手写一个SpringBoot Starter自动配置动态的创建RestHighLevelClient

1、背景

最近做了个elasticsearch搜索的功能，因为会有多个项目模块用到所以决定做成一个springboot starter以便于其他地方用到就不需要作过多的配置，加之不久前出现一个项目的es服务出问题导致几个共用一个es集群的项目都受到影响，需要不同项目之间用的es隔离开来，所以决定手写自定义的Starter并动态的创建RestHighLevelClient。

2、starter的大概结构

3、开发流程

第一步创建一个springboot项目，命名为starter-elasticsearch

创建DynamicElasticsearchProperties类，通过@ConfigurationProperties注解把yml或者properties配置文件转化为bean，从而获取yml或者properites配置文件的属性

创建ElasticsearchProperties类，对应yml或者properites配置文件的属性

创建自动配置的bean，命名为ElasticsearchAutoConfiguration，通过@EnableConfigurationProperties()注解使得@ConfigurationProperties注解生效，后续自动装配动态的创建RestHighLevelClient将在这个类进行

假设需要引入es的项目自定义yml配置如下：

maoshi8:
  elasticsearch:
    datasource:
      node1:
        host: 123.123.123.123
        port: 9200
      node2:
        host: 111.111.111.111
        port: 9200

DynamicElasticsearchProperties类：

@ConfigurationProperties(prefix = "maoshi8.elasticsearch")
public class DynamicElasticsearchProperties {

    /**
     * 配置多动态数据源key datasource id
     */
    private Map<String, ElasticsearchProperties> datasource = new HashMap<String, ElasticsearchProperties>();

    public Map<String, ElasticsearchProperties> getDatasource() {
        return datasource;
    }

    public void setDatasource(Map<String, ElasticsearchProperties> datasource) {
        this.datasource = datasource;
    }

}

@Data
public class ElasticsearchProperties {

	private String schema = "http";
	private String host;
	private Integer port;
	private String username = "elastic";
	private String password = "******";
	private int connectTimeOut = 1000;
	private int socketTimeOut = 30000;
	private int connectionRequestTimeOut = 500;
	private int maxConnectNum = 100;
	private int maxConnectPerRoute = 100;
	private boolean uniqueConnectTimeConfig = true;
	private boolean uniqueConnectNumConfig = true;
	}

DynamicElasticsearchProperties类里ConfigurationProperties注解到maoshi8.elasticsearch就能截取到key=node1和key=node2的value值为ElasticsearchProperties对象的map=>datasource。其中除了host和port属性都给了默认值，我们只需要拿到动态配置的host和port就能动态的创建不同的RestHighLevelClient。这里是定义了两个node1和node2，如果定义了10个就能相应的拿到一个有10个配置对象的datasource。

第二步创建一个META-INF目录

创建META-INF目录后，在里面新建一个spring.factories文件，添加以下代码：

org.springframework.boot.autoconfigure.EnableAutoConfiguration=\
ElasticsearchAutoConfiguration

这里的ElasticsearchAutoConfiguration是我自定义自动配置的类的reference路径。spring boot项目启动以后SpringFactoriesLoader.loadFactoryNames方法调用loadSpringFactories方法从所有的jar包中读取META-INF/spring.factories文件信息，通过反射机制将spring.factories中@Configuration类实例化为对应的java实列，发现要自动配置的bean。

第三步动态的创建RestHighLevelClient

前面已经能自动找到配置类，并且拿到yml或者properties的参数属性了，接下来就只需要在自动配置类ElasticsearchAutoConfiguration里面循环datasource里的定义配置，动态的创建RestHighLevelClient就可以了。
代码如下：
自动配置类ElasticsearchAutoConfiguration：

@Configuration
@EnableConfigurationProperties(DynamicElasticsearchProperties.class)
public class ElasticsearchAutoConfiguration {

    @Autowired
    private DynamicElasticsearchProperties dynamicElasticsearchProperties;

    private final CredentialsProvider credentialsProvider = new BasicCredentialsProvider();

