_id field not returned consistently via Fields API #94515
Description
Elasticsearch Version
8.6.2
Installed Plugins
No response
Java Version
Oracle Corporation/Java HotSpot(TM) 64-Bit Server VM/18.0.2/18.0.2+9-61
OS Version
Mac OS X/12.6.3/aarch64
Problem Description
The fields option lets you query the
metadata _id field, but this field doesn't seem to be returned consistently over larger search result sets.
Steps to Reproduce
add several small documents to a test index with e.g.
POST /test/_doc { "foo" : "bar" } then query e.g. top 10 docs disabling _source and querying _id via fields:
GET test/_search { "_source": false, "fields": [ "_id" ], "size": 10, "query": { "match_all": {} } } In my current reproduction on 8.6.2 this returns e.g.
"hits": [ { "_index": "test", "_id": "082y5YYBJUrNWn3IOYNS", "_score": 1, "fields": { "_id": [ "082y5YYBJUrNWn3IOYNS" ] } }, { "_index": "test", "_id": "1M2y5YYBJUrNWn3IQIOO", "_score": 1, "fields": { "_id": [ "1M2y5YYBJUrNWn3IQIOO" ] } }, { "_index": "test", "_id": "1c2y5YYBJUrNWn3IRYNb", "_score": 1, "fields": { "_id": [ "1c2y5YYBJUrNWn3IRYNb" ] } }, { "_index": "test", "_id": "1s2y5YYBJUrNWn3ITIMM", "_score": 1, "fields": { "_id": [ "1s2y5YYBJUrNWn3ITIMM" ] } }, { "_index": "test", "_id": "182y5YYBJUrNWn3IUIPe", "_score": 1, "fields": { "_id": [ "182y5YYBJUrNWn3IUIPe" ] } }, { "_index": "test", "_id": "2M2z5YYBJUrNWn3IN4O4", "_score": 1, "fields": { "_id": [ "2M2z5YYBJUrNWn3IN4O4" ] } }, { "_index": "test", "_id": "2c2z5YYBJUrNWn3IUoNm", "_score": 1 }, { "_index": "test", "_id": "2s2z5YYBJUrNWn3IVYOr", "_score": 1 }, { "_index": "test", "_id": "282z5YYBJUrNWn3IWYNq", "_score": 1 }, { "_index": "test", "_id": "3M2z5YYBJUrNWn3IW4P9", "_score": 1, "fields": { "_id": [ "3M2z5YYBJUrNWn3IW4P9" ] } } Notice hits 7-9 without the _id field. If queried directly, these documents show the field though:
GET test/_search { "_source": false, "fields": [ "_id" ], "size": 10, "query": { "match": { "_id" : "282z5YYBJUrNWn3IWYNq" } } } ---> "hits": [ { "_index": "test", "_id": "282z5YYBJUrNWn3IWYNq", "_score": 1, "fields": { "_id": [ "282z5YYBJUrNWn3IWYNq" ] } } ] Strangely, also changing the "size" seems to change which docs have the _id in the fields response and which don't.
Logs (if relevant)
No response