assignment for deployment with id not found - 如何解决此 Elasticsearch 异常

ResourceNotFoundException: assignment for deployment with id [your-deployment-id] not found
  at org.elasticsearch.xpack.ml.action.TransportStopTrainedModelDeploymentAction.removeAssignment(TransportStopTrainedModelDeploymentAction.java)
  at org.elasticsearch.xpack.ml.MlMetadata$TrainedModelAssignmentMetadata.removeAssignment(TrainedModelAssignmentMetadata.java)

{
  "error": {
    "root_cause": [
      {
        "type": "resource_not_found_exception",
        "reason": "assignment for deployment with id [model-deployment-123] not found"
      }
    ],
    "type": "resource_not_found_exception",
    "reason": "assignment for deployment with id [model-deployment-123] not found"
  },
  "status": 404
}

# 第一次停止部署（成功）
POST /_ml/trained_models/elastic__distilbert-base-uncased/model_deployment/_stop

# 再次执行相同请求（此时会抛出异常）
POST /_ml/trained_models/elastic__distilbert-base-uncased/model_deployment/_stop
# 返回：assignment for deployment with id [xxx] not found

# 查看所有已部署的模型
GET /_ml/trained_models/_stats

# 查看指定模型的部署状态
GET /_ml/trained_models/elastic__distilbert-base-uncased/_stats

{
  "count": 1,
  "trained_model_stats": [
    {
      "model_id": "elastic__distilbert-base-uncased",
      "deployment_stats": {
        "deployment_id": "elastic__distilbert-base-uncased",
        "deployment_status": "started",
        "number_of_allocations": 1
      }
    }
  ]
}

# 查看当前所有模型分配情况
GET /_cluster/state?filter_path=metadata.ml.trained_model_assignment

# 返回示例
{
  "metadata": {
    "ml": {
      "trained_model_assignment": {
        "elastic__distilbert-base-uncased": {
          "task_parameters": { ... },
          "node_routing_table": { ... }
        }
      }
    }
  }
}

# 先尝试停止（忽略 404 错误）
POST /_ml/trained_models/elastic__distilbert-base-uncased/model_deployment/_stop

# 重新部署模型
POST /_ml/trained_models/elastic__distilbert-base-uncased/model_deployment/_start?wait_for=started
{
  "number_of_allocations": 1,
  "threads_per_allocation": 1
}

# 删除模型（谨慎操作，会同时删除模型定义）
DELETE /_ml/trained_models/elastic__distilbert-base-uncased?force=true

# 重新导入模型（以 ELSER 模型为例）
PUT /_ml/trained_models/.elser_model_1
{
  "input": {
    "field_names": ["text_field"]
  }
}

# 检查集群健康状态
GET /_cluster/health

# 检查主节点状态
GET /_cat/master?v

# 查看 ML 节点是否可用
GET /_cat/nodes?v&h=name,node.role,ml.machine_memory,ml.max_open_jobs

from elasticsearch import Elasticsearch
from elasticsearch.exceptions import NotFoundError

es = Elasticsearch("https://localhost:9200")

def stop_model_deployment(model_id):
    try:
        response = es.ml.stop_trained_model_deployment(
            model_id=model_id,
            wait_for="stopped"
        )
        return response
    except NotFoundError as e:
        # 模型部署不存在，视为已停止
        print(f"Model deployment {model_id} already stopped or not found.")
        return {"acknowledged": True, "status": "already_stopped"}

# 先查询状态
DEPLOYMENT_STATUS=$(curl -s -X GET "localhost:9200/_ml/trained_models/${MODEL_ID}/_stats" | jq -r '.trained_model_stats[0].deployment_stats.deployment_status // "stopped"')

if [ "$DEPLOYMENT_STATUS" = "started" ]; then
  curl -X POST "localhost:9200/_ml/trained_models/${MODEL_ID}/model_deployment/_stop?wait_for=stopped"
else
  echo "Model is not deployed, skipping stop operation."
fi

# 创建包含 ML 元数据的快照
PUT /_snapshot/my_backup_repo/snapshot_with_ml?wait_for_completion=true
{
  "indices": ".ml-*",
  "feature_states": ["ml"],
  "metadata": {
    "description": "Backup ML metadata before model deployment operations"
  }
}

static ClusterState removeAssignment(ClusterState currentState, String deploymentId) {
    TrainedModelAssignmentMetadata.Builder builder = TrainedModelAssignmentMetadata.builder(currentState);
    if (builder.hasModelDeployment(deploymentId) == false) {
        throw new ResourceNotFoundException("assignment for deployment with id [{}] not found", deploymentId);
    }
    logger.debug(() -> format("[%s] removing assignment", deploymentId));
    return update(currentState, builder.removeAssignment(deploymentId));
}