如何知道 DSL Elasticsearch 解析的最终查询？答案

【问题标题】：How to know the final query DSL Elasticsearch parsed?如何知道 DSL Elasticsearch 解析的最终查询？
【发布时间】：2017-12-27 03:11:28
【问题描述】：

我知道以下两个查询可以获得相同的结果。但是有没有办法通过 Elasticsearch 查看最终解析的查询，所以我当然可以知道它们是相同的？（或者，它们实际上并不完全相同，也许一个比另一个花费更少的时间？）

查询 1：

GET /_search
{
  "query": {
    "constant_score": {
      "filter": {
        "term": {
          "price": 20
        }
      }
    }
  }
}

查询 2：

GET /_search
{
  "query": {
    "term": {
      "price": 20
    }
  }
}

【问题讨论】：

标签： elasticsearch elasticsearch-5 elasticsearch-dsl

【解决方案1】：

为了测试和分析您的查询，您可以使用Slow Log，它允许将查询和获取阶段记录到日志文件中。它是高度可配置的——您可以为每个索引定义“慢查询”的含义。

名为“sample”的索引的简单示例（出于测试目的，它将记录时间设置为“0s” - 您可以设置自己的阈值）：

先关闭索引：

curl -X POST http://127.0.0.1:9200/sample/_close

然后配置慢日志：

curl -X PUT \
  'http://127.0.0.1:9200/sample/_settings?preserve_existing=true' \
  -d '{
    "index.indexing.slowlog.threshold.index.debug" : "0s",
    "index.search.slowlog.threshold.fetch.debug" : "0s",
    "index.search.slowlog.threshold.query.debug" : "0s"
}'

打开索引：

curl -X POST http://127.0.0.1:9200/sample/_open

在执行您在帖子中提供的较短查询后（我有 5 个分片，并且针对每个分片执行查询）：

[index.search.slowlog.query] [sample][1] took[594.1micros], ..., source[{"query":{"term":{"price":{"value":33,"boost":1.0}}}}], 
[index.search.slowlog.query] [sample][3] took[649.4micros], ..., source[{"query":{"term":{"price":{"value":33,"boost":1.0}}}}], 
[index.search.slowlog.query] [sample][4] took[575.6micros], ..., source[{"query":{"term":{"price":{"value":33,"boost":1.0}}}}], 
[index.search.slowlog.query] [sample][2] took[1.2ms],       ..., source[{"query":{"term":{"price":{"value":33,"boost":1.0}}}}], 
[index.search.slowlog.query] [sample][0] took[4.3ms],       ..., source[{"query":{"term":{"price":{"value":33,"boost":1.0}}}}], 
...

执行您在帖子中提供的较长查询后：

[index.search.slowlog.query] [sample][1] took[13.2ms], ..., source[{"query":{"constant_score":{"filter":{"term":{"price":{"value":33,"boost":1.0}}},"boost":1.0}}}], 
[index.search.slowlog.query] [sample][4] took[13.2ms], ..., source[{"query":{"constant_score":{"filter":{"term":{"price":{"value":33,"boost":1.0}}},"boost":1.0}}}], 
[index.search.slowlog.query] [sample][3] took[14.7ms], ..., source[{"query":{"constant_score":{"filter":{"term":{"price":{"value":33,"boost":1.0}}},"boost":1.0}}}], 
[index.search.slowlog.query] [sample][2] took[15.5ms], ..., source[{"query":{"constant_score":{"filter":{"term":{"price":{"value":33,"boost":1.0}}},"boost":1.0}}}], 
[index.search.slowlog.query] [sample][0] took[15.5ms], ..., source[{"query":{"constant_score":{"filter":{"term":{"price":{"value":33,"boost":1.0}}},"boost":1.0}}}],
...

当然，这只是一次尝试，但它对于更深入的测试和分析非常有用。

【讨论】：