The business logic involved takes a minuscule amount of time. This is done concurrently using several threads where each thread works on a separate First instance. Save is called because First itself, or any of its children, may be a newly created entity. Then we call firstRepository.save and the transaction is committed. There are no many-to-many relationships.Īs data flows in, we open a transaction, using Spring's fetch an entity First, update some of First's and its childrens properties, and in some occasions add new child entities. The bulk of our model consists of entity First, that has a one-to-many relationship with entity Second, that has a one-to-many relationship with entity Third. We have a Spring boot application, using Spring Data JPA and Hibernate, running in a docker container. Apart from the multithreding constraint, we believe we should be able to make the Db updates faster as a whole. For now we have set it to 10 threads as it appears that this gives us the optimal throughput. The more threads we add, above a certain threshold, the slower becomes each transaction, resulting in relatively constant processing throughput regardless of the count of threads we throw at it. We observe that, although we believe we have free capacity of connections, both in the app's and the DB's connection pools, we can't execute many concurrent transactions effectively. And we believe that our load, although not small, is not something that impressive. Here is the docker compose file: docker-compose.Hi, we seem to have hit a ceiling of the throughput we can squeeze out of our database layer. There are 3 ways to run the Zipkin server which you can find here but for the sake of simplicity in the article, we will use the docker option. Here is the diagram regarding how they interact together. Product-service receives calls from order-service.

Order-service will call REST endpoint in product-service.After that Kafka listener consumes the message then calling order-service. Customer-service which will receive request and then publish a message to Kafka topic.In this demo, there are 3 projects including: Zipkin will be used as a tool to collect and visualize distributed tracing across services. So in this tutorial, you will see how to use Spring Cloud Sleuth to record distributed tracing between Spring Boot microservices and Kafka. As we are all living in the microservices world now, it is necessary and important to have a nice and easy way to trace the invocation from one service to another directly or via a message broker like Kafka.

Spring kafka source code