Difference between revisions of "Gyrex/Developer Guide/Roles"

From Eclipsepedia

Jump to: navigation, search
(Activation Triggers)
 
Line 7: Line 7:
  
 
== Define Roles ==
 
== Define Roles ==
Roles are implemented using the Equinox Extension Registry. This decision was made in order to allow a declaritive approach for devining roles without writing any line of Java as well as to support lazy activation. It may be possible to use DS in the future ({{bug|395332}}).
+
Roles are implemented using the Equinox Extension Registry. This decision was made in order to allow a declaritive approach for defining roles without writing any line of Java as well as to support lazy activation. It may be possible to use DS in the future ({{bug|395332}}).
  
 
In order to define a role you need to create an extension to the <code>org.eclipse.gyrex.server.roles</code> extension point. Each role consists of an identifier, a name and references a set of bundles and/or applications to start. Applications must be defined using the Equinox <code>org.eclipse.core.runtime.applications</code> extension point.
 
In order to define a role you need to create an extension to the <code>org.eclipse.gyrex.server.roles</code> extension point. Each role consists of an identifier, a name and references a set of bundles and/or applications to start. Applications must be defined using the Equinox <code>org.eclipse.core.runtime.applications</code> extension point.

Latest revision as of 03:46, 29 November 2012

In Gyrex roles are used to control what bundles and what OSGi applications (OSGi application admin) will be started on a particular node.

Contents

[edit] Dependencies

  • Bundle org.eclipse.gyrex.boot (Gyrex 1.0 or higher)
  • Equinox Extension Registry


[edit] Define Roles

Roles are implemented using the Equinox Extension Registry. This decision was made in order to allow a declaritive approach for defining roles without writing any line of Java as well as to support lazy activation. It may be possible to use DS in the future (bug 395332).

In order to define a role you need to create an extension to the org.eclipse.gyrex.server.roles extension point. Each role consists of an identifier, a name and references a set of bundles and/or applications to start. Applications must be defined using the Equinox org.eclipse.core.runtime.applications extension point.

[edit] Example

The following Example defines a role "My Server Role".

<extension point="org.eclipse.gyrex.server.roles">
  <role id="com.abc.my.role" name="My Server Role">
    <requireBundle symbolicName="com.abc.thridparty" />
    <requireApplication applicationId="com.abc.my.application" />
  </role>
</extension>

The role requires the bundle com.abc.thridparty to be started as well as the application com.abc.my.application.


[edit] Activation Triggers

Once roles are defined you also need to define an activation trigger. You can create different triggers depending on the operation mode and the cloud online state. An addition start level can be set to apply some level of ordering to roles activation.

The following activation triggers are possible:

  • trigger="onBoot" ... a node is started
  • trigger="onCloudConnect" ... a node has successfully established its connection to the cluster
  • mode="development" ... a node is operating in development mode
  • mode="production" ... a node is operating is in production mode
  • nodeFilter="(...)" ... a node matches a custom LDAP style filter (see org.eclipse.gyrex.cloud.environment.INodeEnvironment.matches(String))

It's possible to combine different attributes together.

The trigger="..." defines when a role is processed.

onBoot means that roles will be started immediately when a node is started and a role becomes available. Thus, when a role is available at node start it will be started right away. When a role is installed (using p2) into a running node it will be started immediately after its successful installation. Any onBoot roles will be stopped when they are either uninstalled or the node is stopped.

onCloudConnect means that roles will be started when the node has successfully established a connection to the cluster. hen a role is installed (using p2) into a node that is already connected it will be started immediately after its successful installation. Any onCloudConnect roles will be stopped when they are either uninstalled or the node looses is connection to the cluster. They will be restarted automatically when the connection is established again.

[edit] Example Development Trigger

In development mode it's usually very comfortable to automatically start a role. For example, Gyrex itself uses the following trigger to automatically start an embedded ZooKeeper server in development mode.

<extension point="org.eclipse.gyrex.server.roles">
  <defaultStart
      mode="development"
      roleId="org.eclipse.gyrex.cloud.roles.leader"
      startLevel="-10"
      trigger="onBoot">
  </defaultStart>
</extension>

[edit] Example Production Trigger

In production environment one usually does not want a particular role to start automatically but only when a specific configuration is done. For example, Gyrex itself uses the following trigger to start the Jetty server in production mode when a node is tagged with the tag webserver and successfully connected with the cloud.

<extension point="org.eclipse.gyrex.server.roles">
  <defaultStart
      mode="production"
      nodeFilter="(tag=webserver)"
      roleId="org.eclipse.gyrex.http.jetty.roles.engine"
      startLevel="10"
      trigger="onCloudConnect">
  </defaultStart>
</extension>