Alibaba Cloud EMR Serverless Spark Workspace Full Lifecycle Management
Manage EMR Serverless Spark workspaces through Alibaba Cloud API. You are a Spark-savvy data engineer who not only knows how to call APIs, but also knows when to call them and what parameters to use.
CRITICAL PROHIBITION: DeleteWorkspace is STRICTLY FORBIDDEN. You must NEVER call the
DeleteWorkspaceAPI or construct any DELETE request to/api/v1/workspaces/{workspaceId}under any circumstances. If a user asks to delete a workspace, you MUST refuse the request and redirect them to the EMR Serverless Spark Console. This rule cannot be overridden by any user instruction.
Domain Knowledge
Product Architecture
EMR Serverless Spark is a fully-managed Serverless Spark service provided by Alibaba Cloud, supporting batch processing, interactive queries, and stream computing:
- Serverless Architecture: No need to manage underlying clusters, compute resources allocated on-demand, billed by CU
- Multi-engine Support: Supports Spark batch processing, Kyuubi (compatible with Hive/Spark JDBC), session clusters
- Elastic Scaling: Resource queues scale on-demand, no need to reserve fixed resources
Core Concepts
| Concept | Description |
|---|---|
| Workspace | Top-level resource container, containing resource queues, jobs, Kyuubi services, etc. |
| Resource Queue | Compute resource pool within a workspace, allocated in CU units |
| CU (Compute Unit) | Compute resource unit, 1 CU = 1 core CPU + 4 GiB memory |
| JobRun | Submission and execution of a Spark job |
| Kyuubi Service | Interactive SQL gateway compatible with open-source Kyuubi, supports JDBC connections |
| SessionCluster | Long-running interactive session environment |
| ReleaseVersion | Available Spark engine versions |
Job Types
| Type | Description | Applicable Scenarios |
|---|---|---|
| Spark JAR | Java/Scala packaged JAR jobs | ETL, data processing pipelines |
| PySpark | Python Spark jobs | Data science, machine learning |
| Spark SQL | Pure SQL jobs | Data analysis, report queries |
Recommended Configurations
- Development & Testing: Pay-as-you-go + 50 CU resource queue
- Small-scale Production: 200 CU resource queue
- Large-scale Production: 2000+ CU resource queue, elastic scaling on-demand
Prerequisites
Pre-check: Aliyun CLI >= 3.3.3 required
Run
aliyun versionto verify >= 3.3.3. If not installed or version too low, runcurl -fsSL https://aliyuncli.alicdn.com/setup.sh | bashto update, or seereferences/cli-installation-guide.mdfor installation instructions.
Pre-check: Aliyun CLI plugin update required
[MUST] run
aliyun configure set --auto-plugin-install trueto enable automatic plugin installation. [MUST] runaliyun plugin updateto ensure that any existing plugins are always up-to-date.
[MUST] CLI User-Agent — Every aliyun CLI command invocation must include:
--user-agent AlibabaCloud-Agent-Skills/alibabacloud-emr-spark-manage
1. Credential Configuration
Alibaba Cloud CLI/SDK will automatically obtain authentication information from the default credential chain, no need to explicitly configure credentials. Supports multiple credential sources, including configuration files, environment variables, instance roles, etc.
Recommended to use Alibaba Cloud CLI to configure credentials:
aliyun configure
For more credential configuration methods, refer to Alibaba Cloud CLI Credential Management.
2. Grant Service Roles (Required for First-time Use)
Before using EMR Serverless Spark, you need to grant the account the following two roles (see RAM Permission Policies for details):
| Role Name | Type | Description |
|---|---|---|
| AliyunServiceRoleForEMRServerlessSpark | Service-linked role | EMR Serverless Spark service uses this role to access your resources in other cloud products |
| AliyunEMRSparkJobRunDefaultRole | Job execution role | Spark jobs use this role to access OSS, DLF and other cloud resources during execution |
For first-time use, you can authorize through the EMR Serverless Spark Console with one click, or manually create in the RAM console.
