_mvt endpoint returns 400 when results are incomplete (shard errors) instead of partial results like _search endpoint

[nreese]

Elasticsearch Version

main

Installed Plugins

none

Java Version

bundled

OS Version

21.6.0 Darwin Kernel Version 21.6.0

Problem Description

_mvt endpoint returns 400 when results are incomplete (shard errors) instead of partial results like _search endpoint

Steps to Reproduce

Run commands to set up data

PUT local1
{}

PUT local1/_mapping
{
  "properties": {
    "value": {
      "type": "keyword"
    },
    "location": {
      "type": "geo_point"
    }
  }
}

PUT local1/_doc/1
{
    "value" : "foo1",
    "location": [0, 0]
}

PUT local2
{}

PUT local2/_mapping
{
  "properties": {
    "value": {
      "type": "keyword"
    },
    "location": {
      "type": "geo_point"
    }
  }
}

PUT local2/_doc/1
{
    "value" : "foo2",
    "location": [1, 1]
}

Run _mvt request

POST /local*/_mvt/location/1/1/1
{
  "buffer": 0,
  "grid_precision": 0,
  "exact_bounds": true,
  "extent": 4096,
  "query": {
    "bool": {
      "filter": [
        {
          "error_query": {
            "indices": [
              {
                "error_type": "exception",
                "message": "local shard failure message 123",
                "name": "local2"
              }
            ]
          }
        }
      ],
      "must": [],
      "must_not": [],
      "should": []
    }
  }
}

Notice how the response returns 400.

{
  "error": {
    "root_cause": [
      {
        "type": "illegal_argument_exception",
        "reason": "Unsupported vector tile type for field [_shards.failures.0] : java.util.LinkedHashMap"
      }
    ],
    "type": "illegal_argument_exception",
    "reason": "Unsupported vector tile type for field [_shards.failures.0] : java.util.LinkedHashMap"
  },
  "status": 400
}

The same request to _search endpoint returns partial results. _mvt should mimic this behavior.

POST /local*/_search
{
  "query": {
    "bool": {
      "filter": [
        {
          "error_query": {
            "indices": [
              {
                "error_type": "exception",
                "message": "local shard failure message 123",
                "name": "local2"
              }
            ]
          }
        }
      ],
      "must": [],
      "must_not": [],
      "should": []
    }
  }
}

{
  "took": 0,
  "timed_out": false,
  "_shards": {
    "total": 2,
    "successful": 1,
    "skipped": 0,
    "failed": 1,
    "failures": [
      {
        "shard": 0,
        "index": "local2",
        "node": "XPQtz28bSLW26Oynf6oUmg",
        "reason": {
          "type": "query_shard_exception",
          "reason": "failed to create query: [local2][0] local shard failure message 123",
          "index_uuid": "sg-JHdyNRUqszXjYiQTBNQ",
          "index": "local2",
          "caused_by": {
            "type": "runtime_exception",
            "reason": "[local2][0] local shard failure message 123"
          }
        }
      }
    ]
  },
  "hits": {
    "total": {
      "value": 1,
      "relation": "eq"
    },
    "max_score": 0,
    "hits": [
      {
        "_index": "local1",
        "_id": "1",
        "_score": 0,
        "_source": {
          "value": "foo1",
          "location": [
            0,
            0
          ]
        }
      }
    ]
  }
}

Logs (if relevant)

No response

[elasticsearchmachine]

Pinging @elastic/es-analytics-geo (Team:Analytics)

[iverase]

This a side effect of #97619. Before we were just ignoring that part of the result, now we are just reporting we cannot handle it. Maybe we need to go back to the old behaviour.

[nreese]

This a side effect of #97619. Before we were just ignoring that part of the result, now we are just reporting we cannot handle it. Maybe we need to go back to the old behaviour.

Just checked 8.8 and response from _mvt endpoint is

{
	"meta": [
		{
			"geometry": {
				"type": "Point",
				"coordinates": [
					0,
					0
				]
			},
			"properties": {
				"_shards.failed": 1,
				"_shards.failures.0.index": "local2",
				"_shards.failures.0.node": "-ydL7NYTR-qgITqZrOHRdQ",
				"_shards.failures.0.reason.index": "local2",
				"_shards.failures.0.reason.index_uuid": "aGlYyc6OQ7iTUIp2yJ7uxw",
				"_shards.failures.0.reason.reason": "Field [location] is of unsupported type [keyword] for [geo_shape] query",
				"_shards.failures.0.reason.type": "query_shard_exception",
				"_shards.failures.0.shard": 0,
				"_shards.skipped": 0,
				"_shards.successful": 1,
				"_shards.total": 2,
				"hits.total.relation": "eq",
				"hits.total.value": 1,
				"timed_out": false,
				"took": 56
			}
		}
	],
	"hits": [
		{
			"geometry": {
				"type": "Point",
				"coordinates": [
					0,
					0
				]
			},
			"properties": {
				"_id": "1",
				"_index": "local1"
			}
		}
	]
}

Reverting to this response would be great.