kellow
kellow

Reputation: 83

How to create subquery with elasticsearch

I need to create a query on elasticsearch for the condition below.

When the last element added with the "rabbitmq.queue.name" attribute is equal to "service_test_error" and the value from "rabbitmq.queue.messages.total.count" is different of "0"

The sql query below works for my search, but I could not to do the same query with elasticsearch

select * from metric where rabbitmq.queue.messages.total.count != '0' and rabbitmq.queue.name = 'service_test_error' and timestamp = (select max(timestamp) from metric where rabbitmq.queue.name = 'service_test_error')

These records below are examples that exist in my metric-xpto index

[
  {
    "_index": "metric-xpto",
    "_type": "_doc",
    "_id": "jYP1WnEBmYyEo7K68Zme",
    "_version": 1,
    "_score": null,
    "_source": {
      "@timestamp": "2020-04-08T18:03:14.899Z",
      "rabbitmq": {
        "queue": {
          "name": "service_test_error",
          "messages": {
            "total": {
              "count": 0
            }
          }
        }
      }
    }
  },
  {
    "_index": "metric-xpto",
    "_type": "_doc",
    "_id": "jYP1WnEBmYyEo7K68Zme",
    "_version": 1,
    "_score": null,
    "_source": {
      "@timestamp": "2020-04-07T18:03:14.899Z",
      "rabbitmq": {
        "queue": {
          "name": "service_test_error",
          "messages": {
            "total": {
              "count": 3
            }
          }
        }
      }
    }
  },
  {
    "_index": "metric-xpto",
    "_type": "_doc",
    "_id": "jYP1WnEBmYyEo7K68Zme",
    "_version": 1,
    "_score": null,
    "_source": {
      "@timestamp": "2020-04-03T17:03:14.899Z",
      "rabbitmq": {
        "queue": {
          "name": "service_alpha_test_error",
          "messages": {
            "total": {
              "count": 8
            }
          }
        }
      }
    }
  },
  {
    "_index": "metric-xpto",
    "_type": "_doc",
    "_id": "jYP1WnEBmYyEo7K68Zme",
    "_version": 1,
    "_score": null,
    "_source": {
      "@timestamp": "2020-04-03T18:03:14.899Z",
      "rabbitmq": {
        "queue": {
          "name": "service_test_error",
          "messages": {
            "total": {
              "count": 8
            }
          }
        }
      }
    }
  }
]

How can I create a similar query using elasticsearch?

Upvotes: 1

Views: 3189

Answers (1)

jaspreet chahal
jaspreet chahal

Reputation: 9109

In query part you can only fetch top most record and then check has to be in client side on message count.

In aggregation part this can be done.

{
  "size": 0,
  "query": {
    "term": {
      "rabbitmq.queue.name.keyword": {
        "value": "service_test_error"
      }
    }
  },
  "aggs": {
    "date": {
      "terms": {
        "field": "@timestamp",
        "size": 1,
        "order": {
          "_term": "desc"
        }
      },
      "aggs": {
        "message_count": {
          "terms": {
            "field": "rabbitmq.queue.messages.total.count",
            "size": 10
          },
          "aggs": {
            "filter_count": {
              "bucket_selector": {
                "buckets_path": {
                  "key": "_key"
                },
                "script": "if(params.key>0) return true; else return false;"
              }
            }
          }
        },
        "select_timestamp": {
          "bucket_selector": {
            "buckets_path": {
              "key": "message_count._bucket_count"
            },
            "script": "if(params.key>0) return true;else false;"
          }
        }
      }
    }
  }
}

Result: If top most records message count is zero then buckets will be empty else there will be data

 "hits" : {
    "total" : {
      "value" : 7,
      "relation" : "eq"
    },
    "max_score" : null,
    "hits" : [ ]
  },
  "aggregations" : {
    "date" : {
      "doc_count_error_upper_bound" : 0,
      "sum_other_doc_count" : 6,
      "buckets" : [ ]
    }
  }

Elastic search doesn't have joins , So in query it is not possible to compare one document with another

Upvotes: 2

Related Questions