接着上图分析,经过Consensus Commit流程生成批数据后,是如何送入到ChainCode呢?我们还是以Invoke命令分析。1)在consensus的helper中调用chaincode的ExecuteTransactions 进入transaction处理流程
func (h *Helper) ExecTxs(id interface{}, txs []*pb.Transaction) ([]byte, error) { succeededTxs, res, ccevents, txerrs, err := chaincode.ExecuteTransactions(context.Background(), chaincode.DefaultChain, txs) }
2)该函数在core/chaincode 中处理,将命令封装成ChainCode识别的格式。其中的chain对象则是访问ChainCode对应的ChainCodeSupport,这样就说明访问ChainCode的接口类是ChainCodeSupportServer。
func Execute(ctxt context.Context, chain *ChaincodeSupport, t *pb.Transaction) ([]byte, *pb.ChaincodeEvent, error) { ccMsg, err = createTransactionMessage(t.Txid, cMsg) resp, err := chain.Execute(ctxt, chaincode, ccMsg, timeout, t) }
3)该函数在ChainCodeSupport文件中,首先检测ChainCode是否建立成功、能否正常运行。其中chrte.handler的得来是比较复杂的,见下描述
func (chaincodeSupport *ChaincodeSupport) Execute(ctxt context.Context, chaincode string, msg *pb.ChaincodeMessage, timeout time.Duration, tx *pb.Transaction) (*pb.ChaincodeMessage, error) { chrte, ok := chaincodeSupport.chaincodeHasBeenLaunched(chaincode) chrte.handler.sendExecuteMessage(msg, tx) }
3.1)在创建ChainCodeSupport的时候registerChaincodeSupport 调用 NewChaincodeSupport 实例化ChainCodeSupport(start.go),服务器的Name:
protos.ChaincodeSupport
ccStartupTimeout := time.Duration(tOut) * time.Millisecond ccSrv := chaincode.NewChaincodeSupport(chainname, peer.GetPeerEndpoint, userRunsCC, ccStartupTimeout, secHelper) //Now that chaincode is initialized, register all system chaincodes. system_chaincode.RegisterSysCCs() pb.RegisterChaincodeSupportServer(grpcServer, ccSrv)
var _ChaincodeSupport_serviceDesc = grpc.ServiceDesc{ ServiceName: "protos.ChaincodeSupport", HandlerType: (*ChaincodeSupportServer)(nil), Methods: []grpc.MethodDesc{}, Streams: []grpc.StreamDesc{ { StreamName: "Register", Handler: _ChaincodeSupport_Register_Handler, ServerStreams: true, ClientStreams: true, }, }, }
3.2)ChainCode 调用 err := shim.Start(new(SimpleChaincode)) 接入到ChainCodeSupportServer
err := shim.Start(new(SimpleChaincode))
3.3)连接ChainCodeSupprotServer同时调用Register函数
func Start(cc Chaincode) error { chaincodeSupportClient := pb.NewChaincodeSupportClient(clientConn) stream, err := chaincodeSupportClient.Register(context.Background()) err = chatWithPeer(chaincodename, stream, cc) }
3.4)与此同时ChainCodeSupportServer会根据Client调用注册函数创建该Stream的Handler处理句柄,创建消息响应循环,等待Client发送命令.(注意该handler就是我们关心的
handler.sendExecuteMessage)func HandleChaincodeStream(chaincodeSupport *ChaincodeSupport, ctxt context.Context, stream ccintf.ChaincodeStream) error {
deadline, ok := ctxt.Deadline()
chaincodeLogger.Debugf("Current context deadline = %s, ok = %v", deadline, ok)
handler := newChaincodeSupportHandler(chaincodeSupport, stream)
return handler.processStream()
}
3.5) 客户端shim/Chaincode发送RegisterMessage
handler.serialSend(&pb.ChaincodeMessage{Type: pb.ChaincodeMessage_REGISTER, Payload: payload})
3.6)对于Server而言,我们刚刚创建了handler又有ProcessStream消息响应循环,这样RegisterMessage就交到了ProcessStream手里,ProcessStream根据消息类型执行命令分发
调用beforeRegisterEvent函数。
func (handler *Handler) beforeRegisterEvent(e *fsm.Event, state string) { err = handler.chaincodeSupport.registerHandler(handler) }
3.7)同理在client端(shim/chaincode)也建立响应的消息响应循环。
4)到目前为止还没有完,我们将invoke命令送给了Client的委托模块Shim进行处理。Shim模块根据来访事件类型送入指定处理函数
5)调用ChainCode的invoke函数func (handler *Handler) enterTransactionState(e *fsm.Event) { msg, ok := e.Args[0].(*pb.ChaincodeMessage) if !ok { e.Cancel(fmt.Errorf("Received unexpected message type")) return } chaincodeLogger.Debugf("[%s]Received %s, invoking transaction on chaincode(Src:%s, Dst:%s)", shorttxid(msg.Txid), msg.Type.String(), e.Src, e.Dst) if msg.Type.String() == pb.ChaincodeMessage_TRANSACTION.String() { // Call the chaincode's Run function to invoke transaction handler.handleTransaction(msg) } }
func (handler *Handler) handleTransaction(msg *pb.ChaincodeMessage) { res, err := handler.cc.Invoke(stub, function, params) }
以上分析涉及两个过程 1) consensus结束后如何将命令送入ChainCode 2)ChainCodeSupportServer与ChainCode如何建立通信关系。
画一个 ChainCodeSupportServer与ChainCode如何建立通信关系 图: