When you register an E-MapReduce (EMR) cluster to DataWorks, you select one of two registration modes:
When to Choose Each Mode
-
Choose
shortcut mode if your workspace does not require strict permission control between users and your priority is getting EMR tasks running quickly. -
Choose
security mode if different users or task owners must have isolated data permissions when running EMR tasks in DataWorks.
Mode Comparison
|
Dimension |
Shortcut mode |
Security mode |
|
Primary use case |
Rapid EMR task creation and execution |
EMR task execution with strict per-user data permission management |
|
Job execution identity |
All jobs run as a |
Jobs run as the cluster account mapped to the submitting Alibaba Cloud account or RAM user |
|
Permission granularity |
Fine-grained permission management is not supported. The hadoop user holds all permissions on the cluster. |
You configure an account mapping that grants the cluster account's permissions to the DataWorks identity. |
|
Permission isolation |
Not enforced — all task owners share the same hadoop identity |
Enforced — different Alibaba Cloud accounts, task owners, and RAM users have distinct data permissions |
|
Suitable for |
Workspaces that do not require strict permission management for users who run tasks |
Workspaces that require strict management and isolation of data permissions for users who run tasks |
Shortcut Mode
When a DataWorks user (Alibaba Cloud account or RAM user) runs or schedules an EMR task to generate data, the task code is committed to the EMR cluster and run by a
Limitations
-
Fine-grained permission management is not supported.
-
All jobs share the same Hadoop user identity. Data access controls are not enforced at the user level.
Because all task code runs under a single shared identity with full cluster permissions, proceed with caution when using
Security Mode
When a DataWorks user (Alibaba Cloud account or RAM user) runs or schedules an EMR task, the task code is committed to the EMR cluster and executed as the cluster account to which that DataWorks identity is mapped.
You configure a
|
Cluster type |
Supported Mapping Type values |
|
|
|
|
|
|
Permission management
-
You configure a mapping between each Alibaba Cloud account or RAM user and a cluster account. This mapping grants the cluster account's permissions to that DataWorks identity and isolates permissions on task data.
-
EMR Ranger can be used to manage the permissions of each Hadoop user in the EMR cluster. -
Different Alibaba Cloud accounts, task owners, and RAM users have different data permissions when running EMR tasks in DataWorks.