Trident tutorial

tridenter Characteristics of ：

1. Adopt the idea of decentralization to manage clusters
2. Support message multicast and unicast between clusters
3. Support various load balancing strategies
4. Support for multiple Leader Election algorithm
5. Provide process pool / Implementation of scheduling process pool
6. Built in microservice registry
7. Built in flow limiting and degradation strategy for calls between multiple microservices
8. Built in microservices Rest client
9. built-in HTTP The service gateway
10. Cluster status monitoring and alarm
11. Distributed transaction management

Message multicast in a cluster is tridenter Very important function ,tridenter The bottom is through Redis（PubSub） Realize the multicast function, so as to realize the mutual discovery of applications , Then form the application cluster . Each member of the cluster supports the ability of message multicast and unicast .

utilize tridenter The ability to support message unicast ,tridenter Provides Leader Election algorithm interface , There are two built-in Leader Election algorithm , Fast Leader Election algorithm （ be based on Redis queue ） And consistent election algorithm ( be based on Paxos Algorithm )

meanwhile , utilize tridenter The ability to support message unicast ,tridenter Provides a process pool , It realizes the ability of cross process method call and method fragmentation

On the other hand ,tridenter It also provides the basic functions of microservice governance ：

tridenter Bring your own registry , Using the principle of message multicast , Application is mutual discovery , Mutually registered , Therefore, each member in the cluster has a full list of members , That is, every application is a registry , It embodies the design idea of decentralization . Each member through a naming service , Between applications HTTP The ability of interfaces to call each other , It also provides various notes and Restful Configuration decouples service publishers and consumers

tridenter With gateway function , Applications can be published as gateway services independently , Can distribute HTTP Request and download tasks （ Uploading is not supported at the moment ）

tridenter A variety of load balancing algorithms and current limiting degradation strategies are also built in , Users can also customize load balancing algorithms or degradation policies

tridenter Realized spring actuator Health check interface , In addition to monitoring cluster status , It also has functions such as statistical analysis of its own interface , The unified management and monitoring of the interface are preliminarily realized

therefore , be based on tridenter frame , We can also build a set similar to Spring Cloud Our microservice system

tridenter The design idea of decentralization is adopted , That is, developers do not need to know which is the current master node , Which nodes are slave nodes , More importantly, you should not explicitly define an application as the master node , This is from tridenter Adopted Leader The election algorithm determines , The default election algorithm is fast election algorithm . According to the election algorithm , Any application node in the cluster may become the master node , By default, the first application launched is the master node , But if the consistent election algorithm is used , It may be different . According to the author's description , The consistent election algorithm is not stable at present , It is recommended to use fast election algorithm in application .

Maven:

        <dependency>

            <groupId>com.github.paganini2008.atlantis</groupId>

            <artifactId>tridenter-spring-boot-starter</artifactId>

            <version>1.0-RC3</version>

        </dependency>

tridenter The basic function of is to make different Spring Boot The application becomes a cluster mode , So when configuring , We will be having application.properties

Define two system variables , Otherwise, an error will be reported

spring.application.name=chaconne-management  #  The service name 

spring.application.cluster.name=chaconne-management-cluster  # Cluster name 

After starting , If in Console The following information indicates that the cluster configuration is effective

2021-06-05 18:20:11 [INFO ] io.undertow - starting server: Undertow - 2.0.29.Final

2021-06-05 18:20:11 [INFO ] org.xnio - XNIO version 3.3.8.Final

2021-06-05 18:20:11 [INFO ] org.xnio.nio - XNIO NIO Implementation Version 3.3.8.Final

2021-06-05 18:20:11 [INFO ] i.a.f.t.m.ApplicationMulticastGroup - Registered candidate: {applicationContextPath: http://192.168.159.1:6543, applicationName: chaconne-management, clusterName: chaconne-management-cluster, id: fafdc9ada3a5d1de3836b1a0ba4ef174, leader: false, startTime: 1622888405582, weight: 1}, Proportion: 1/1

2021-06-05 18:20:11 [INFO ] i.a.f.t.m.ApplicationRegistryCenter - Register application: [{applicationContextPath: http://192.168.159.1:6543, applicationName: chaconne-management, clusterName: chaconne-management-cluster, id: fafdc9ada3a5d1de3836b1a0ba4ef174, leader: false, startTime: 1622888405582, weight: 1}] to ApplicationRegistryCenter

2021-06-05 18:20:11 [INFO ] i.a.f.c.SerialDependencyListener - SerialDependencyHandler initialize successfully.

