Elasticsearch在2.0以前版本,删除操作有两种方式,一种是通过id来进行删除,但是这种方式一般不常用,因为id不容易得到;另一种方式是通过先查询操作,然后删除,也就是通过client.prepareDeleteByQuery这种方式来根据条件批量删除数据:
DeleteByQueryResponse response = client.prepareDeleteByQuery("library")
.setQuery(QueryBuilders.termQuery("title", "ElasticSearch"))
.execute().actionGet();
但是Delete by Query在2.0版本及其以上的版本已经被移除了,因为这种方式会自动强制刷新,所以在大量索引并发的情况下,会很快造成内存溢出。
详情可查看:https://www.elastic.co/guide/en/elasticsearch/client/java-api/1.7/delete-by-query.html
那么在2.0以后的版本,我们如何来进行批量的删除呢?
我们可以先通过Search API查询,然后得到需要删除的批量数据的id,然后再通过id来删除,但是这种方式在大批量数据的删除的时候,依然是行不通的。
具体实现代码:
public void deleteByTerm(Client client){
BulkRequestBuilder bulkRequest = client.prepareBulk();
SearchResponse response = client.prepareSearch("megacorp").setTypes("employee")
.setSearchType(SearchType.DFS_QUERY_THEN_FETCH)
.setQuery(QueryBuilders.termQuery("first_name", "xiaoming"))
.setFrom(0).setSize(20).setExplain(true).execute().actionGet();
for(SearchHit hit : response.getHits()){
String id = hit.getId();
bulkRequest.add(client.prepareDelete("megacorp", "employee", id).request());
}
BulkResponse bulkResponse = bulkRequest.get();
if (bulkResponse.hasFailures()) {
for(BulkItemResponse item : bulkResponse.getItems()){
System.out.println(item.getFailureMessage());
}
}else {
System.out.println("delete ok");
}
}
同样通过delete-by-query插件,我们还可以根据type来批量删除数据,这种方式能够删除大批量的数据,他是现将要删除的数据一个一个做标记,然后再删除,于是效率会比较低。下面是官网的说明:https://www.elastic.co/guide/en/elasticsearch/plugins/2.3/plugins-delete-by-query.html
Queries which match large numbers of documents may run for a long time, as every document has to be deleted individually. Don’t use delete-by-query to clean out all or most documents in an index. Rather create a new index and perhaps reindex the documents you want to keep.
可见这种删除方式并不适合大批量数据的删除,因为效率真的是很低,我是亲身体验过了。
这种方式需要先引入delete-by-query插件包,然后使用插件的api来删除:
<dependency>
<groupId>org.elasticsearch.plugin</groupId>
<artifactId>delete-by-query</artifactId>
<version>2.3.2</version>
</dependency>
具体实现代码:
import java.net.InetAddress;
import java.net.UnknownHostException;
import java.util.ResourceBundle;
import java.util.Stack;
import org.elasticsearch.action.deletebyquery.DeleteByQueryAction;
import org.elasticsearch.action.deletebyquery.DeleteByQueryRequestBuilder;
import org.elasticsearch.action.deletebyquery.DeleteByQueryResponse;
import org.elasticsearch.action.search.SearchRequestBuilder;
import org.elasticsearch.action.search.SearchResponse;
import org.elasticsearch.action.search.SearchType;
import org.elasticsearch.client.Client;
import org.elasticsearch.client.transport.TransportClient;
import org.elasticsearch.common.settings.Settings;
import org.elasticsearch.common.transport.InetSocketTransportAddress;
import org.elasticsearch.plugin.deletebyquery.DeleteByQueryPlugin;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.xgd.log.common.ExceptionUtil;
public class EsDeleteByType {
private static final Logger logger = LoggerFactory.getLogger(EsDeleteByType.class);
private Client client;
private static ResourceBundle getEsConfig(){
return ResourceBundle.getBundle("elasticsearch");
}
private void getClient(){
String clusterName = getEsConfig().getString("clusterName");
String hosts = getEsConfig().getString("hosts");
if (hosts == null || clusterName == null) {
throw new IllegalArgumentException("hosts or clusterName was null.");
}
Settings settings = Settings.settingsBuilder().put("cluster.name", clusterName).build();
client = TransportClient.builder()
.addPlugin(DeleteByQueryPlugin.class)
.settings(settings).build();
String[] hostsArray = hosts.split(",");
for(String hostAndPort : hostsArray){
String[] tmpArray = hostAndPort.split(":");
try {
client = ((TransportClient)client).addTransportAddress(new InetSocketTransportAddress(InetAddress.getByName(tmpArray[0]), Integer.valueOf(tmpArray[1])));
} catch (NumberFormatException e) {
logger.error(ExceptionUtil.getTrace(e));
} catch (UnknownHostException e) {
logger.error(ExceptionUtil.getTrace(e));
}
}
}
/**
* 判断一个index中的type是否有数据
* @param index
* @param type
* @return
* @throws Exception
*/
public Boolean existDocOfType(String index, String type) throws Exception {
SearchRequestBuilder builder = client.prepareSearch(index).setTypes(type)
.setSearchType(SearchType.QUERY_THEN_FETCH)
.setSize(1);
SearchResponse response = builder.execute().actionGet();
long docNum = response.getHits().getTotalHits();
if (docNum == 0) {
return false;
}
return true;
}
/**
* 根据type来删除数据
* @param index
* @param types
* @return
*/
public long deleteDocByType(String index, String[] types) {
getClient();
long oldTime = System.currentTimeMillis();
StringBuilder b = new StringBuilder();
b.append("{\"query\":{\"match_all\":{}}}");
DeleteByQueryResponse response = new DeleteByQueryRequestBuilder(client, DeleteByQueryAction.INSTANCE)
.setIndices(index).setTypes(types)
.setSource(b.toString())
.execute().actionGet();
Stack<String> allTypes = new Stack<String>();
for(String type : types){
allTypes.add(type);
}
while(!allTypes.isEmpty()){
String type = allTypes.pop();
while(true){
try {
if (existDocOfType(index, type) == false) {
break;
}
} catch (Exception e) {
logger.error("queryError: " + e.getMessage());
}
}
}
System.out.println(System.currentTimeMillis() - oldTime);
return response.getTotalDeleted();
}
}
那么当我们在开发中,使用到elasticsearch的时候,总会涉及到大批量数据的删除,我们要怎么办呢?
经过很长时间的纠结,我发现使用elasticsearch存储数据的时候,千万不要把所有数据都存储于一个index,这样一个是不利于查询的效率,一个是不利于后面的删除,既然我们不能index中去删除部分的大批量数据,那么我们为啥不改变一种思路呢,就是分索引,然后通过索引来删除数据,例如:我在生产上面,每天有5亿的数据,那么我每天在集群中生成一个index用于存储这5亿的数据,如果我们的elasticsearch集群对数据只要求保存7天的数据,超过7天的数据就可以删除了,这样我们可以通过index直接删除7天以前的数据,这种方式,我们在查询的时候不会在所有数据中查询,只需要在所要查询的时间段内查询,便提高了查询的效率,同时删除效率的问题也解决了,能够很快删除不需要的数据,释放掉磁盘空间。
针对于elasticsearch大批量数据删除效率的问题,目前官网上面也没有一个特别好的解决办法,这种方式算是目前还算能行得通的方式了。