Java中ES5.3批量插入_bulk实现方案

ES5.5.3众所周知是十分尴尬的，用不了RestHighLevelClient， TransportClient可以用但是网上各种不建议使用，其实TransportClient在7.0.0才开始废弃，8.0.0之后才正式移除，现在的5.5.3版本完全是可以光明正大使用的，不过考虑到后面的升级，还是做出了妥协。

这里分享一下两种方案：

方案一：使用TransportClient：

1.pom文件

repository务必要指定，在mvnrepository是没有的，另外网上给出大多数都不用引x-pack-transport，直接使用transport，这对于自建的MySQL数据中是没有问题的，但是要同步阿里云的RDS就会报错

... <dependency>     <groupId>org.elasticsearch.clientgroupId>     <artifactId>transportartifactId>     <version>5.5.3version> dependency> <dependency>     <groupId>org.elasticsearch.plugingroupId>     <artifactId>transport-netty3-clientartifactId>     <version>5.5.3version> dependency> <dependency>     <groupId>org.elasticsearch.clientgroupId>     <artifactId>x-pack-transportartifactId>     <version>5.5.3version> dependency> ... <repository>     <id>elasticsearch-releasesid>     <url>https://artifacts.elastic.co/mavenurl>     <releases>         <enabled>trueenabled>     releases>     <snapshots>         <enabled>falseenabled>     snapshots> repository> ...

2. 配置

需要注意的是的（1）网上普遍的TransportClient都是通过PreBuiltTransportClient，这在RDS上面也是行不通的，需要使用PreBuiltXPackTransportClient；（2）阿里云的clusterName是es的实例id，并不是名称。

 

@Configuration @Log4j2 public class ElasticSearchConfig {       @Value("${icec.elasticsearch.host}")     private String host;       @Value("${icec.elasticsearch.tcpPort}")     private int tcpPort;       @Value("${icec.elasticsearch.clusterName}")     private String clusterName;       @Value("${icec.elasticsearch.username}")     private String username;       @Value("${icec.elasticsearch.password}")     private String password;       @Bean     public TransportClient transportClient() {           TransportClient transportClient = null;         TransportClient preBuiltTransportClient = new PreBuiltXPackTransportClient(Settings.builder()                 .put("cluster.name", clusterName)                 .put("xpack.security.user", username + ":" + password)                 .put("client.transport.sniff", false)                 .build());           try {             transportClient = preBuiltTransportClient                     .addTransportAddress(new InetSocketTransportAddress(InetAddress.getByName(host), tcpPort));         } catch (UnknownHostException e) {             log.warn(e);         }           return transportClient;     } }

3. 使用示例：

public void batchInsert(List       if (CollectionUtils.isEmpty(datas)) {         return;     }       BulkRequestBuilder bulkRequestBuilder = transportClient.prepareBulk();     datas.forEach(data -> bulkRequestBuilder.add(transportClient.prepareIndex(ES_INDEX, ES_TYPE, (String) data.get(OrderEsConstant.ORDER_ID)).setSource(data)));       BulkResponse bulkResponse = bulkRequestBuilder.execute().actionGet();       List failuerMessages = Arrays.asList(bulkResponse.getItems())             .stream().filter(p -> StringUtils.isNotBlank(p.getFailureMessage()))             .map(BulkItemResponse::getFailureMessage).collect(Collectors.toList());       if (CollectionUtils.isNotEmpty(failuerMessages)) {         log.info("同步失败订单->{}", failuerMessages);     }       log.info("批处理完成 总条数：{}", datas.size());   }

 

方案二：手写批量操作，通过RestClient实现

public void batchInsert(List       StringBuilder bulkRequestBody = new StringBuilder();     int count = 1;       for (Map order : orders) {           String actionMetaData = String.format("{ \"index\" : {\"_id\" : \"%s\"} }%n", order.get(OrderEsConstant.ORDER_ID));         String orderJson = JSON.toJSONString(order, SerializerFeature.WriteNullStringAsEmpty);         bulkRequestBody.append(actionMetaData);         bulkRequestBody.append(orderJson);         bulkRequestBody.append("\n");           if (count % 5000 == 0 || count == orders.size()) {               Response response;             Map result = Maps.newHashMap();             String router = String.format(BASIC_FORMAT, ES_INDEX, ES_TYPE, "_bulk");               try {                 response = restClient.performRequest(POST, router, Collections.emptyMap(), new StringEntity(bulkRequestBody.toString(), ContentType.APPLICATION_JSON));                 result = objectMapper.readValue(response.getEntity().getContent(), Map.class);             } catch (IOException e) {                 log.info("ES批量插入异常");             }               List"items")).stream().map(p -> (Map) p.get("index")).collect(Collectors.toList());             List errorMsgs = itemResults.stream().filter(p -> 200 != (Integer) p.get("status")).map(this::getErrorMsg).collect(Collectors.toList());             if (CollectionUtils.isNotEmpty(errorMsgs)) {                 log.warn("数据插入失败 -> {}", errorMsgs);             }               log.info("批处理完成，总条数： {}", itemResults.size());         }         count++;     } }

方案一可以一劳永逸，后续一些高级的操作都可以通过TransportClient去实现，但是如果ES升级的话，就会有问题，官方是在8.0.0才移除该功能，但是阿里云的数据订阅的话只支持到5.5.3；

方案二暂时不用考虑后续升级的问题，但是只要涉及到高级操作，都需要自己去实现，例如upsert操作。