Vertex AI托管笔记本创建定时执行计划报错:无法提交计划
Vertex AI Workbench定时执行提交失败:Unknown Error排查方案
问题描述
在Vertex AI Workbench托管笔记本中,交互式执行和**单次执行(One-time execution)均正常,但创建定时重复执行(Schedule-based recurring executions)**时,提交等待约30秒后报错:2: Unknown Error.: Unable to submit schedule,执行参数与单次执行完全一致。
完整错误日志如下:
{ "protoPayload": { "@type": "type.googleapis.com/google.cloud.audit.AuditLog", "status": { "code": 2, "message": "Error 0: googleapi: Error 503: The service is currently unavailable. Details: [ { "@type": "type.googleapis.com/google.rpc.DebugInfo", "detail": "[ORIGINAL ERROR] RPC::DEADLINE_EXCEEDED: Deadline exceeded while waiting for channels to become available, and/or be able to accept RPCs. Channel count = 0. Queue size = 0. Deadline=28.992677276s; Extensible Stubs capped the outgoing deadline: see go/stubs-longest-allowed-deadline; \ncom.google.net.rpc3.RpcException: DEADLINE_EXCEEDED: Deadline exceeded while waiting for channels to become available, and/or be able to accept RPCs. Channel count = 0. Queue size = 0. Deadline=28.992677276s; Extensible Stubs capped the outgoing deadline: see go/stubs-longest-allowed-deadline; \n\tat com.google.apphosting.admin.mixer.actions.CreateApplicationAction.checkRpcPresentStatus(CreateApplicationAction.java:194)\n\tSuppressed: com.google.common.labs.concurrent.LabsFutures$LabeledExecutionException: GraphFuture{key=@com.google.apphosting.admin.mixer.actions.CreateApplicationAction$CreateReturnMessage com.google.protobuf.Message} failed: com.google.net.rpc3.RpcException: DEADLINE_EXCEEDED: Deadline exceeded while waiting for channels to become available, and/or be able to accept RPCs. Channel count = 0. Queue size = 0. Deadline=28.992677276s; Extensible Stubs capped the outgoing deadline: see go/stubs-longest-allowed-deadline; \n" } ] , backendError" }, "authenticationInfo": { "principalEmail": "xxxxxxxxxx@developer.gserviceaccount.com", "serviceAccountDelegationInfo": [ { "firstPartyPrincipal": { "principalEmail": "xxxxxxxxxx@gcp-sa-notebooks.iam.gserviceaccount.com" } } ] }, "requestMetadata": { "requestAttributes": {}, "destinationAttributes": {} }, "serviceName": "notebooks.googleapis.com", "methodName": "google.cloud.notebooks.v1.NotebookService.CreateSchedule", "resourceName": "projects/MY-PROJECT/locations/europe-west4/schedules/test_rnn_aria__1673534440196", "resourceLocation": { "currentLocations": [ "europe-west4" ] } }, "insertId": "184n01ud3oul", "resource": { "type": "audited_resource", "labels": { "method": "google.cloud.notebooks.v1.NotebookService.CreateSchedule", "service": "notebooks.googleapis.com", "project_id": "MY-PROJECT" } }, "timestamp": "2023-01-12T14:42:42.563449708Z", "severity": "ERROR", "logName": "projects/MY-PROJECT/logs/cloudaudit.googleapis.com%2Factivity", "operation": { "id": "projects/MY-PROJECT/locations/europe-west4/operations/operation-XXXXXXXXXX", "producer": "notebooks.googleapis.com", "last": true }, "receiveTimestamp": "2023-01-12T14:42:43.526355300Z" }
排查与解决步骤
从日志中的DEADLINE_EXCEEDED和503: The service is currently unavailable可以看出,问题核心是服务端超时或临时不可用,结合场景可按以下步骤排查:
- 检查目标区域服务健康状态:登录GCP控制台,查看Notebooks服务在
europe-west4区域的运行状态,确认是否存在临时故障或维护。 - 验证服务账号权限:
- 确认执行定时任务的服务账号(
gcp-sa-notebooks.iam.gserviceaccount.com)拥有notebooks.schedules.create权限,同时需具备Cloud Scheduler、Cloud Functions(若涉及)的相关权限,以及笔记本输出存储桶的读写权限。 - 在项目IAM页面检查该服务账号是否被授予
Notebooks Admin角色,或包含上述权限的自定义角色。
- 确认执行定时任务的服务账号(
- 尝试用gcloud命令行创建定时任务:界面操作的超时限制可能较严格,使用命令行工具可配置更灵活的参数,示例命令:
gcloud notebooks schedules create SCHEDULE_NAME \ --location=europe-west4 \ --vm-image-project=deeplearning-platform-release \ --vm-image-family=common-cpu \ --notebook-file=gs://PATH/TO/YOUR/NOTEBOOK.ipynb \ --schedule="0 1 * * *" - 检查VPC网络配置:
- 若笔记本部署在自定义VPC中,确保防火墙规则允许出站流量访问Notebooks API和Cloud Scheduler API。
- 若启用了私有服务访问,确认Notebooks API端点在VPC内可达。
- 重试或切换区域测试:
- 临时服务不可用通常是短暂的,间隔一段时间后重新提交定时任务。
- 若条件允许,切换到其他稳定区域(如
europe-west1)创建笔记本和定时任务,排除区域特定问题。
内容的提问来源于stack exchange,提问作者MariusPontmercy
相关产品推荐
相关产品推荐

