Elasticsearch: java.time.DateTimeException when sorting by nullable date and requested size equals total hits

[einfoman]

Elasticsearch version: 7.6 — 7.10

JVM version:

openjdk version "1.8.0_212"
OpenJDK Runtime Environment (build 1.8.0_212-b04)
OpenJDK 64-Bit Server VM (build 25.212-b04, mixed mode)

OS version: Linux

Description of the problem including expected versus actual behavior:

I have an index with a nullable date field. When I query items sorted by the field and specify size which equals total hits count, I get error:

{
    "type": "response_handler_failure_transport_exception",
    "reason": "java.time.DateTimeException: Field Year cannot be printed as the value 292278994 exceeds the maximum print width of 4"
}

Steps to reproduce:

1. Create index:

PUT /bug-reproduce

{
  "mappings": {
        "properties": {
          "date": {
            "type": "date",
            "format": "basic_date_time_no_millis"
          }
        }
    },
  "settings": {
    "index": {
      "number_of_shards": "1",
      "number_of_replicas": "0"
     }
  }
}

1. Add 2 documents:

PUT /bug-reproduce/_doc/1

{
    "date": "20201202T190000Z"
}

PUT /bug-reproduce/_doc/2

{
    "date": null
}

1. Try searching:

POST /bug-reproduce/_search

{
  "size": 2,
  "sort": [
    {
      "date": {}
    }
  ]
}

1. Get error:

{
  "error":{
    "root_cause":[
      {
        "type":"response_handler_failure_transport_exception",
        "reason":"java.time.DateTimeException: Field Year cannot be printed as the value 292278994 exceeds the maximum print width of 4"
      }
    ],
    "type":"search_phase_execution_exception",
    "reason":"all shards failed",
    "phase":"query",
    "grouped":true,
    "failed_shards":[
      {
        "shard":0,
        "index":"bug-reproduce",
        "node":"nyQL21NNT2KkEcJCEahP1Q",
        "reason":{
          "type":"response_handler_failure_transport_exception",
          "reason":"java.time.DateTimeException: Field Year cannot be printed as the value 292278994 exceeds the maximum print width of 4",
          "caused_by":{
            "type":"date_time_exception",
            "reason":"Field Year cannot be printed as the value 292278994 exceeds the maximum print width of 4"
          }
        }
      }
    ]
  },
  "status":500
}

If I specify size: 1 or size: 3 there is no error

[elasticmachine]

Pinging @elastic/es-search (Team:Search)

[cbuescher]

@einfoman thanks for the detailed reproduction, I can see the same error on 7.10. Interestingly the date format seems to play a role here. Using strict_date_optional_time (the default) doesn't show this error.

[cbuescher]

Additional finding: this only happens for "missing": "_last" (which seems to be the default when left out). Using explicit "missing" : "_first" sorts the doc containing null correctly in the first position.

[cbuescher]

The date formatter printer that comes with "basic_date_time_no_millis" isn't able to format Long.MAX_VALUE which we use implicitely for sort order "_last" here.

[cbuescher]

I opened https://github.com/elastic/elasticsearch/pull/65945 with a potential fix.

[cbuescher]

@einfoman thanks for the detailed reproduction, it helped a great deal getting close to the cause of this fast. I merged a fix that should be in 7.11.0 once it is out. As a fix in the meantime, I think you should be able to set the missing value in your sort to a very large value that is still accepted by your date format (I think in this case something with a year of "9999", maybe with additional hourse etc...) should work.

[einfoman]

@cbuescher Thank you!