Elastic Search - Is there a way to ignore certain documents based on condition during aggregation?

Question

I am logging analytics for the flow. For the field "start" set to "true" when a flow starts and "true" will be set to the field "end" on flow ends. Few flows may not contain "end" field "true". I want to find where the flow exactly stops.

I have tried with nested aggregation, but not able to get the documents for the flows which are not ends.

This is the data stored in the elastic search

    [
    {
        "date": 1565094409535,
        "start": "true",
        "end": "",
        "message": "Select Option",
        "context": "third",
        "account_id": "123"
    },
    {
        "date": 1565094411152,
        "start": "",
        "end": "",
        "message": "Select Me",
        "context": "third",
        "account_id": "123"
    },
    {
        "date": 1565094409652,
        "start": "true",
        "end": "",
        "message": "Select option",
        "context": "third",
        "account_id": "123"
    },
    {
        "date": 1565094409751,
        "start": "",
        "end": "",
        "message": "Select Me",
        "context": "third",
        "account_id": "123"
    },
    {
        "date": 1565094411187,
        "start": "",
        "end": "true",
        "message": "Bye Bye",
        "context": "third",
        "account_id": "123"
    },
    {
        "date": 1565094411211,
        "start": "true",
        "end": "",
        "message": "Select option",
        "context": "third",
        "account_id": "123"
    },
    {
        "date": 1565094411311,
        "start": "true",
        "end": "",
        "message": "How are you",
        "context": "second",
        "account_id": "123"
    }
]

Query used:

 {
"size": 0,
"query": {
    "bool": {
        "must": [{
                "term": {
                    "context.keyword": "third"
                }
            }
        ]
    }
},
"aggs": {
    "sessions": {
        "terms": {
            "field": "account_id.keyword",
            "size": 25000
        },

        "aggs": {
            "top_sessions_hits": {
                "top_hits": {
                    "sort": [{
                        "date": {
                            "order": "asc"
                        }
                    }],
                    "_source": {
                        "includes": ["date", "message", "account_id", "start", "end", "context"]
                    },
                    "size": 10000
                }
            }
        }
    }
}

}

I am getting the below output

    {
      "took": 37,
      "timed_out": false,
      "_shards": {
        "total": 5,
        "successful": 5,
        "failed": 0
      },
      "hits": {
        "total": 4,
        "max_score": 0,
        "hits": []
      },
      "aggregations": {
        "sessions": {
          "doc_count_error_upper_bound": 0,
          "sum_other_doc_count": 0,
          "buckets": [
            {
              "key": "123",
              "doc_count": 6,
              "top_sessions_hits": {
                "hits": {
                  "total": 6,
                  "max_score": null,
                  "hits": [
                    {
                      "_index": "messages",
                      "_type": "doc",
                      "_id": "********",
                      "_score": null,
                      "_source": {
                        "date": 1565094409535,
                        "start": "true",
                        "end": "",
                        "message": "Select Option",
                        "context": "third",
                        "account_id": "123"
                      },
                      "sort": [
                        1565094409535
                      ]
                    },{
                      "_index": "messages",
                      "_type": "doc",
                      "_id": "********",
                      "_score": null,
                      "_source": {
                        "date": 1565094411152,
                        "start": "",
                        "end": "",
                        "message": "Select Me",
                        "context": "third",
                        "account_id": "123"
                      },
                      "sort": [
                        1565094411152
                      ]
                    },
                    {
                      "_index": "messages",
                      "_type": "doc",
                      "_id": "********",
                      "_score": null,
                      "_source": {
                        "date": 1565094409652,
                        "start": "true",
                        "end": "",
                        "message": "Select option",
                        "context": "third",
                        "account_id": "123"
                      },
                      "sort": [
                        1565094409652
                      ]
                    },
                    {
                      "_index": "messages",
                      "_type": "doc",
                      "_id": "********",
                      "_score": null,
                      "_source": {
                        "date": 1565094409751,
                        "start": "",
                        "end": "",
                        "message": "Select Me",
                        "context": "third",
                        "account_id": "123"
                      },
                      "sort": [
                        1565094409751
                      ]
                    },
                    {
                      "_index": "messages",
                      "_type": "doc",
                      "_id": "********",
                      "_score": null,
                      "_source": {
                        "date": 1565094411187,
                        "start": "",
                        "end": "true",
                        "message": "Bye Bye",
                        "context": "third",
                        "account_id": "123"
                      },
                      "sort": [
                        1565094411187
                      ]
                    },
                    {
                      "_index": "messages",
                      "_type": "doc",
                      "_id": "********",
                      "_score": null,
                      "_source": {
                        "date": 1565094411211,
                        "start": "true",
                        "end": "",
                        "message": "Select option",
                        "context": "third",
                        "account_id": "123"
                      },
                      "sort": [
                        1565094411211
                      ]
                    }
                  ]
                }
              }
            }
          ]
        }
      }
    }

But I don't want to get the document #3, #4 & #5 since the flow was completed.

I am very new to elastic search. Since the core guy went on leave, I am trying to fix a few issues. Please guide me to get the document #1, #2 & #6.

Answer 1

I understand that there are 2 messages in each flow - one with "start":true and one with "end":true . in order to find the flows that have only start, but no end, you need to have a unique identifier on each flow, like lets say flow-id .

In case the messages will contain flow-id you coult run a term aggregation on the flow-id, to count how many messages exists for each flow, and then sort ascending the results based on the _count of aggregation result - the first agg results would be with count=1, therefore flows which have only start and no end.

The query should look like:

GET /flows_index/_search {
"size": 0,
"aggs": {
    "flow_id_agg": {
        "terms": {
            "field": "flow_id",
            "order": {
                "_count": "asc"
            },
            "aggs": {
                "flow_id_samples": {
                    "top_hits": {
                      "sort": [{
                        "date": {
                            "order": "asc"
                        }
                       }],
                      "_source": {
                         "includes": ["date", "message", "account_id", "start", "end", "context"]
                    },
                    "size": 10000
                   }
                }
            }
          }
       }
    }
}

Take a look at a similar need here: Elasticsearch terms aggregation and querying