2021-06-05 18:20:11 [INFO ] i.a.f.t.e.ApplicationLeaderElection - This is the leader of application cluster 'chaconne-management-cluster'. Current application event type is 'indi.atlantis.framework.tridenter.election.ApplicationClusterLeaderEvent'

2021-06-05 18:20:11 [INFO ] i.a.f.t.e.ApplicationLeaderElection - Current leader: {applicationContextPath: http://192.168.159.1:6543, applicationName: chaconne-management, clusterName: chaconne-management-cluster, id: fafdc9ada3a5d1de3836b1a0ba4ef174, leader: true, startTime: 1622888405582, weight: 1}

2021-06-05 18:20:11 [INFO ] o.s.b.w.e.u.UndertowServletWebServer - Undertow started on port(s) 6543 (http) with context path ''

2021-06-05 18:20:12 [INFO ] i.a.f.c.m.ChaconneManagementMain - Started ChaconneManagementMain in 12.134 seconds (JVM running for 12.829)

First ：

2021-06-05 18:20:11 [INFO ] i.a.f.t.m.ApplicationMulticastGroup - Registered candidate: {applicationContextPath: http://192.168.159.1:6543, applicationName: chaconne-management, clusterName: chaconne-management-cluster, id: fafdc9ada3a5d1de3836b1a0ba4ef174, leader: false, startTime: 1622888405582, weight: 1}, Proportion: 1/1

2021-06-05 18:20:11 [INFO ] i.a.f.t.m.ApplicationRegistryCenter - Register application: [{applicationContextPath: http://192.168.159.1:6543, applicationName: chaconne-management, clusterName: chaconne-management-cluster, id: fafdc9ada3a5d1de3836b1a0ba4ef174, leader: false, startTime: 1622888405582, weight: 1}] to ApplicationRegistryCenter

These two lines of logs indicate the successful registration of the message multicast group and the application registry

2021-06-05 18:20:11 [INFO ] i.a.f.t.e.ApplicationLeaderElection - This is the leader of application cluster 'chaconne-management-cluster'. Current application event type is 'indi.atlantis.framework.tridenter.election.ApplicationClusterLeaderEvent'

2021-06-05 18:20:11 [INFO ] i.a.f.t.e.ApplicationLeaderElection - Current leader: {applicationContextPath: http://192.168.159.1:6543, applicationName: chaconne-management, clusterName: chaconne-management-cluster, id: fafdc9ada3a5d1de3836b1a0ba4ef174, leader: true, startTime: 1622888405582, weight: 1}

These two lines of logs represent the use of ApplicationLeaderElection The current application of the election algorithm is leader,（ By default, the fast election algorithm takes the first application to start as Leader,

It's kind of similar Jgroups）

tridenter-spring-boot-starter It is a basic framework , Provides a variety of distributed capabilities , Here are some abilities ：

The process of pool

Multiple applications with the same name （${spring.application.name}） It can be built into a process pool , Just as the thread pool allocates different threads to call a method , The process pool can make cross application method calls , The premise is that this method exists

Sample code ：

    @MultiProcessing(value = "calc", defaultValue = "11")

    public int calc(int a, int b) {

        if (a % 3 == 0) {

            throw new IllegalArgumentException("a ==> " + a);

        }

        log.info("[" + counter.incrementAndGet() + "]Port: " + port + ", Execute at: " + new Date());

        return a * b * 20;

    }

    @OnSuccess("calc")

    public void onSuccess(Object result, MethodInvocation invocation) {

        log.info("Result: " + result + ", Take: " + (System.currentTimeMillis() - invocation.getTimestamp()));

    }

    @OnFailure("calc")

    public void onFailure(ThrowableProxy info, MethodInvocation invocation) {

        log.info("========================================");

        log.error("{}", info);

    }

explain ：

1. annotation @MultiProcessing Modification methods calc, Indicates that this method is called by multiple processes
2. onSuccess and onFailure Both methods are called asynchronously

