https://github.com/raomuyang/plum-mesh-agent
A convenient and efficient service discovery and network communication middleware
https://github.com/raomuyang/plum-mesh-agent
etcd load-balancer loadbalance middleware netty ribbon service-discovery service-mesh zk zookeeper
Last synced: 2 months ago
JSON representation
A convenient and efficient service discovery and network communication middleware
- Host: GitHub
- URL: https://github.com/raomuyang/plum-mesh-agent
- Owner: raomuyang
- License: apache-2.0
- Created: 2018-06-22T07:19:06.000Z (almost 7 years ago)
- Default Branch: master
- Last Pushed: 2018-09-20T09:08:50.000Z (over 6 years ago)
- Last Synced: 2025-04-13T04:07:07.957Z (2 months ago)
- Topics: etcd, load-balancer, loadbalance, middleware, netty, ribbon, service-discovery, service-mesh, zk, zookeeper
- Language: Java
- Homepage:
- Size: 154 KB
- Stars: 8
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
## plum-mesh-agent
[](https://travis-ci.org/suniper/plum-mesh-agent) [](https://coveralls.io/github/suniper/plum-mesh-agent?branch=master)
pma是便捷高效的服务发现和网络通信中间件，封装了ribbon、netty、http等工具，可以快速地实现服务注册和负载均衡。
### 模块
* plum-mesh-discovery-core Service discovery核心模块
* plum-mesh-discovery-etcd 支持基于etcd的服务发现
* plum-mesh-discovery-zk 支持基于Zookeeper的服务发现
* plum-mesh-transport netty、okhttp的
* plum-mesh-discovery-start
### 快速开始
> 当前版本 1.0.2
以Zookeeper为例，使用 `plum-mesh-discovery-start` 快速开始一个pma程序，`plum-mesh-discovery-start`可以通过classpath中添加的依赖决定自己的行为：
```xml
cn.suniper
plum-mesh-discovery-start
${version}
cn.suniper
plum-mesh-discovery-zk
${version}
org.apache.zookeeper
zookeeper
${zk.version}
```
pma不会指定zookeeper/etcd库的版本，所以需要添加相应的依赖，除此之外，用户对程序的初始化过程都是无感知的。
* 简单的服务注册（假设这个服务可以在8080端口上提供一个服务）
```java
@AsProvider
public class SimpleDiscoveryEnabledServerDemo {
public static void main(String[] args) {
Object primary = new SimpleDiscoveryEnabledServerDemo();
// 设置服务名、开放的端口，所属的server group（用于服务发现）
ProviderInfo providerInfo = new ProviderInfo("test-provider-1", 8080);
Application application = new Application(null, providerInfo, "demo");
// 初始化ConfigManager实例
ConfigManager configManager = ConfigManager.newBuilder().withAppInfo(application).build();
try {
PlumApplication.launch(primary, configManager);
} catch (Exception e) {
e.printStackTrace();
}
// do sth start the server
}
@KvStoreBean
public ZooKeeper getZkClient() throws IOException, InterruptedException {
return ConnAutoInitializer.zkConnection(
"192.168.1.111:2181,192.168.1.112:2181", 1000);
}
}
```
* 简单的服务发现/负载均衡
```java
Object primary = new SimpleClientSideLoadBalance();
AppParameters parameters = AppParameters.newBuilder().enableOkHttp().build();
// 指定初始化使用动态列表的okhttp客户端
Application application = new Application(null, null, "demo");
ConfigManager configManager = ConfigManager.newBuilder().withAppInfo(application).build();
try {
PlumContext context = PlumApplication.launch(primary, parameters, configManager);
// 获取负载均衡客户端
LoadBalancingHttpClient client = (LoadBalancingHttpClient) context.getClient();
// 创建并发送请求
