Using rabbitmq to implement distributed transaction

 

RabbitMQ The idea of solving distributed transaction ：

Case study ： Classic case , In the case of the current popular takeout , After the user orders , Call the order service , Let the post order service call the dispatch system to inform the delivery personnel to deliver the order , At this time, the order system and dispatch system adopt MQ Asynchronous communication .

RabbitMQ Solve the principle of distributed transaction ： Using the principle of final consistency .

 

The following three elements need to be guaranteed

1、 Confirm that the producer must deliver the data to MQ Server （ use MQ Message confirmation mechanism ）

2、MQ Consumer news can consume information correctly , Use manual ACK Pattern （ Note the idempotency problem of retrial ）

3、 How to ensure that the first transaction is executed first , Using compensation mechanism , Create a replenishment consumer to monitor , If the order is not created successfully , Make up the order .

 

If the producer delivers a message to MQ Server success

scene 1： If the consumer fails to consume the news , The producer does not need to roll back the transaction .

Solution ： Consumers use manual ack Response mode , use MQ Compensation and retrial mechanism , Be careful MQ The problem of compensating idempotency .

 

problem ： How to ensure that producers deliver messages to MQ The server will succeed ？

Solution ：confirm Mechanism ( Confirmation response mechanism ).

scene 2 If the producer delivers a message to MQ Server failed , How to solve ？

  Solution ： Using the producer retry mechanism to send messages , Pay attention to idempotency .

scene 3   How to ensure that a transaction is executed first , The producer delivers a message to MQ Server success , Consumer spending is successful , But the order rolled back . 

Solution ： Replenishment mechanism .

 

Traditional solutions ：

RabbitMq Solution ：