{section}
{column}
h1. On the Petals ESB Topology

Petals ESB is a fully distributed ESB that features many types of topology both dynamically or statically. The topology of a Petals ESB cluster can be configured through a topology definition (ie. {{topology.xml}}) and [server properties|Container Configuration] (ie. {{server.properties}}).

This document will explain to you how to properly configure the topology of your Petals ESB cluster.

h1. Topology composition

A Petals ESB cluster is composed of a *domain* divided into *sub-domains* that contain the *containers*. A sub-domain aims to limit the visibility of service endpoints.
Inside a sub-domain:
* all service providers can be seen by all service consumers without restrictions,
* all service providers must be accessible directly to all service consumers.

Using a gateway, it is possible to interact between sub-domains pairs:
* by default, service endpoints of a sub-domain are not propagated to other sub-domains,
* the gateway is configured to propagated a selection of service endpoints in both ways,

A Petals sub-domain supports two different modes :
* static : no new container can be added to a running Petals sub-domain. All containers of the sub-domain must be stop and restart to take into account changes
* dynamic : a container can be attached to or detached from a running sub-domain, without requiring a full start/stop of the sub-domain

h2. Choosing the sub-domain mode

The sub-domain mode 'static' is recommended in following use-case:
* you want to prevent to add or remove container from your sub-domain,
* you don't plan hot topology refactoring

If you have a worst constraint about the availability of the Petals ESB (24/24, 7/7), you shoudl use the mode 'dynamic' to be able to apply topology changes without stopping the Petals ESB.


h2. Using the static mode

To use the static mode, in the sub-domain element of the topology definition, you just have to change the mode property to *static* :

{code:lang=xml}<tns:sub-domain name="your-subdomain-name" mode="static">{code}

See : [^static-topology-sample.xml]

h2. Using the dynamic mode

To use the dynamic mode, in the sub-domain element of the topology definition, you just have to change the mode property to *dynamic* :

{code:lang=xml}<tns:sub-domain name="your-sub-domain-name" mode="dynamic">{code}

See : [^dynamic-topology-sample.xml]

h1. Configuring the subdomain mode

Apart from configuring Petals ESB domain mode, you can also, for each subdomain, choose various modes :
* *standalone* : the subdomain can only have one node.
* *flooding* : the subdomain can have many nodes, each one knowing each other.
* *master-slave* : the subdomain can have many nodes, each node knows only one node called the master node.

The domain mode doesn't impact the way Petals ESB nodes communicate business data with each other. This only change the way endpoints-related informations are exchanged between nodes.

This chapter will explain you how to choose and configure the subdomain mode according your needs.

h2. Standalone configuration

In standalone mode, the Petals ESB cluster is limited to one node. You can only have one subdomain, also in standalone mode, with only one node, again declared as working in standalone mode.

{center}!standalone.png|height=285,width=500!{center}

{warning:title=Warning}
Again, since Petals ESB 3.x, the standalone mode is *not supported anymore*. Instead, use a the static domain with one subdomain in master/slave mode. The only node of the cluster should then be configured in master mode.
{warning}

h2. Flooding configuration

In flooding mode, each time a new service unit (registering endpoints) is installed, the node on which it has been installed propagates immediately the modification to all its peers. Business data of course is still exchanged directly between nodes, from the source directly to the destination.

{center}
!flooding.png|height=375,width=500!
{center}

A subdomain configured to use the flooding strategy can only contain node working in *peer* mode. The minimum number of peer node is one, while there is not limit on the number of peer nodes.

{warning:title=Warning}
This mode is only available when the whole domain (eg. the whole cluster) is in *static* mode.
{warning}

h2. Master/Slave configuration

In master-slave mode, every node of the cluster resynchronises itself regularly against the configuration of the master node. For now on, each node knows every all the endpoints and services deployed on the cluster. In the future, this strategy will be the only one to be maintained and will be refined to control the propagation of the informations on the cluster. As ever, business data is still exchanged directly between nodes, from the source directly to the destination.

{center}!master-slave.png|height=375,width=500!{center}

A subdomain using the master/slave strategy can only contains nodes using master or slave nodes. Each subdomain using this mode must have at least one master node. The number of slave nodes is unbounded.

{note:title=Note}
This strategy can be used both in dynamic or static domain mode. In static domain mode, every node of the cluster should be declared within the {{topology.xml}} file while in dynamic mode, you can only register the master node and the local node.
{note}

h1. Declaring containers configurations

As we have seen, the {{topology.xml}} file is used to declare the configuration of a Petals ESB cluster.

h2. Generic container configuration

Each Petals ESB container declared within must conforms to the following XML structure :

{code:lang=xml}
<tns:container name="0" type="master">
<tns:description>description of the container 0</tns:description>
<tns:host>localhost</tns:host>
<tns:user>petals</tns:user>
<tns:password>petals</tns:password>
<tns:webservice-service>
<tns:port>7600</tns:port>
<tns:prefix>petals/ws</tns:prefix>
</tns:webservice-service>
<tns:jmx-service>
<tns:rmi-port>7700</tns:rmi-port>
</tns:jmx-service>
<tns:transport-service>
<tns:tcp-port>7800</tns:tcp-port>
</tns:transport-service>
<tns:registry-service>
<tns:port>7900</tns:port>
</tns:registry-service>
</tns:container>
{code}

