輪詢 BigQuery 工作完成情況

使用 BigQuery 連接器啟動 BigQuery 工作時，工作流程可能不會自動等待工作完成。如果後續步驟取決於作業是否成功，您應實作輪詢機制。

許多 Workflows 連接器會傳回可輪詢的長時間執行的作業 (LRO) 資源。雖然部分連接器會傳回 Operation 類型的資源來代表 LRO，但 BigQuery 會使用 Job 資源來達到這個目的。其他服務 (例如 Cloud Build 或 Compute Engine) 則會傳回實際的 Operation 資源。如需 LRO 資源類型清單，請參閱支援的連接器。

自動輪詢

使用會傳回 LRO 資源的連接器方法 (例如 googleapis.bigquery.v2.jobs.insert) 時，您可以自動輪詢 BigQuery 工作是否完成。

使用這類連接器方法時，Workflows 會封鎖工作流程執行作業，直到作業成功或失敗為止。您可以使用 connector_params 欄位自訂輪詢行為，包括總逾時和狀態檢查頻率。詳情請參閱「叫用連接器呼叫」。

舉例來說，下列工作流程示範如何使用 jobs.insert 方法啟動 BigQuery 工作，並使用自訂設定 (而非預設設定) 的政策，設定工作的輪詢作業：

YAML

# This workflow demonstrates how to automatically poll for the completion of
# a BigQuery job when using a connector method that return an LRO resource.
main:
  steps:
    - init:
        assign:
          - project_id: ${sys.get_env("GOOGLE_CLOUD_PROJECT_ID")}
          - query: "SELECT * FROM `bigquery-public-data.usa_names.usa_1910_2013` LIMIT 10"
    - run_bigquery_job:
        call: googleapis.bigquery.v2.jobs.insert
        args:
          projectId: ${project_id}
          body:
            configuration:
              query:
                query: ${query}
                useLegacySql: false
          connector_params:
            timeout: 3600  # Total time in seconds to wait for the job; default is 1800
            polling_policy:
              initial_delay: 2.0  # Seconds to wait before the first poll; default is 1.0
              multiplier: 1.5     # Factor by which to increase delay between polls; default is 1.25
              max_delay: 60.0     # Maximum delay in seconds between polls; default is 60.0
        result: job_status
    - finish:
        return: ${job_status}

JSON

{
  "main": {
    "steps": [
      {
        "init": {
          "assign": [
            {
              "project_id": "${sys.get_env(\"GOOGLE_CLOUD_PROJECT_ID\")}"
            },
            {
              "query": "SELECT * FROM `bigquery-public-data.usa_names.usa_1910_2013` LIMIT 10"
            }
          ]
        }
      },
      {
        "run_bigquery_job": {
          "call": "googleapis.bigquery.v2.jobs.insert",
          "args": {
            "projectId": "${project_id}",
            "body": {
              "configuration": {
                "query": {
                  "query": "${query}",
                  "useLegacySql": false
                }
              }
            },
            "connector_params": {
              "timeout": 3600,
              "polling_policy": {
                "initial_delay": 2,
                "multiplier": 1.5,
                "max_delay": 60
              }
            }
          },
          "result": "job_status"
        }
      },
      {
        "finish": {
          "return": "${job_status}"
        }
      }
    ]
  }
}

手動輪詢

自動輪詢僅適用於傳回 LRO 資源的方法。標準同步方法 (例如 googleapis.bigquery.v2.jobs.query) 不支援自動輪詢，如果查詢時間超過內部 API 逾時時間，可能需要手動輪詢。

舉例來說，下列工作流程示範如何使用 getQueryResults 方法，直接在工作流程中輪詢 BigQuery 工作是否完成：

YAML

# This workflow demonstrates how to manually poll for the completion of
# a BigQuery job by using the `getQueryResults` method.
main:
  steps:
    - init:
        assign:
          - project_id: ${sys.get_env("GOOGLE_CLOUD_PROJECT_ID")}
    - startQuery:
        call: googleapis.bigquery.v2.jobs.query
        args:
          projectId: ${project_id}
          body:
            useLegacySql: false
            query: "SELECT * FROM `bigquery-public-data.usa_names.usa_1910_2013` LIMIT 10"
        result: queryResponse
    - getJobId:
        assign:
          - jobId: ${queryResponse.jobReference.jobId}
    - pollingLoop:
        steps:
          - checkStatus:
              call: googleapis.bigquery.v2.jobs.getQueryResults
              args:
                projectId: ${project_id}
                jobId: ${jobId}
                timeoutMs: 10000  # Default wait time per call
              result: jobStatus
          - checkIfDone:
              switch:
                - condition: ${jobStatus.jobComplete}
                  return: ${jobStatus}  # Job is finished
          - wait:
              call: sys.sleep
              args:
                seconds: 5
              next: checkStatus  # Repeat check

JSON

{
  "main": {
    "steps": [
      {
        "init": {
          "assign": [
            {
              "project_id": "${sys.get_env(\"GOOGLE_CLOUD_PROJECT_ID\")}"
            }
          ]
        }
      },
      {
        "startQuery": {
          "call": "googleapis.bigquery.v2.jobs.query",
          "args": {
            "projectId": "${project_id}",
            "body": {
              "useLegacySql": false,
              "query": "SELECT * FROM `bigquery-public-data.usa_names.usa_1910_2013` LIMIT 10"
            }
          },
          "result": "queryResponse"
        }
      },
      {
        "getJobId": {
          "assign": [
            {
              "jobId": "${queryResponse.jobReference.jobId}"
            }
          ]
        }
      },
      {
        "pollingLoop": {
          "steps": [
            {
              "checkStatus": {
                "call": "googleapis.bigquery.v2.jobs.getQueryResults",
                "args": {
                  "projectId": "${project_id}",
                  "jobId": "${jobId}",
                  "timeoutMs": 10000
                },
                "result": "jobStatus"
              }
            },
            {
              "checkIfDone": {
                "switch": [
                  {
                    "condition": "${jobStatus.jobComplete}",
                    "return": "${jobStatus}"
                  }
                ]
              }
            },
            {
              "wait": {
                "call": "sys.sleep",
                "args": {
                  "seconds": 5
                },
                "next": "checkStatus"
              }
            }
          ]
        }
      }
    ]
  }
}

後續步驟