ZooKeeper allows for very simple and effective leader election out of the box. This component exploits this election capability in a RoutePolicy to control when and how routes are enabled. This policy would typically be used in fail-over scenarios, to control identical instances of a route across a cluster of Camel based servers. A very common scenario is a simple 'Master-Slave' setup where there are multiple instances of a route distributed across a cluster but only one of them, that of the master, should be running at a time. If the master fails, a new master should be elected from the available slaves and the route in this new master should be started.
The policy uses a common znode path across all instances of the RoutePolicy that will be involved in the election. Each policy writes its id into this node and zookeeper will order the writes in the order it received them. The policy then reads the listing of the node to see what postion of its id; this postion is used to determine if the route should be started or not. The policy is configured at startup with the number of route instances that should be started across the cluster and if its position in the list is less than this value then its route will be started. For a Master-slave scenario, the route is configured with 1 route instance and only the first entry in the listing will start its route. All policies watch for updates to the listing and if the listing changes they recalculate if their route should be started. The following example uses the node '/someapplication/somepolicy' for the election and is set up to start only the top '1' entries in the node listing i.e. elect a master:
ZooKeeperRoutePolicy policy = new ZooKeeperRoutePolicy( "zookeeper:localhost:39913/someapp/somepolicy", 1); from("direct:policy-controlled").routePolicy(policy) .to("mock:controlled");