Distributed application refers to the way in which applications are distributed on different computers and work together to complete a task over the network. Take javaEE to implement an e-commerce website as an example:

• Single application: all functions are written in one project; Package as a runnable WAR package; The deployment of this WAR package completes the functionality of the entire site.
• Distributed applications: different functions are written in different projects; Packaged into multiple runnable WAR packages; Multiple running services work together to complete the functionality of the entire site.
  
  

As shown in the figure above, this e-commerce website contains four modules (also known as services) : user management, commodity management, order management and payment management. In distributed applications, many functions are completed by the cooperation of multiple services. How to efficiently and orderly cooperate among services has become an important issue in distributed development.

• Analogy:

  • Single application –> All work done by one person
  • Distributed applications –> Many tasks are performed by multiple people working together

The architecture of ZooKeeper is shown as follows:

A ZooKeeper cluster consists of a group of Server nodes. In this group, one node serves as Leader and the other nodes serve as followers. The management and ZooKeeper architecture have the following key points:

• Read operation that reads one of the follwers directly and returns directly,

• For write operations, the ZooKeeper cluster performs the following operations:

  • These requests are sent to the Leader node
  • Data changes on the Leader node are synchronized to other Follower nodes in the cluster

• After receiving a data change request, the Leader node first writes the change to the local disk for recovery. Changes are not applied to memory until all write requests are persisted to disk.

• Note: Data persisted to the hard disk is only used for data recovery when the service restarts.

• If the Leader node fails, the cluster automatically elects a new Leader node.

• Permanent node (default) : The permanent node does not disappear once it is successfully created.
• Temporary node: After a temporary node is created, it is automatically deleted if the client and server are disconnected

The ZooKeeper client can monitor zNode data changes. When zNode data changes, the ZooKeeper client automatically notifies the listening client.

A distributed lock corresponds to a node of ZooKeeper. Each client thread that needs to obtain the distributed lock creates a temporary ordered node under this node. In this case, there are two situations:

1. If the temporary sequential node created is the first node under the folder, the distributed lock is considered successful.
2. If the temporary sequential node created is not the first node in the folder, the current lock is considered to have been acquired by another client thread, and the node needs to enter the blocking stateWakes up the current thread when the previous node in the sequence releases the lock.
   
   

In the production environment, you need to call the interface (such as HTTP interface) of another business server from one business server (such as Tomcat), so you need to know the IP address and port of the called party. There are three solutions:

• IP& port write dead in project (not recommended)

  • Pros: Simple and straightforward.

  • Disadvantages: If the IP address and port of the caller changes or the number of deployed nodes changes, the caller code needs to be modified to bring the caller back online

  • analogy

    • App1 – > A classmate
    • App2 – > classmate B
    • IP+ Port – > Phone number
    • Call — > Student A dials student B’s phone number

• Reverse proxy server, such as Nginx

  • Advantage: The caller only needs to configure the Nginx address, and does not care about the real IP address and port of the caller.
    
    

  • Analogy:

    • Nginx – > monitor
    • Reverse proxy server forwarding – > Student A asks the monitor to send A message to student B

• Registries, such as ZooKeeper

  • Advantages: The caller does not need to configure the IP address of the called party, and the information of the called party can be automatically updated when it changes.

  • Process:

    • Register, app2 starts to send IP, port and other information to the registry
    • When APP1 starts, deregister and pull app2 information (IP, port, etc.)
    • Initiating a call (such as an HTTP request)

  • The distributed framework Dubbo recommends using ZookeeEPr as the registry.
    
    

  • Analogy:

    • Registration center –> Class teacher
    • Registration — all students will give their mobile phone number to the class teacher when they enter the school
    • Pull the target server information and initiate the call — > Student A asks the head teacher for student B’s phone number and then calls