[BUG] MissingRequiredPropertyException: Missing required property 'ResourceStats.average' when polling _tasks API

[LanceGrahamTR]

What is the bug?

The ResourceStats schema in spec/schemas/_common.yaml of the opensearch-api-specification marks average, max, min, thread_info, and total as all required. However, the _tasks API on OpenSearch 1.3 only returns total in the resource_stats block when a task is in-flight.

Per the compatibility matrix, opensearch-java 3.x supports OpenSearch 1.x-3.x, so the spec should reflect what 1.x actually returns.

This causes the opensearch-java 3.8.0 client to throw:

org.opensearch.client.util.MissingRequiredPropertyException: Missing required property 'ResourceStats.average'
    at org.opensearch.client.util.ApiTypeHelper.requireNonNull(ApiTypeHelper.java:90)
    at org.opensearch.client.opensearch._types.ResourceStats.<init>(ResourceStats.java:82)
    at org.opensearch.client.opensearch._types.ResourceStats$Builder.build(ResourceStats.java:302)
    at org.opensearch.client.json.ObjectBuilderDeserializer.deserialize(ObjectBuilderDeserializer.java:92)

How can one reproduce the bug?

1. Index test documents (100k+):

POST /test-index/_bulk
{"index":{"_id":"1"}}
{"date":"2020-01-01T00:00:00Z","status":"OPEN"}
{"index":{"_id":"2"}}
{"date":"2020-01-01T00:00:00Z","status":"OPEN"}
... (repeat to 100k+ documents)

2. Submit async delete-by-query:

POST /test-index/_delete_by_query?wait_for_completion=false&conflicts=proceed
{
  "query": {
    "range": {
      "date": {
        "lt": "now-6M/d"
      }
    }
  }
}

Response:

{
  "task": "nodeId:1234567"
}

3. Poll the task while it's still running:

GET /_tasks/nodeId:1234567

In-flight response (before task completes):

{
  "completed": false,
  "task": {
    "node": "nodeId",
    "id": 1234567,
    "type": "transport",
    "action": "indices:data/write/delete/byquery",
    "status": {
      "total": 100000,
      "updated": 0,
      "created": 0,
      "deleted": 11000,
      "batches": 12,
      "version_conflicts": 0,
      "noops": 0,
      "retries": {
        "bulk": 0,
        "search": 0
      },
      "throttled_millis": 0,
      "requests_per_second": -1.0,
      "throttled_until_millis": 0
    },
    "description": "delete-by-query [test-index]",
    "start_time_in_millis": 1781528541973,
    "running_time_in_nanos": 587603829,
    "cancellable": true,
    "cancelled": false,
    "headers": {},
    "resource_stats": {
      "total": {
        "cpu_time_in_nanos": 0,
        "memory_in_bytes": 0
      }
    }
  }
}

Note: resource_stats contains only total. The fields average, max, min, and thread_info are absent. The opensearch-java 3.8.0 client fails to deserialize this response because ResourceStats marks all five fields as required.

4. After task completes, resource_stats is no longer present in the response, which is why this only surfaces when polling in-flight tasks with large enough data volumes for the task to take seconds.

Environment

• opensearch-java client version: 3.8.0
• OpenSearch server: 1.3
• Java version: 17