    @Bean
    public RestHighLevelClientUtils restHighLevelClientUtils(){
        RestHighLevelClientUtils restHighLevelClientUtils = new RestHighLevelClientUtils();
        Map<String, ElasticsearchProperties> datasourceMap = dynamicElasticsearchProperties.getDatasource();
        Map<String, RestHighLevelClient> restHighLevelClientMap = new HashMap<>();
        for (String key : datasourceMap.keySet()) {
            ElasticsearchProperties elasticsearchProperties = datasourceMap.get(key);
            HttpHost httpHost = new HttpHost(elasticsearchProperties.getHost(),
                    elasticsearchProperties.getPort(),elasticsearchProperties.getSchema());
            RestClientBuilder builder = RestClient.builder(httpHost);
            if (elasticsearchProperties.isUniqueConnectTimeConfig()) {
                setConnectTimeOutConfig(builder ,elasticsearchProperties);
            }
            if (elasticsearchProperties.isUniqueConnectNumConfig()) {
                setMutiConnectConfig(builder ,elasticsearchProperties);
            }
            RestHighLevelClient client = new RestHighLevelClient(builder);
            restHighLevelClientMap.put(key,client);

            RestHighLevelClientUtils.registerRestHighLevelClient(key, client);
        }
        return restHighLevelClientUtils;
    }

    /** * 异步httpclient的连接延时配置 */
    public void setConnectTimeOutConfig(RestClientBuilder builder, ElasticsearchProperties elasticsearchProperties) {
        builder.setRequestConfigCallback(new RestClientBuilder.RequestConfigCallback() {
            @Override
            public RequestConfig.Builder customizeRequestConfig(RequestConfig.Builder requestConfigBuilder) {
                requestConfigBuilder.setConnectTimeout(elasticsearchProperties.getConnectTimeOut());
                requestConfigBuilder.setSocketTimeout(elasticsearchProperties.getSocketTimeOut());
                requestConfigBuilder.setConnectionRequestTimeout(elasticsearchProperties.getConnectionRequestTimeOut());
                return requestConfigBuilder;
            }
        });
    }

    /** * 异步httpclient的连接数配置 */
    public void setMutiConnectConfig(RestClientBuilder builder, ElasticsearchProperties elasticsearchProperties) {
        builder.setHttpClientConfigCallback(new RestClientBuilder.HttpClientConfigCallback() {
            @Override
            public HttpAsyncClientBuilder customizeHttpClient(HttpAsyncClientBuilder httpClientBuilder) {
                httpClientBuilder.setMaxConnTotal(elasticsearchProperties.getMaxConnectNum());
                httpClientBuilder.setMaxConnPerRoute(elasticsearchProperties.getMaxConnectPerRoute());
                //elasticsearch7.x版本需要密码
                credentialsProvider.setCredentials(AuthScope.ANY, new UsernamePasswordCredentials(elasticsearchProperties.getUsername(), elasticsearchProperties.getPassword()));
                httpClientBuilder.setDefaultCredentialsProvider(credentialsProvider);
                return httpClientBuilder;
            }
        });
    }


}

工具类RestHighLevelClientUtils ：

public class RestHighLevelClientUtils {

    private static Map<String, RestHighLevelClient> restHighLevelClientMap = new ConcurrentHashMap<>();

    public RestHighLevelClientUtils() {
    }

    public static RestHighLevelClient getRestHighLevelClient(String restHighLevelClientId) {
        RestHighLevelClient restHighLevelClient = restHighLevelClientMap.get(restHighLevelClientId);
        if(restHighLevelClient == null){
            throw new RuntimeException("找不到RestHighLevelClient,restHighLevelClientId:"+ restHighLevelClientId);

        }
        return restHighLevelClient;
    }

    public RestHighLevelClient getClient(String restHighLevelClientId) {
        return RestHighLevelClientUtils.getRestHighLevelClient(restHighLevelClientId);
    }

    public static void registerRestHighLevelClient(String restHighLevelClientId, RestHighLevelClient restHighLevelClient) {
        RestHighLevelClientUtils.restHighLevelClientMap.put(restHighLevelClientId, restHighLevelClient);
    }
}

这里我是用的一个工具类RestHighLevelClientUtils定义一个ConcurrentHashMap来存放动态创建的RestHighLevelClient，map的key就是datasource的key：node1、node2，value就是创建的RestHighLevelClient。

第四步引入starter使用

引入pom文件：

        <!--        添加自定义的starter-elasticsearch    -->
        <dependency>
            <groupId>com.maoshi8</groupId>
            <artifactId>starter-elasticsearch</artifactId>
            <version>0.0.1-SNAPSHOT</version>
        </dependency>

需要RestHighLevelClient使用处添加代码：

    private static RestHighLevelClient client;

    @Resource
    private RestHighLevelClientUtils restHighLevelClientUtils;

    @PostConstruct
    public void init(){
        client = restHighLevelClientUtils.getClient("node1");
    }

需要引入的模块只需要加上自定义的yml配置，并在pom文件引入该starter，再restHighLevelClientUtils工具类就能拿到对应的RestHighLevelClient了