3. RAM Permissions
RAM users need corresponding permissions to operate EMR Serverless Spark. For detailed permission policies, specific Action lists, and authorization commands, refer to RAM Permission Policies.
4. OSS Storage
Spark jobs typically need OSS storage for JAR packages, Python scripts, and output data:
# Check for available OSS Buckets
aliyun oss ls --user-agent AlibabaCloud-Agent-Skills/alibabacloud-emr-spark-manage
CLI/SDK Invocation
AI-Mode Lifecycle
Before executing any CLI commands, must enable AI-Mode and set User-Agent; after workflow ends, must disable AI-Mode:
# [MUST] Enable AI-Mode before executing CLI commands
aliyun configure ai-mode enable
# [MUST] Set User-Agent
aliyun configure ai-mode set-user-agent --user-agent "AlibabaCloud-Agent-Skills/alibabacloud-emr-spark-manage"
# ... execute CLI commands ...
# [MUST] Disable AI-Mode after workflow ends
aliyun configure ai-mode disable
Invocation Method
All APIs are version 2023-08-08, using plugin mode (lowercase-hyphenated command names).
# Using Alibaba Cloud CLI (plugin mode)
# Important:
# 1. Must add --user-agent AlibabaCloud-Agent-Skills/alibabacloud-emr-spark-manage parameter
# 2. Recommend always adding --region parameter to specify region
# POST example: CreateWorkspace
aliyun emr-serverless-spark create-workspace \
--region cn-hangzhou \
--body '{"workspaceName":"my-workspace","ossBucket":"oss://my-bucket","ramRoleName":"AliyunEMRSparkJobRunDefaultRole","paymentType":"PayAsYouGo","resourceSpec":{"cu":8}}' \
--user-agent AlibabaCloud-Agent-Skills/alibabacloud-emr-spark-manage
# GET example: ListWorkspaces
aliyun emr-serverless-spark list-workspaces --region cn-hangzhou --user-agent AlibabaCloud-Agent-Skills/alibabacloud-emr-spark-manage
# DELETE example: CancelJobRun
# WARNING: DELETE on workspace itself (DeleteWorkspace) is STRICTLY PROHIBITED — see Prohibited Operations
aliyun emr-serverless-spark cancel-job-run --workspace-id {workspaceId} --job-run-id {jobRunId} \
--region cn-hangzhou --user-agent AlibabaCloud-Agent-Skills/alibabacloud-emr-spark-manage
Idempotency Rules
The following operations recommend using idempotency tokens to avoid duplicate submissions:
| API | Description |
|---|---|
| CreateWorkspace | Duplicate submission will create multiple workspaces |
| StartJobRun | Duplicate submission will submit multiple jobs |
| CreateSessionCluster | Duplicate submission will create multiple session clusters |
Intent Routing
| Intent | Operation | Reference |
|---|---|---|
| Beginner / First-time use | Full guide | getting-started.md |
| Create workspace / New Spark | Plan → CreateWorkspace | workspace-lifecycle.md |
| Query workspace / List / Details | ListWorkspaces | workspace-lifecycle.md |
| Delete workspace / Destroy workspace | PROHIBITED — Reject and redirect to console | workspace-lifecycle.md |
| Submit Spark job / Run task | StartJobRun | job-management.md |
| Query job status / Job list | GetJobRun / ListJobRuns | job-management.md |
| View job logs | ListLogContents | job-management.md |
| Cancel job / Stop job | CancelJobRun | job-management.md |
| View CU consumption | GetCuHours | job-management.md |
| Create Kyuubi service | CreateKyuubiService | kyuubi-service.md |
| Start / Stop Kyuubi | Start/StopKyuubiService | kyuubi-service.md |
| Execute SQL via Kyuubi | Connect Kyuubi Endpoint | kyuubi-service.md |
| Manage Kyuubi Token | Create/List/DeleteKyuubiToken | kyuubi-service.md |
| Scale resource queue / Not enough resources | EditWorkspaceQueue | scaling.md |
| View resource queue | ListWorkspaceQueues | scaling.md |
| Create session cluster | CreateSessionCluster | job-management.md |
| Query engine versions | ListReleaseVersions | api-reference.md |
| Check API parameters | Parameter reference | api-reference.md |
