PoolBoy is an RPC connection pool with support for multiple Corda nodes.

See complete documentation at https://manosbatsis.github.io/corda-rpc-poolboy

Note this is an early release. It is highly unstable but provided in hope it can be useful. Contributions are welcome.

Add PoolBoy to your Cordapp's Gradle dependencies:

dependencies{
    // PoolBoy dependency
    compile("com.github.manosbatsis.corda.rpc.poolboy:corda-rpc-poolboy:$poolboy_version")

    // Corda dependencies etc.
    // ...

}    

To use PoolBoy you will have to implement a RpcConfigurationService. The service provides RPC pool and connection configuration. Implementations can be "fixed" or fully dynamic, e.g. based on a properties file or database connection etc. respectively.

A future release will provide a sample, properties-based implementation. In the meantime you can find an example in Corbeans' default
RpcConfigurationService.

The main configuration items provided by such a service are as follows:

1. getRpcPoolParams(): A PoolParams instance, in each turn specifies:
   • rpcClientsMode: An RpcClientsMode value to control the CordaRPCClient creation strategy, one of SHARED, POOLED, DEDICATED. Default is SHARED.
   • rpcClientPool: The commons-pool2 keyed configuration to use if rpcClientsMode is set to POOLED, equivalent to a GenericKeyedObjectPoolConfig.
   • rpcOpsPool: The commons-pool2 keyed configuration to use for NodeRpcConnection instances - those start and wrap a CordaRPCOps
2. getRpcNodeParams(String): Provides the RPC connection configuration (i.e. NodeParams) corresponding to the given node key per your configuration implementation, e.g. X500 name, application.properties key, database primary key etc. Note that the implementor is solely responsible for any caching.
3. getGracefulReconnect(NodeParams): Override to change the GracefulReconnect implementation to use when NodeParams.disableGracefulReconnect is false.
4. getCustomSerializers(List<String>): Custom serializer types found in the configured cordapp packages. Override to bypass classpath scanning and improve search discovery performance.
5. buildPoolKey(String): Builds a PoolKey for the given node name. Override to customise PoolKey.externalTrace and PoolKey.impersonatedActor, i.e. the equivalent parameters passed to CordaRPCClient.start()

Using a full PoolBoy:

// Create the PoolBoy instance
val poolBoy = PoolBoy(myRpcConfigurationService)

// Obtain a pool key for the target node
val poolKey = myRpcConfigurationService.buildPoolKey(nodeName)

// Do something with the CordaRPCOps
// for the given node name 
poolBoy.withConnection(poolKey){
    // e.g. start a flow
   it.proxy.startFlow(MyFlow::class.java, foo, bar, baz)
}

Using a PoolBoyConnection, i.e. a pooled connection handle to a target node directly:

// Create the PoolBoy instance
val poolBoy = PoolBoy(myRpcConfigurationService)

// Obtain a pool key for the target node
val poolKey = myRpcConfigurationService.buildPoolKey(nodeName)

// Get a pool-able connection handle
val pbCconn: PoolBoyConnection = poolBoy.forKey(poolKey)

// Do something with the CordaRPCOps
// for the target node
pbCconn.withConnection {
    // e.g. start a flow
   it.proxy.startFlow(MyFlow::class.java, foo, bar, baz)
}

Both PoolBoy and PoolBoyConnection have borrowConnection() and returnConnection() methods, but it's highly discouraged; you'd better know what you're doing if you use them. The withConnection approach is preferred.