一、三者之间的区别

1. from size：

   1. 深度分页或者size特别大的情况，会出现deep pagination问题。且es的自保机制max_result_window是10000，当查询数量超过一万就会报错
   2. 该查询的实现原理类似于mysql中的limit，比如查询第10001条数据，那么需要将前面的1000条都拿出来，进行过滤，最终才得到数据。（性能较差，实现简单，适用于少量数据）

2. search after

   1. search_after缺点是不能够随机跳转分页，只能是一页一页的向后翻（当有新数据进来，也能实时查询到），并且需要至少指定一个唯一不重复字段来排序（一般是_id和时间字段）
   2. 当使用search_after时，from值必须设置为0或者-1

3. scroll

   1. 高效进行滚动查询，首次查询会在内存中保存一个历史快照以及游标（scroll_id）,记录当前消息查询的终止位置，下次查询的时候将基于游标进行消费（性能良好，不具备实时性，一般是用于大量数据导出或者索引重建）

二、代码测试类

1. from size

package com.example.es.test;import org.apache.http.HttpHost;
import org.elasticsearch.action.search.SearchRequest;
import org.elasticsearch.action.search.SearchResponse;
import org.elasticsearch.client.RequestOptions;
import org.elasticsearch.client.RestClient;
import org.elasticsearch.client.RestHighLevelClient;
import org.elasticsearch.search.SearchHit;
import org.elasticsearch.search.builder.SearchSourceBuilder;
import org.elasticsearch.search.sort.SortBuilders;
import org.elasticsearch.search.sort.SortOrder;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;import java.util.ArrayList;
import java.util.List;
import java.util.Map;/*** @author * @Description  es的from-size用法* @date 2022/01/26 10:04*/
public class ESTest_from_size {public static final Logger logger = LoggerFactory.getLogger(ESTest_searchAfter.class);public static void main(String[] args) throws Exception{long startTime = System.currentTimeMillis();// 创建ES客户端RestHighLevelClient esClient = new RestHighLevelClient(RestClient.builder(new HttpHost("localhost", 9200, "http")));// 1、创建searchRequestSearchRequest searchRequest = new SearchRequest("audit2");// 2、指定查询条件SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();//必须加上track_total_hits，不然就只显示10000// 页面上的第一页等同于在es中的 0sourceBuilder.from(0);// 每页多少条数据sourceBuilder.size(1000);// 设置唯一排序值定位sourceBuilder.sort(SortBuilders.fieldSort("operationtime").order(SortOrder.DESC));//将sourceBuilder对象添加到搜索请求中searchRequest.source(sourceBuilder);// 发送请求SearchResponse searchResponse = esClient.search(searchRequest, RequestOptions.DEFAULT);SearchHit[] hits = searchResponse.getHits().getHits();List<Map<String, Object>> result = new ArrayList<>();if (hits != null && hits.length > 0) {for (SearchHit hit : hits) {// 获取需要数据Map<String, Object> sourceAsMap = hit.getSourceAsMap();result.add(sourceAsMap);}}logger.info("查询出来的数据个数为：{}", result.size());// 关闭客户端esClient.close();logger.info("运行时间: " + (System.currentTimeMillis() - startTime) + "ms");}
}

运行结果：

10:08:40.466 [main] INFO com.example.es.test.ESTest_searchAfter - 查询出来的数据个数为：1000
10:08:40.474 [main] INFO com.example.es.test.ESTest_searchAfter - 运行时间: 1506ms

现象：

如果from size 查询的数据超过10000条，会报错误 

2、search after

package com.example.es.test;import org.apache.http.HttpHost;
import org.elasticsearch.action.search.SearchRequest;
import org.elasticsearch.action.search.SearchResponse;
import org.elasticsearch.client.RequestOptions;
import org.elasticsearch.client.RestClient;
import org.elasticsearch.client.RestHighLevelClient;
import org.elasticsearch.search.SearchHit;
import org.elasticsearch.search.builder.SearchSourceBuilder;
import org.elasticsearch.search.sort.SortBuilders;
import org.elasticsearch.search.sort.SortOrder;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;import java.util.ArrayList;
import java.util.List;
import java.util.Map;/*** @author* @Description  es 的search_after方法* @date 2022/01/11 14:04*/
public class ESTest_searchAfter {public static final Logger logger = LoggerFactory.getLogger(ESTest_searchAfter.class);public static void main(String[] args) throws Exception{long startTime = System.currentTimeMillis();// 创建ES客户端RestHighLevelClient esClient = new RestHighLevelClient(RestClient.builder(new HttpHost("localhost", 9200, "http")));// 1、创建searchRequestSearchRequest searchRequest = new SearchRequest("audit2");// 2、指定查询条件SearchSourceBuilder sourceBuilder = new SearchSourceBuilder().trackTotalHits(true);//必须加上track_total_hits，不然就只显示10000//设置每页查询的数据个数sourceBuilder.size(1000);// 设置唯一排序值定位sourceBuilder.sort(SortBuilders.fieldSort("operationtime").order(SortOrder.DESC));//多条件查询//将sourceBuilder对象添加到搜索请求中searchRequest.source(sourceBuilder);// 发送请求SearchResponse searchResponse = esClient.search(searchRequest, RequestOptions.DEFAULT);SearchHit[] hits1 = searchResponse.getHits().getHits();List<Map<String, Object>> result = new ArrayList<>();if (hits1 != null && hits1.length > 0) {do {for (SearchHit hit : hits1) {// 获取需要数据Map<String, Object> sourceAsMap = hit.getSourceAsMap();result.add(sourceAsMap);}// 取得最后得排序值sort，用于记录下次将从这个地方开始取数SearchHit[] hits = searchResponse.getHits().getHits();Object[] lastNum = hits[hits.length - 1].getSortValues();// 设置searchAfter的最后一个排序值sourceBuilder.searchAfter(lastNum);searchRequest.source(sourceBuilder);// 进行下次查询searchResponse = esClient.search(searchRequest, RequestOptions.DEFAULT);} while (searchResponse.getHits().getHits().length != 0);}logger.info("查询出来的数据个数为：{}", result.size());// 关闭客户端esClient.close();logger.info("运行时间: " + (System.currentTimeMillis() - startTime) + "ms");}}

