Grid Failover
What does the sample do?
This demo illustrates a JPPF grid failover mechanism which ensures that a JPPF driver is not a single point of failure.
How does it work?
The failover mechanism relies on two main components:
- the nodes have a specific node connection strategy
which handles an ordered list of drivers to connect to, as follows:
- the first driver in the list is considered the primary driver, all others are backups in case the primary goes down
- when initially connecting or reconnecting due to a mangement request, the node attemps to connect
to the first (highest priority) driver in the list, then goes down the list if the driver is not online
- when reconnecting due to a disconnection from the driver, the node goes down the list of drivers,
rolling back to the top when the end of the list is reached
- a JPPF client is used as a controller, to detect when a driver is back up after going down, and decide whether it is the new primary and force the nodes to reconnect to it.
This client is used as follows:
- it is configured
to connect to the same drivers as provided to the nodes. Each configured driver has a priority
in descending order of the list given to the nodes
- for each driver connection, a connection status listener is registered
- the listener detects situations when the primary driver goes down or gets back up and determines which dfriver is the current primary
- when a new primary driver goes back up, the listener issues a management request to force the nodes to reconnect, once they complete their remaining tasks
How do I run it?
- Before running this sample application, you must have at least two JPPF servers running.
For information on how to set up a node and server, please refer to the JPPF documentation.
- Update the drivers definition file to reflect the information on your JPPF servers
- build the sample with this command: ant zip. This will create a file named jppf-grid-failover.zip
- configure a node to use the driver discovery:
- unzip the jppf-grid-failover.zip file into the root installation directory of a node. It will copy the required libraries into the node's /lib directory,
and the drivers.yaml file in the node's root installation directory.
- in a text editor, open the node's configuration file located at <node_install_root>/config/jppf-node.properties and add this line:
jppf.server.connection.strategy = org.jppf.example.gridfailover.NodeSideDiscovery
- start the node: ./startNode.sh or startNode.bat
- configure an administration console to use the driver discovery:
- unzip the jppf-grid-failover.zip file into the root installation directory of an adminsitration console. It will copy the required libraries into the console's /lib directory,
and the drivers.yaml file in the console's root installation directory.
- in a text editor, open the admin console's configuration file located at <console_install_root>/config/jppf-gui.properties and add this line:
jppf.remote.execution.enabled = false
this will disable the default built-in discovery mechanism, such that only our custom one is used
- the client-side discovery auto-installs through the Service Provider Interface (SPI), thus no configuration update is needed for this
- start the admin console: ./startConsole.sh or startConsole.bat
- once started, the admin console should show the drivers defined in the drivers.yaml file, along with a node connected to the first driver in the list
- from this sample's root directory, start the sample grid controller: ./run.sh or run.bat. The console output will show something similar to this:
client process id: 37556, uuid: A9619789-6E4B-4DA9-B267-FC014A811275
press [Enter] to exit
discovered new driver: ClientConnectionPoolInfo[name=primary, secure=false, host=localhost,
port=11111, priority=2, poolSize=1, jmxPoolSize=1, heartbeatEnabled=false, maxJobs=2147483647]
discovered new driver: ClientConnectionPoolInfo[name=backup, secure=false, host=localhost,
port=11112, priority=1, poolSize=1, jmxPoolSize=1, heartbeatEnabled=false, maxJobs=2147483647]
[client: primary-1 - ClassServer] Attempting connection to the class server at localhost:11111
[client: backup-1 - ClassServer] Attempting connection to the class server at localhost:11112
[client: backup-1 - ClassServer] Reconnected to the class server
[client: primary-1 - ClassServer] Reconnected to the class server
[client: primary-1 - TasksServer] Attempting connection to the task server at localhost:11111
[client: backup-1 - TasksServer] Attempting connection to the task server at localhost:11112
[client: primary-1 - TasksServer] Reconnected to the JPPF task server
[client: backup-1 - TasksServer] Reconnected to the JPPF task server
- shutdown the primary server:
- in the administration console's "Tree view", select the driver to which a node is connected
- click on the "Server restart or shutdown" button (
 )
- in the displayed dialog, enter "0" for the "Shutdown delay" and uncheck the "Restart" checkbox:
- click the "Ok" button to terminate the server
- you will notice that the node reconnects to the "backup" server
- now start the terminated server again
- observe how the node disconnects from the backup server and reconnects to the one that was just restarted
Related source files
I have additional questions and comments, where can I go?
There are 2 privileged places you can go to:
|
