Suddenly need to add track_scores to queries

Cluster: 2c4be5

Hello support!

We've been developing against 2.3.3 locally and recently created and deployed our application to your Cloud Service (cluster: 2c4be5, v2.3.3). What I've noticed is that our queries seem to now require 'track_scores' in order to limit our search results appropriately (we're sorting by published_at).

With our local install of ES 2.3.3, and setting track_scores: false or leaving out all together, we receive 34 results. When we set track_scores: true, we also receive 34 results. In production (in the cluster), if we set track_scores: false or leave out, we get 2186 results, but if we set track_scores: true, we get the expected 34.

Here's an example of the query:

    GET /video_catalogs/catalog_entry/_search?from=0&size=20
    {  
      "track_scores":false,
       "query":{  
          "query_string":{  
             "default_operator":"AND",
             "query":"waterparks OR waterpark"
          }
       },
       "filter":{  
          "and":[  
             {  
                "term":{  
                   "active":true
                }
             },
             {  
                "or":[  
                   {  
                      "range":{  
                         "sunset_at":{  
                            "gt":"2016-06-24T07:10:08.658-07:00"
                         }
                      }
                   },
                   {  
                      "missing":{  
                         "field":"sunset_at"
                      }
                   }
                ]
             },
             {  
                "range":{  
                   "published_at":{  
                      "lte":"2016-06-24T07:10:08-07:00"
                   }
                }
             },
             {  
    
             },
             {  
                "has_parent":{  
                   "type":"video_catalog",
                   "query":{  
                      "filtered":{  
                         "query":{  
                            "match_all":{  
    
                            }
                         },
                         "filter":{  
                            "or":[  
                               {  
                                  "term":{  
                                     "whitelist_size":0
                                  }
                               },
                               {  
                                  "terms":{  
                                     "whitelist":[  
                                        948
                                     ]
                                  }
                               }
                            ]
                         }
                      }
                   }
                }
             },
             {  
                "terms":{  
                   "network_id":[  
                      215
                   ]
                }
             },
             {  
    
             },
             {  
    
             }
          ]
       },
       "sort":[  
          {  
             "published_at":"desc"
          }
       ]
    }`

Did this change recently (track_scores now defaults to false?), or is there some cluster level configuration that we can change?

Thanks for your help!
Aaron.

Hi @arustad,

I moved this post to the elasticsearch category as it seems that you should be able to get a better help there. From the cloud perspective we expose the elasticsearch api as is, without modifying it.

Thanks,
Igor

Thanks @igor_k, I was hoping that since I included a cluster ID that it might get some priority treatment from the Elasticsearch cloud support team.