Elasticsearch noop更新响应无法反序列化的问题咨询

当执行脚本化更新（请求路径/my_idx/_update/my_doc_id）且条件未满足时，Elasticsearch 7.2.0返回如下noop响应：

{
    "_index": "my_idx",
    "_type": "_doc",
    "_id": "my_doc_id",
    "_version": 4,
    "result": "noop",
    "_shards": {
        "total": 0,
        "successful": 0,
        "failed": 0
    }
}

但该响应无法被elasticsearch-java客户端（版本8.4.2）反序列化为UpdateResponse<TDocument>，原因是其父类WriteResponseBase的构造器强制要求primaryTerm和seqNo字段，而这两个字段在noop响应中缺失：

protected WriteResponseBase(AbstractBuilder<?> builder) {
    this.id = ApiTypeHelper.requireNonNull(builder.id, this, "id");
    this.index = ApiTypeHelper.requireNonNull(builder.index, this, "index");
    this.primaryTerm = ApiTypeHelper.requireNonNull(builder.primaryTerm, this, "primaryTerm");
    this.result = ApiTypeHelper.requireNonNull(builder.result, this, "result");
    this.seqNo = ApiTypeHelper.requireNonNull(builder.seqNo, this, "seqNo");
    this.shards = ApiTypeHelper.requireNonNull(builder.shards, this, "shards");
    this.type = builder.type;
    this.version = ApiTypeHelper.requireNonNull(builder.version, this, "version");
    this.forcedRefresh = builder.forcedRefresh;
}

已通过hack手段让Elasticsearch服务端7.2.0与Java客户端8.4.2基本兼容，现在需要确认两个问题：

1. primaryTerm和seqNo在Elasticsearch 7.2.0的noop响应中是否属于可选字段？
2. 除了捕获异常处理外，是否存在其他可行的解决办法？

1. 关于primaryTerm和seqNo的可选性 #

在Elasticsearch 7.2.0中，primaryTerm和seqNo确实是可选字段。这两个字段用于实现乐观并发控制和数据版本追踪，但仅在文档实际发生变更（如创建、更新、删除）时才会返回。当更新操作触发noop（无实际变更）时，服务端不会生成新的序列号和主分片任期，因此响应中不会包含这两个字段。

2. 除捕获异常外的解决办法 #

（1）自定义反序列化逻辑 #

通过Jackson的Module扩展客户端的反序列化规则，为WriteResponseBase类添加自定义反序列化器：

• 当响应中不存在primaryTerm和seqNo时，默认赋值为0或其他合法默认值；
• 注册该自定义模块到Elasticsearch客户端的RestClientBuilder中，覆盖默认的反序列化行为。

示例代码片段：

public class WriteResponseBaseDeserializer extends StdDeserializer<WriteResponseBase> {
    public WriteResponseBaseDeserializer() {
        super(WriteResponseBase.class);
    }

    @Override
    public WriteResponseBase deserialize(JsonParser p, DeserializationContext ctxt) throws IOException {
        JsonNode node = p.getCodec().readTree(p);
        // 提取必填字段
        String id = node.get("_id").asText();
        String index = node.get("_index").asText();
        String result = node.get("result").asText();
        ShardStatistics shards = p.getCodec().treeToValue(node.get("_shards"), ShardStatistics.class);
        long version = node.get("_version").asLong();
        
        // 处理可选字段，默认值设为0
        long primaryTerm = node.has("primary_term") ? node.get("primary_term").asLong() : 0;
        long seqNo = node.has("_seq_no") ? node.get("_seq_no").asLong() : 0;
        
        // 构建WriteResponseBase实例（借助反射绕过Builder的访问限制）
        try {
            Constructor<?> builderConstructor = Class.forName("org.elasticsearch.action.support.WriteResponseBase$Builder").getDeclaredConstructor();
            builderConstructor.setAccessible(true);
            Object builder = builderConstructor.newInstance();
            
            // 调用Builder的setter方法
            Method setId = builder.getClass().getMethod("id", String.class);
            setId.invoke(builder, id);
            Method setIndex = builder.getClass().getMethod("index", String.class);
            setIndex.invoke(builder, index);
            Method setPrimaryTerm = builder.getClass().getMethod("primaryTerm", long.class);
            setPrimaryTerm.invoke(builder, primaryTerm);
            Method setResult = builder.getClass().getMethod("result", WriteResponse.Result.class);
            setResult.invoke(builder, WriteResponse.Result.valueOf(result.toUpperCase()));
            Method setSeqNo = builder.getClass().getMethod("seqNo", long.class);
            setSeqNo.invoke(builder, seqNo);
            Method setShards = builder.getClass().getMethod("shards", ShardStatistics.class);
            setShards.invoke(builder, shards);
            Method setVersion = builder.getClass().getMethod("version", long.class);
            setVersion.invoke(builder, version);
            
            Method buildMethod = builder.getClass().getMethod("build");
            return (WriteResponseBase) buildMethod.invoke(builder);
        } catch (Exception e) {
            throw new IOException("Failed to deserialize WriteResponseBase", e);
        }
    }
}

注册反序列化器到客户端：

RestClientBuilder builder = RestClient.builder(new HttpHost("localhost", 9200));
JacksonJsonpMapper jsonMapper = new JacksonJsonpMapper();
ObjectMapper objectMapper = jsonMapper.objectMapper();
SimpleModule module = new SimpleModule();
module.addDeserializer(WriteResponseBase.class, new WriteResponseBaseDeserializer());
objectMapper.registerModule(module);
RestHighLevelClient client = new RestHighLevelClient(
    builder.setHttpClientConfigCallback(httpClientBuilder -> 
        httpClientBuilder.setDefaultIOReactorConfig(IOReactorConfig.custom().setIoThreadCount(Runtime.getRuntime().availableProcessors()).build())
    ).setMapper(jsonMapper)
);

（2）修改客户端依赖的反序列化约束 #

通过字节码修改工具（如ASM、ByteBuddy）修改WriteResponseBase的构造器，移除primaryTerm和seqNo的非空校验：

• 定位到WriteResponseBase类的构造器，将ApiTypeHelper.requireNonNull(builder.primaryTerm, this, "primaryTerm")替换为builder.primaryTerm != null ? builder.primaryTerm : 0；
• 同理处理seqNo字段，使其允许空值并赋予默认值。
  这种方式需要注意客户端版本兼容性，升级客户端时需重新处理。

（3）在服务端添加默认字段 #

通过Elasticsearch的插件或自定义响应拦截器，在noop响应中主动添加primary_term和_seq_no字段，默认值设为0。但这种方式需要修改服务端逻辑，对于已上线的集群风险较高，需谨慎操作。

内容的提问来源于stack exchange，提问作者linusv