HttpRequest request = HttpRequest
.newBuilder()
.verb(HttpRequest.Verb.GET)
.uri("https://LB-APP/test:8080")
.build();
HttpResponse response = client.executeWithLoadBalancer(request, null);
System.out.println(response.getStatus());
// 获取可用服务列表
RegisteredServerDynamicList list = context.getDynamicServerList();
System.out.println(list.getCachedListOfServers());
} catch (Exception e) {
e.printStackTrace();
}
}
@KvStoreBean
public ZooKeeper getZkClient() throws IOException, InterruptedException {
return ConnAutoInitializer.zkConnection(
"192.168.1.111:2181,192.168.1.112:2181", 1000);
}
```
### 注册中心
理论上所有的KV存储都可以作为注册中心，只需实现KVStore接口中的所有方法，pma目前提供了对以下服务的支持：
* Zookeeper
* etcd
> 将来可以陆续增加更多的注册中心的支持，比如 redis、mongo等
### Ribbon自动配置
pma 集成了Ribbon的功能，如果需要对ribbon的参数做一些调整，只需修改配置文件（client name 为 `plum`），例如：
```properties
# Max number of retries on the same server (excluding the first try)
plum.ribbon.MaxAutoRetries=3
# Max number of next servers to retry (excluding the first server)
plum.ribbon.MaxAutoRetriesNextServer=1
# Whether all operations can be retried for this client
plum.ribbon.OkToRetryOnAllOperations=true
# Connect timeout used by Apache HttpClient
plum.ribbon.ConnectTimeout=3000
# Read timeout used by Apache HttpClient
plum.ribbon.ReadTimeout=3000
```
若pma的初始化参数没有显式地注明初始化netty或okhttp客户端，则会通过ribbon配置初始化负载均衡客户端，具体可见ribbon的[文档](https://github.com/Netflix/ribbon/wiki/Programmers-Guide)
### Netty自动配置，池化连接
作为高性能的网络通信框架，netty在不少项目中都有引入，这里同样提供了对netty的支持。Netty客户端同样可以通过配置初始化，配置示例如下：
```properties
# EventLoopGroup的类型，默认为NioEventLoopGroup
plum.tcp.groupEventType=io.netty.channel.nio.NioEventLoopGroup
# socket通道类型默认为NioSocketChannel
plum.tcp.socketChannelType=io.netty.channel.socket.nio.NioSocketChannel
# group工作线程数，默认为4
plum.tcp.workers=4
# 连接池最大的连接数，默认为20
plum.tcp.maxPoolConn=30
# channel pipeline从头到尾将要添加的handler
plum.tcp.channelPipelines=io.netty.handler.codec.http.HttpClientCodec, io.netty.handler.codec.http.HttpContentDecompressor, SimpleHttpHandler
```
只需在`AppParameters`中将`autoTcpClient`设为`true`
（命令行中`--plum.auto.tcp=true`），pma将自动初始化基于netty的负载均衡客户端
### OkHttp客户端
只需在`AppParameters`中将`okHttpClient`设为`true` （命令行中`--plum.auto.http=true`），pma的客户端即可初始化为基于OkHttp的负载均衡客户端
### 使用负载均衡
`plum-mesh-transport`模块提供的tcp和http负载均衡可以单独使用，以下是一个简单的负载均衡示例：
```java
ILoadBalancer loadBalancer = LoadBalancerBuilder
.newBuilder()
.withRule(new RandomRule())
.buildFixedServerListLoadBalancer(Arrays.asList(
new Server("atomicer.cn", 80),
new Server("baidu.com", 443),
new Server("zhihu.com", 443)
));
LoadBalancingHttpClient httpClient = new LoadBalancingHttpClient(loadBalancer);
HttpRequest request = new HttpRequest.Builder()
.verb(HttpRequest.Verb.GET)
.uri("https://LB-APP")
.build();
for (int i = 0; i < 10; i++) {
// 通过打印的结果我们可以看到列表中的三个host被随机请求
HttpResponse response = httpClient.executeWithLoadBalancer(request);
System.out.println(String.format("\n--------- %s ----------", response.getRequestedURI()));
System.out.println(response.getStatus());
System.out.println(response.getPayload());
}
```