当前位置:网站首页>Nacos conformance protocol cp/ap/jraft/distro protocol

Nacos conformance protocol cp/ap/jraft/distro protocol

2022-07-07 15:22:00 Grab

Nacos Because we need to support the registration of multiple service types , And it can have the disaster recovery function of the computer room 、 Essential capabilities such as cluster expansion , Run simultaneously in a single cluster CP The protocol and AP Agreement two kinds of agreements . In the decision to use CP still AP⼀ Sexual time , Use ⼀ A broker , Forwarding through controllable rules .
current ⼀ Conformance protocol implementation ,⼀ It's based on Raft Of CP⼀ Sexual nature ,⼀ One is based on self-developed agreement Distro Of AP Uniformity .


Raft agreement

Nacosv1.x It's using Raft agreement ,v2.x Change it to JRaft agreement .Raft The agreement needs no more words , yes CP agreement , be based on Leader To write .


Distro agreement

Distro Agreement is Nacos Community self-developed ⼀ Kind of AP Distributed protocol , It is designed for temporary instances ⼀ Distributed protocol , It ensures that in some Nacos After node downtime , The whole temporary instance processing system can still work normally . Its data is stored in the cache , And full data synchronization will be carried out at startup , And conduct data verification regularly . Every Distro All nodes can receive read and write requests .

Distro The main design idea of the protocol is as follows :

  1. Nacos Each node is equal and can handle write requests , Synchronize new data to other nodes at the same time .
  2. Each node is only responsible for part of the data , Regularly send the verification value of your own data to other nodes to keep the data ⼀ Sexual nature .
  3. Each node handles read requests independently , Respond locally in a timely manner .

Distro How the protocol works :

  1. Data initialization , Newly added Distro The node will pull the full amount of data . The specific operation is to poll all Distro node , Pull the full amount of data by sending a request to other machines . After the full pull operation is completed ,Nacos All currently registered non persistent instance data are maintained on each machine .
  2. data verification , stay Distro After cluster startup , The heartbeat will be sent between the machines regularly for data verification . If the data of one machine is inconsistent with that of other machines , Then it will pull all the data to request to complete .
  3. Write operations , When a write request to register a non persistent instance hits a certain computer Nacos Server time , First by Filter Intercept , On request IP The port information is forwarded to the corresponding Distro Processing requests on the responsibility node .Distro The agreement will also be implemented regularly Sync Mission , Synchronize all the instance information of the local machine to other nodes .
  4. Read operations , Due to the full amount of data stored on each machine , So in every ⼀ In the first read operation ,Distro The machine will pull data directly from the local , Respond quickly .


Service registration found using AP agreement

To ensure availability , Try to ensure the service registration and discovery ability as much as possible, and can provide external services ,Nacos The service registration discovery of adopts the mechanism that heartbeat can automatically complete the service data compensation . If the data is lost , This mechanism can quickly compensate for data loss , And ensure final consistency .


Configuration management adopts CP agreement

The configuration data , Is directly in Nacos The server side creates and manages , You must ensure that most nodes have saved this configuration data before you can think that the configuration has been saved successfully . In order to avoid serious failure caused by missing configuration changes , Strong ⼀ Consensus algorithm , Ensure that most nodes in the cluster are strong ⼀ To .


Configuration center consistency protocol

Server Between the ⼀ Agreement on sex , Yes DB Pattern ( Read write separation architecture ),⼀ The core of consistence is Server And DB Keep the data ⼀ Sexual nature , To ensure that Server data ⼀ Cause .Server They are all equal , Write any data ⼀ individual Server, Priority persistence , After successful persistence, asynchronously notify other nodes to pull the latest configuration value from the database , And notify that the writing is successful .
Server Between the ⼀ Agreement on sex , nothing DB Pattern Server Inter use Raft Agreement guarantee data ⼀ Sexual nature .
SDK And Server Of ⼀ Agreement on sex , adopt MD5 Whether the value is ⼀ To realize , If you don't ⼀ To pull the latest value .

原网站

版权声明
本文为[Grab]所创,转载请带上原文链接,感谢
https://yzsam.com/2022/188/202207071316592916.html