运行结果：

16:11:44.057 [main] INFO com.example.es.test.ESTest_searchAfter - 查询出来的数据个数为：64000
16:11:44.061 [main] INFO com.example.es.test.ESTest_searchAfter - 运行时间: 20979ms

现象：audit2该索引里面总共就69873条数据，控制台打印的信息是每1000条的查询打印出来，最终查询出来64000条记录，还有5873条数据丢失了。还有size如果超过10000，也会报错。

我自己的疑问：search after既然不能跳页查询，只能一页一页的查询出来，那前端调用这个接口后端不是还是返回全部的数据吗。那如果前端设置成向下滚动查询，然后滚轮向下几页后端就返回几页数据，后端不是会更省查询的时间吗。现在search after还是一次性将数据查询出来，只是内部它是一页一页查询出来的，最终展示出来的还是全部的数据。这个我有疑问，我应该怎么与前端对接这个接口。

3、scroll

package com.example.es.test;import org.apache.http.HttpHost;
import org.elasticsearch.action.search.*;
import org.elasticsearch.client.RequestOptions;
import org.elasticsearch.client.RestClient;
import org.elasticsearch.client.RestHighLevelClient;
import org.elasticsearch.common.unit.TimeValue;
import org.elasticsearch.index.query.QueryBuilders;
import org.elasticsearch.search.SearchHit;
import org.elasticsearch.search.builder.SearchSourceBuilder;
import org.elasticsearch.search.sort.SortBuilders;
import org.elasticsearch.search.sort.SortOrder;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;import java.util.ArrayList;
import java.util.List;
import java.util.Map;/*** @author * @Description  java 实现scroll滚动查询* @date 2021/12/08 14:09*/
public class ESTest_Scroll {public static final Logger logger = LoggerFactory.getLogger(ESTest_Scroll.class);public static void main(String[] args) throws Exception{long startTime = System.currentTimeMillis();// 创建ES客户端RestHighLevelClient esClient = new RestHighLevelClient(RestClient.builder(new HttpHost("localhost", 9200, "http")));// 1、创建searchRequestSearchRequest searchRequest = new SearchRequest("audit2");// 2、指定scroll信息searchRequest.scroll(TimeValue.timeValueMinutes(1L));// 3、指定查询条件SearchSourceBuilder searchSourceBuilder = new SearchSourceBuilder();searchSourceBuilder.size(1000);searchSourceBuilder.sort(SortBuilders.fieldSort("operationtime").order(SortOrder.DESC));//多条件查询searchRequest.source(searchSourceBuilder);//4、获取返回结果scrollId，sourceSearchResponse searchResponse = esClient.search(searchRequest, RequestOptions.DEFAULT); //通过发送初始搜索请求来初始化搜索上下文String scrollId = searchResponse.getScrollId();SearchHit[] searchHits = searchResponse.getHits().getHits();List<Map<String, Object>> result = new ArrayList<>();for (SearchHit hit: searchHits) {result.add(hit.getSourceAsMap());}// java也是一样要查询两次，先把我们的首页给查询出来// 查询出来之后我们要获取他的id// 然后利用他的id去查询他的下一页while (true) {//5、循环  -  创建 SearchScrollRequest  创建一个新的搜索滚动请求，保存最后返回的滚动标识符和滚动间隔// 获取 scrollId 去查询下一页SearchScrollRequest scrollRequest = new SearchScrollRequest(scrollId);//6、指定scrollId的生存时间scrollRequest.scroll(TimeValue.timeValueMinutes(1L));//7、执行查询获取返回结果SearchResponse scrollResp = esClient.scroll(scrollRequest, RequestOptions.DEFAULT);//8、判断是否查询到了数据，输出SearchHit[] hits = scrollResp.getHits().getHits();//循环输出下一页if (hits != null && hits.length > 0) {for (SearchHit hit : hits) {result.add(hit.getSourceAsMap());}} else {//9、判断没有查询到数据，退出循环break;}}//查完之后我们把存进缓存的id给删除  完成滚动后，清除滚动上下文//10、创建ClearScrollRequestClearScrollRequest clearScrollRequest = new ClearScrollRequest();//11、指定scrollIdclearScrollRequest.addScrollId(scrollId);//12、删除scrollIdClearScrollResponse clearScrollResponse = esClient.clearScroll(clearScrollRequest, RequestOptions.DEFAULT);//13、输出结果boolean succeeded = clearScrollResponse.isSucceeded();logger.info("删除scrollId：{}", succeeded);logger.info("查询总个数：{}", result.size());// 关闭客户端esClient.close();logger.info("运行时间: " + (System.currentTimeMillis() - startTime) + "ms");}}

运行结果：

16:20:54.794 [main] INFO com.example.es.test.ESTest_Scroll - 删除scrollId：true
16:20:54.795 [main] INFO com.example.es.test.ESTest_Scroll - 查询总个数：69873
16:20:54.797 [main] INFO com.example.es.test.ESTest_Scroll - 运行时间: 5716ms

现象：

audit2该索引里面总共就69873条数据，最终查询出来69873条记录，一条记录都没有丢失。还有size如果超过10000，也会报错。很奇怪，search after会丢失数据，而scroll一条记录没有丢失。