The meaning of these properties is summarized in the following table :

{center}
|| Property name || Default value || Required || Description ||
| {{name}} | 0 | yes | name of the container |
| {{mode}} | master | yes | mode of the node (master, slave, peer, standalone) |
| {{description}} | description of the container 0 | no | description of the container |
| {{host}} | localhost | yes | host name or ip address of the container |
| {{user}} | petals | no | jmx username |
| {{password}} | petals | no | jmx password |
|| {{webservice-service}} || || || ||
| {{port}} | 7600 | yes | administration webservices port (topology, information, ...) |
| {{prefix}} | petals/ws | yes | path of the webservice |
|| {{jmx-service}} || || || ||
| {{rmi-port}} | 7700 | yes | jmx service port |
|| {{transport-service}} || || || ||
| {{tcp-port}} | 7800 | yes |NIO transporter port |
|| {{registry-service}} || || || ||
| {{port}} | 7900 | yes | registry service port |
{center}


h2. Standalone container configuration

A container is declared to be working in standalone mode using the following XML code :

{code:lang=xml}
<tns:container name="0" type="standalone">
...
</tns:container>
{code}

{warning:title=Warning}
Again, since Petals ESB 3.x, the standalone mode is *not supported anymore*. Instead, use a the static domain with one subdomain in master/slave mode. The only node of the cluster should then be configured in master mode.
{warning}


h2. Peer container configuration

A container is declared to be working in peer mode using the following XML code :

{code:lang=xml}
<tns:container name="0" type="peer">
...
</tns:container>
{code}

This kind of containers can only be found in subdomains using the flooding strategy.


h2. Master container configuration

A container is declared to be working in master mode using the following XML code :

{code:lang=xml}
<tns:container name="0" type="master">
...
</tns:container>
{code}

A domain using the master-slave strategy must have at least one and only one master node.

h2. Slave container configuration

A container is declared to be working in mode mode using the following XML code :

{code:lang=xml}
<tns:container name="0" type="slave">
...
</tns:container>
{code}

A domain using the master-slave strategy can have an unbounded number of slave nodes.

When using the dynamic domain mode, for a slave node, you can declare only the master configuration and the slave configuration. The first topology update will rewrite the {{topology.xml}} file and update the containers list using the ones registered to the master node.

{warning:title=Warning}
You can declare two slaves node using the same name with different configuration, but in this case, the master node will only remembers the latest one.
{warning}

h1. Configuring the local container

As we have seen, the {{topology.xml}} file contains the declaration of a Petals ESB cluster. This information is not sufficient to determine which configuration must be chosen by the local container at startup. To do so, you must give a container name declared within the {{topology.xml}} file to the property {{petals.container.name}} declared in the {{server.properties}}.

{noformat}
#This property specifies the name of the container. In distributed mode, this property is mandatory
#and must match a container name in the topology.xml file
petals.container.name=0
{noformat}

The container name given to the {{petals.container.name}} property must be declared within the {{topology.xml}} file, otherwise, Petals ESB won't be able to start. This mean that you must have declared the local container configuration within the {{topology.xml}} file.

h1. Summary

The table below summarize the valid configuration for the {{topology.xml}} file :

{center}
|| Domain mode || Subdomain mode || Container mode|| Supported || Won't be deprecated ||
| standalone | standalone | standalone | (x) | (x) |
| dynamic | master-slave | master, slave | (/) | (/) |
| static | flooding | peer | (/) | (x) |
| static | master-slave | master, slave | (/) | (/) |
{center}

h1. Appendix

h2. Default configuration

The default {{topology.xml}} file is :

{code:lang=xml}
<?xml version="1.0" encoding="UTF-8"?>
<tns:topology xmlns:tns="http://petals.ow2.org/topology"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://petals.ow2.org/topology petalsTopology.xsd">
<tns:domain mode="static" name="PEtALS">
<tns:description>The static domain configuration</tns:description>
<tns:sub-domain name="subdomain1" mode="master-slave">
<tns:description>description of the subdomain</tns:description>
<tns:container name="0" type="master">
<tns:description>description of the container 0</tns:description>
<tns:host>localhost</tns:host>
<tns:user>petals</tns:user>
<tns:password>petals</tns:password>
<tns:webservice-service>
<tns:port>7600</tns:port>
<tns:prefix>petals/ws</tns:prefix>
</tns:webservice-service>
<tns:jmx-service>
<tns:rmi-port>7700</tns:rmi-port>
</tns:jmx-service>
<tns:transport-service>
<tns:tcp-port>7800</tns:tcp-port>
</tns:transport-service>
<tns:registry-service>
<tns:port>7900</tns:port>
</tns:registry-service>
</tns:container>
</tns:sub-domain>
</tns:domain>
</tns:topology>
{code}

Download link : [^default-topology.xml]

h2. Topology XML Schema

For more information on the structure of the topology.xml file, you can consult its XML schema definition here : [^petalsTopology.xsd].
{column}
{column:width=350px}
{panel:title=Table of contents}{toc}{panel}
{panel:title=Contributors}{contributors:order=name|mode=list}{panel}
{column}
{section}