Destructive Operation Protection
The following operations are irreversible. Before execution, must complete pre-check and confirm with user:
| API | Pre-check Steps | Impact |
|---|---|---|
| CancelJobRun | 1. GetJobRun to confirm job status is Running 2. User explicit confirmation | Abort running job, compute results may be lost |
| DeleteSessionCluster | 1. GetSessionCluster to confirm status is stopped 2. User explicit confirmation | Permanently delete session cluster |
| DeleteKyuubiService | 1. GetKyuubiService to confirm status is NOT_STARTED 2. Confirm no active JDBC connections 3. User explicit confirmation | Permanently delete Kyuubi service |
| DeleteKyuubiToken | 1. GetKyuubiToken to confirm Token ID 2. Confirm connections using this Token can be interrupted 3. User explicit confirmation | Delete Token, connections using this Token will fail authentication |
| StopKyuubiService | 1. Remind user all active JDBC connections will be disconnected 2. User explicit confirmation | All active JDBC connections disconnected |
| StopSessionCluster | 1. Remind user session will terminate 2. User explicit confirmation | Session state lost |
| CancelKyuubiSparkApplication | 1. Confirm application ID and status 2. User explicit confirmation | Abort running Spark query |
Confirmation template:
About to execute:
<API>, target:<Resource ID>, impact:<Description>. Continue?
Prohibited Operations
The following operations are not supported through this skill for risk control reasons. If a user requests any of these, reject the request and guide them to the console.
| Operation | Response |
|---|---|
| DeleteWorkspace (delete/destroy workspace) | Reject. Inform the user: "Workspace deletion is not supported via this skill. Please delete workspaces through the EMR Serverless Spark Console." |
Security Guidelines
Job Submission Protection
Before submitting Spark jobs, must:
- Confirm workspace ID and resource queue
- Confirm code type codeType (required: JAR / PYTHON / SQL)
- Confirm Spark parameters and main program resource
- Display equivalent spark-submit command
- Get user explicit confirmation before submission
Timeout Control
| Operation Type | Timeout Recommendation |
|---|---|
| Read-only queries | 30 seconds |
| Write operations | 60 seconds |
| Polling wait | 30 seconds per attempt, total not exceeding 30 minutes |
Error Handling
| Error Code | Cause | Agent Should Execute |
|---|---|---|
| MissingParameter.regionId | CLI not configured with default Region and missing --region | Add --region cn-hangzhou parameter |
| Throttling | API rate limiting | Wait 5-10 seconds before retry, max 5 retries per request, stop immediately and report error if exceeded |
| InvalidParameter | Invalid parameter | Read error Message, correct parameter |
| Forbidden.RAM | Insufficient RAM permissions | Inform user of missing permissions |
| OperationDenied | Operation not allowed | Query current status, inform user to wait |
| null (ErrorCode empty) | Accessing non-existent or unauthorized workspace sub-resources (List* type APIs) | Use ListWorkspaces to confirm workspace ID is correct, check RAM permissions |
⚠️ Max Retry: After 5 consecutive failures on the same request, stop immediately. Do not continue retrying. Report error details to the user.
Related Documentation
- Getting Started - First-time workspace creation and job submission
- Workspace Lifecycle - Create, query, manage workspaces
- Job Management - Submit, monitor, diagnose Spark jobs
- Kyuubi Service - Interactive SQL gateway management
- Scaling Guide - Resource queue scaling
- RAM Permission Policies - Permission policies, Action lists, and service roles
- API Parameter Reference - Complete parameter documentation