Elasticsearch7.9.2与Nebula Spark连接器冲突RestHighLevelClient实例化失败

• 当前运行环境使用Elasticsearch 7.9.2版本，集成Nebula Graph官方Spark连接器nebula-spark-connector时出现依赖冲突
• 此前了解过Spark与Elasticsearch依赖冲突的相关案例，但未找到可直接落地的解决方案
• 已尝试对nebula-spark-connector的jar包执行shade重打包处理，报错未消除

服务启动初始化ES高级REST客户端时抛出如下错误：

Failed to instantiate [org.elasticsearch.client.RestHighLevelClient]: Factory method 'restHighLevelClient' threw exception; nested exception is java.lang.NoSuchFieldError: INSTANCE
    at org.springframework.beans.factory.support.SimpleInstantiationStrategy.instantiate(SimpleInstantiationStrategy.java:185)
    at org.springframework.beans.factory.support.ConstructorResolver.instantiate(ConstructorResolver.java:653)
    ... 47 common frames omitted
Caused by: java.lang.NoSuchFieldError: INSTANCE
    at org.apache.http.impl.nio.conn.PoolingNHttpClientConnectionManager$InternalAddressResolver.<init>(PoolingNHttpClientConnectionManager.java:619)
    at org.apache.http.impl.nio.conn.PoolingNHttpClientConnectionManager.<init>(PoolingNHttpClientConnectionManager.java:165)
    at org.apache.http.impl.nio.conn.PoolingNHttpClientConnectionManager.<init>(PoolingNHttpClientConnectionManager.java:149)
    at org.apache.http.impl.nio.conn.PoolingNHttpClientConnectionManager.<init>(PoolingNHttpClientConnectionManager.java:121)
    at org.apache.http.impl.nio.client.HttpAsyncClientBuilder.build(HttpAsyncClientBuilder.java:668)
    at java.security.AccessController.doPrivileged(Native Method)
    at org.elasticsearch.client.RestClientBuilder.createHttpClient(RestClientBuilder.java:219)
    at java.security.AccessController.doPrivileged(Native Method)
    at org.elasticsearch.client.RestClientBuilder.build(RestClientBuilder.java:191)
    at org.elasticsearch.client.RestHighLevelClient.<init>(RestHighLevelClient.java:287)
    at org.elasticsearch.client.RestHighLevelClient.<init>(RestHighLevelClient.java:279)
    at com.bybit.byassets.collection.config.ElasticSearchConfig.restHighLevelClient(ElasticSearchConfig.java:54)
    at com.bybit.byassets.collection.config.ElasticSearchConfig$$EnhancerBySpringCGLIB$$e050e7f6.CGLIB$restHighLevelClient$0(<generated>)
    at com.bybit.byassets.collection.config.ElasticSearchConfig$$EnhancerBySpringCGLIB$$e050e7f6$$FastClassBySpringCGLIB$$43cd2bed.invoke(<generated>)
    at org.springframework.cglib.proxy.MethodProxy.invokeSuper(MethodProxy.java:244)
    at org.springframework.context.annotation.ConfigurationClassEnhancer$BeanMethodInterceptor.intercept(ConfigurationClassEnhancer.java:331)
    at com.bybit.byassets.collection.config.ElasticSearchConfig$$EnhancerBySpringCGLIB$$e050e7f6.restHighLevelClient(<generated>)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:498)
    at org.springframework.beans.factory.support.SimpleInstantiationStrategy.instantiate(SimpleInstantiationStrategy.java:154)
    ... 48 common frames omitted

这个NoSuchFieldError: INSTANCE是典型的依赖版本不匹配问题，冲突点在Apache HttpAsyncClient相关包：

1. Elasticsearch 7.9.2 对http依赖的版本要求固定：httpasyncclient:4.1.4、httpcore-nio:4.4.14、httpclient:4.5.13，这些版本中PoolingNHttpClientConnectionManager$InternalAddressResolver类依赖的INSTANCE静态字段存在
2. nebula-spark-connector会传递引入版本不匹配的http相关依赖，类加载时优先加载了错误版本的类，就会报字段不存在的错误
3. 之前shade重打包没生效，基本是因为配置没有覆盖全所有http相关包，要么没做类路径重定位，要么打包后错误版本的类还是从更高优先级的classpath路径被加载了

方案1：依赖排除+版本锁定（优先使用，改造成本最低） #

直接在项目构建配置中排除nebula-spark-connector传递进来的所有http相关依赖，全局锁定为ES 7.9.2匹配的版本。
如果是Maven项目，在引入connector的依赖块中添加排除规则：

<dependency>
    <groupId>com.vesoft</groupId>
    <artifactId>nebula-spark-connector</artifactId>
    <version>${你实际使用的connector版本}</version>
    <exclusions>
        <exclusion>
            <groupId>org.apache.httpcomponents</groupId>
            <artifactId>httpasyncclient</artifactId>
        </exclusion>
        <exclusion>
            <groupId>org.apache.httpcomponents</groupId>
            <artifactId>httpcore</artifactId>
        </exclusion>
        <exclusion>
            <groupId>org.apache.httpcomponents</groupId>
            <artifactId>httpcore-nio</artifactId>
        </exclusion>
        <exclusion>
            <groupId>org.apache.httpcomponents</groupId>
            <artifactId>httpclient</artifactId>
        </exclusion>
    </exclusions>
</dependency>

然后在dependencyManagement节点全局锁定http依赖版本：

<dependencyManagement>
    <dependencies>
        <dependency>
            <groupId>org.apache.httpcomponents</groupId>
            <artifactId>httpasyncclient</artifactId>
            <version>4.1.4</version>
        </dependency>
        <dependency>
            <groupId>org.apache.httpcomponents</groupId>
            <artifactId>httpcore</artifactId>
            <version>4.4.14</version>
        </dependency>
        <dependency>
            <groupId>org.apache.httpcomponents</groupId>
            <artifactId>httpcore-nio</artifactId>
            <version>4.4.14</version>
        </dependency>
        <dependency>
            <groupId>org.apache.httpcomponents</groupId>
            <artifactId>httpclient</artifactId>
            <version>4.5.13</version>
        </dependency>
    </dependencies>
</dependencyManagement>

如果是Spark集群提交作业，额外加两个提交参数，让用户jar包的类优先被加载，避免集群内置的旧版本依赖抢占优先级：

--conf spark.driver.userClassPathFirst=true
--conf spark.executor.userClassPathFirst=true

方案2：修正shade重打包配置 #

如果必须用shade方式处理依赖，要在maven-shade-plugin的配置中补全http包的重定位规则，把connector自带的http类全部重写到独立路径下，和ES依赖的http包隔离开：

<relocation>
    <pattern>org.apache.http</pattern>
    <shadedPattern>shaded.nebula.org.apache.http</shadedPattern>
</relocation>

注意：这个规则必须覆盖所有org.apache.http开头的包路径，漏了任何一个http相关的子包都会导致冲突复现

方案3：类加载隔离（适合复杂集群环境） #

如果集群环境依赖复杂，前两种方案都不生效，可以把nebula-spark-connector相关依赖放在独立目录，通过自定义类加载器加载connector相关类，和ES客户端的类加载器完全隔离，不过这个方案改造成本较高，非必要不选。

修改配置后，执行依赖树检查命令，确认整个依赖树中没有其他版本的http相关包：

• Maven项目执行：mvn dependency:tree | grep http
• Gradle项目执行：gradle dependencies | grep http
  确认所有http依赖版本都和ES 7.9.2要求的版本一致后，再启动服务即可。

内容的提问来源于stack exchange，提问作者sim