A problem involving parent-child documents and using top_hits

[zp-shadowolf]

Error Description

Index ptedata designed two documents, the session and user. User for the session's parent document. Two documents have same fields (only save the contents of different), only through an rs_type field to distinguish session or user. When aggs use top_hits return field content, but returned the following error.

Code

GET 45_hot/ptedata/_search
{
  "size": 0,
  "query": {
    "bool": {
      "must": [
        {
          "term": {
            "rs_type": {
              "value": "session"
            }
          }
        },
        {
          "range": {
            "v_ymd": {
              "format": "yyyy-MM-dd", 
              "gte": "2017-12-03",
              "lte": "2017-12-03"
            }
          }
        },
        {
          "term": {
            "sid": {
              "value": "45341cfb"
            }
          }
        },
        {
          "nested": {
            "path": "scene",
            "query": {
              "bool": {
                "must": [
                  {
                    "term": {
                      "scene.cutpath": {
                        "value": "https://www.biccamera.com/bc/category/"
                      }
                    }
                  }
                ]
              }
            }
          }
        }
      ]
    }
  },
  "aggs": {
    "pv": {
      "nested": {
        "path": "scene"
      },
      "aggs": {
        "pv2": {
          "filter": {
            "bool": {
              "must": [
                {
                  "term": {
                    "scene.cutpath": "https://www.biccamera.com/bc/category/"
                  }
                }  
              ]
            }
          },
          "aggs": {
            "scr": {
              "nested": {
                "path": "scene.scroll"
              },
              "aggs": {
                "scroll_details": {
                  "top_hits": {
                    "_source": [
                      "scene.scroll.pos",
                      "scene.scroll.vh"
                    ], 
                    "size": 10
                  }
                },
                "scroll_sum": {
                  "filter": {
                    "exists": {
                      "field": "scene.scroll.vh"
                    }
                  }
                }
              }
            }
          }
        }
      }
    }
  }
}

Error

{
  "error": {
    "root_cause": [
      {
        "type": "illegal_state_exception",
        "reason": "Child query must not match same docs with parent filter. Combine them as must clauses (+) to find a problem doc. docId=2147483647, class org.apache.lucene.search.TermScorer"
      }
    ],
    "type": "search_phase_execution_exception",
    "reason": "all shards failed",
    "phase": "query",
    "grouped": true,
    "failed_shards": [
      {
        "shard": 0,
        "index": "45_hot",
        "node": "RPzpfc_3RAC2b5n69X8xbA",
        "reason": {
          "type": "illegal_state_exception",
          "reason": "Child query must not match same docs with parent filter. Combine them as must clauses (+) to find a problem doc. docId=2147483647, class org.apache.lucene.search.TermScorer"
        }
      }
    ]
  },
  "status": 500
}

Thank you very much for your help.

[martijnvg]

@zp-shadowolf Can you tell what ES version you're using?
Do both user and session documents have scene.scroll.* fields and are do they use the same type of field mappings?

This type of error has been reported before and general validation mechanism is needed to prevent this error #23290, but I'm still unsure how this can happen in your case, because in previous reports a nested query was used inside a filter aggregation that was then nested under a nested aggregation.

[zp-shadowolf]

Yes, es now can not use more than one type in the same index, so the user and session are using the same type, using the join method.

[zp-shadowolf]

@martijnvg

[martijnvg]

@zp-shadowolf Can you share your mapping or at least the relevant portions of it?

[mayya-sharipova]

@zp-shadowolf Can you please provide the mappings and ES version or will be closing the issue as we don't have enough feedback.

[colings86]

No further feedback. Closing this issue but if you have the requested information can you share it and we can look at reopening the issue