Method slice

Method slicing is also called method parallel processing , In fact, each parameter of a set of parameters is distributed to different applications using the process pool , Then merge the output , And need to implement the partition rule interface , See source code ：

public interface Parallelization {

    Object[] slice(Object argument); //  section 

    Object merge(Object[] results);  //  Merge 

}

Sample code ：

    @ParallelizingCall(value = "loop-test", usingParallelization = TestCallParallelization.class)

    public long total(String arg) {// 0,1,2,3,4,5,6,7,8,9

        return 0L;

    }

    public static class TestCallParallelization implements Parallelization {

        @Override

        public Long[] slice(Object argument) {

            String[] args = ((String) argument).split(",");

            Long[] longArray = new Long[args.length];

            int i = 0;

            for (String arg : args) {

                longArray[i++] = Long.parseLong(arg);

            }

            return longArray;

        }

        @Override

        public Long merge(Object[] results) {

            long total = 0;

            for (Object o : results) {

                total += ((Long) o).longValue();

            }

            return total;

        }

    }

explain ：

1. annotation @ParallelizingCall modification total Method , It means that this method needs to be segmented
2. Parameters arg, For example, you can pass 0,1,2,3,4,5,6,7,8,9, The fragmentation rule will call slice Method takes the parameter as “,” Division , Become an array , Then convert each value to long type , And then distributed to each application for execution , It's all done , Re execution merge Method to add , somewhat MapReduce The smell of
3. total Method 0, It refers to the default value returned when the parameter is empty or the method is abnormal

Micro service capabilities

Rest client

Sample code ：

@RestClient(provider = "test-service")

// @RestClient(provider = "http://192.168.159.1:5050")

public interface TestRestClient {

    @PostMapping("/metrics/sequence/{dataType}")

    Map<String, Object> sequence(@PathVariable("dataType") String dataType, @RequestBody SequenceRequest sequenceRequest);

}

explain ：

1. annotation @RestClient The modified interface indicates that this is a Http client
2. In the annotations ,provider Attribute represents the service provider , It can be an application name in the cluster （${spring.application.name}）, It can also be specific http Address
3. Support Spring annotation （GetMapping, PostMapping, PutMapping, DeleteMapping）, Besides , With notes @Api More fine-grained parameter settings can be provided

gateway

@EnableGateway

@SpringBootApplication

@ComponentScan

public class GatewayMain {

    public static void main(String[] args) {

        final int port = 9000;

        System.setProperty("server.port", String.valueOf(port));

        SpringApplication.run(GatewayMain.class, args);

    }

}

Reference notes @EnableGateway That's it ,tridenter The bottom layer of the gateway is Netty4 Realized

Configure the routing ：

@Primary

@Component

public class MyRouterCustomizer extends DefaultRouterCustomizer {

    @Override

    public void customize(RouterManager rm) {

        super.customize(rm);

        rm.route("/my/**").provider("tester5");

        rm.route("/test/baidu").url("https://www.baidu.com").resourceType(ResourceType.REDIRECT);

        rm.route("/test/stream").url("  https://www.baidu.com/img/PCtm_d9c8750bed0b3c7d089fa7d55720d6cf.png").resourceType(ResourceType.STREAM);

    }

}

explain ：

ResourceType Of 4 Types ：

DEFAULT, Forward the request

REDIRECT, Jump

STREAM, Binary stream

FILE Save the file

Current limiting the drop

Current limiting refers to limiting current at the client , Instead of the server

Current limiting will depend on 3 Indicators ：

1. Response timeout rate
2. Error rate
3. Concurrency undefined By default , When any of these three indicators exceeds 80%, Current limiting will be triggered , Call downgrade service undefined Current limiting index statistics : RequestStatisticIndicatorundefined Degraded service interface ：FallbackProviderundefined Relevant source code can be studied by yourself

Health monitoring

at present tridenter Provides 3 individual HealthIndicator Subclasses of

1. ApplicationClusterHealthIndicatorundefined Displays the overall health status of the cluster
2. TaskExecutorHealthIndicatorundefined Displays the health status of the cluster thread pool
3. RestClientHealthIndicatorundefined Show Rest The health status of the client （ Response timeout rate , Error rate , Concurrency ）