<rule name="example.ArtifactDeployed.DESTROY" scope="deployed">
    <conditions>
        <type>example.ArtifactDeployed</type>
        <operation>DESTROY</operation>
    </conditions>
    <steps>
        <os-script>
            <script>scripts/undeploy-artifact</script>
        </os-script>
    </steps>
</rule>

Notes:

  • The operation is DESTROY.
  • XL Deploy automatically sets the order and description.
  • The step is an os-script step. The script behind the step is responsible for deleting the file on the server.

Undeploy script

The FreeMarker variable for the previousDeployed object is automatically added to the freemarker-context. This allows the script to refer to the properties of the previous deployed object such as file name.

The Unix script undeploy-artifact.sh.ftl contains:

echo "Undeploying file on Unix"
rm ${previousDeployed.container.home + "/context/" + previousDeployed.file.name}
echo "Done"

Test the undeploy rule

After successfully deploying the artifact, roll back the deployment or undeploy the application. If you have defined undeployment rules for all deployeds or used the sample code provided by XebiaLabs, the deployment plan looks like this:

Undeploy plan

Restart the server

Restarting the server is an advanced procedure because it requires a script rule, which is written in Jython.

You created a rule that copies an artifact to the server. To correctly install the artifact, you must stop the server at the beginning of the deployment plan and start it again in the end. This requires two more steps:

  • One script that stops the server by calling the stop script
  • One script that starts the server by calling the start script

Note: A full implementation requires four scripts:

  • One script that stops the server for Unix
  • One script that starts the server for Unix
  • One script that stops the server for Windows
  • One script that starts the server for Windows

Define a restart rule

The script rule is defined in xl-rules.xml as follows:

<rule name="example.Server.startStop" scope="plan">
    <planning-script-path>planning/start-stop-server.py</planning-script-path>
</rule>

Notes:

  • The scope is plan because the script must inspect all deployeds of the specific sub-plan to make its decision. The rule contributes one start step and stop step per sub-plan, and rules with the plan scope are only triggered once per sub-plan.
  • The rule has no conditions because the script will determine if the rule will contribute steps.
  • The rule refers to an external script file in a location that is relative to the plugin definition.

Restart server script

The script start-stop-server.py contains:

from java.util import HashSet

def containers():
    result = HashSet()
    for _delta in deltas.deltas:
        deployed = _delta.deployedOrPrevious
        current_container = deployed.container
        if _delta.operation != "NOOP" and current_container.type == "example.Server":
            result.add(current_container)
    return result

for container in containers():
    context.addStep(steps.os_script(
        description="Stopping server %s" % container.name,
        order=20,
        script="scripts/stop",
        freemarker_context={'container': container},
        target_host=container.host)
    )
    context.addStep(steps.os_script(
        description="Starting server %s" % container.name,
        order=80,
        script="scripts/start",
        freemarker_context={'container': container},
        target_host=container.host))

Note: The freemarker_context={'container': container} is required to make the container object available in the FreeMarker context.

The rules demo plugin also includes a dummy script called start.sh.ftl that contains:

echo "Starting server on Unix"

In a real implementation, this script must contain the commands required to start the server.

  • The script starts with:

    • An import statement of an utility class.
    • The method definition of containers().
    • A loop that iterates all containers and creates steps; this is the starting point of the code that will be executed.
  • The containers() method determines which containers need to be restarted and collects them in a set. The set data structure prevents duplicate start and stop steps.

    • The method iterates over the deltas and selects the deployed with deployedOrPrevious, regardless if it is DESTROY, CREATE, and so on.
    • It retrieves the container of the deployed and stores it in current_container.
    • The container is added to the set of containers that must be restarted if:

      • The operation is not NOOP. You perform actions when the operation is CREATE, MODIFY, or DESTROY.
      • The type of the container is example.Server. This rule will be triggered for every plan and every deployment. Ensure that the delta is related to a relevant container.
  • The script iterates over all containers that must be restarted.

    • The freemarker_context map contains a reference to the container.
    • In the start and stop steps, the steps factory is used to construct the steps by name. Notes:

      • The os_script step is used to execute the script.
      • The Jython naming convention (with underscores) is used to refer to the step.
      • The orders for the stop (20) and start (80) steps will ensure that they are ran before and after the deployment of the application.
      • Use the addStep method to add the constructed step directly to the context.
  • If XL Deploy does not find deltas for the sub-plan, the start and stop steps will not be created.

Test the server restart

To test the server restart rules, set up a deployment as described in Test the deployment rules. The deployment plan should look like:

Final plan

Note: The steps to start and stop server are added even when application is undeployed:

Final plan

Roll back a deployment

The plugin that you create when following this tutorial does not require any extra updates to support rollbacks. XL Deploy automatically generates checkpoints for the last step of each deployed. When a user rolls back a deployment that has only been partially executed, the roll back plan will contain the steps for the opposite deltas of the deployeds for which all steps have been executed.

If you have more advanced rollback requirements, see Using checkpoints.

Next steps

After finishing this tutorial, you should have a good understanding of rules-based planning, and you should be able to find the information you need to continue creating deployment rules.

The code presented in this tutorial is available in the rules demo plugin, which you can find in the samples directory of your XL Deploy installation. The demo plugin contains additional examples.

If you want to change the behavior of an existing plugin, you can disable predefined rules and redefine the behavior with new rules. For more information about this, see Disable a rule.