Scheduling configuration lets workspace administrators set fine-grained rules for resource usage, including usage limits, priorities, auto-shutdown policies, and cost controls. Use it when multiple users share a workspace or when you need to control resource usage.
Non-workspace members are users who are not added to the workspace by an administrator but are granted RAM permissions by the Alibaba Cloud account, so they can still use resources and submit jobs. Scheduling configuration lets you set specific constraints for these users.
Resource usage constraints
Resource usage constraints define how a resource quota can be used: which modules and roles can use it, and how much quota each can consume. This configuration is optional. If you don't configure resource usage constraints, workspace members can freely use all product modules.
Accessible modules: Restrict the modules (DSW, DLC, and EAS) that can use the resource quota.
Accessible roles: Restrict the resource quota to specific roles.
Usage limits: The amount of resources that each member in a role can use individually. Administrators can limit the resources available to different members for submitting jobs of specified priorities in a specific resource quota. If multiple rules match, the strictest rule applies. For example, in a resource quota, two rules apply to user A: one for administrators (priorities 1–7, up to 100 GPUs) and one for member A (priorities 7–9, up to 10 GPUs). When user A submits a priority 7 job in this quota, they can use at most 10 GPUs.
Total role quota: The amount of resources that all members of a role can use. Administrators can limit the resources available to different roles for submitting jobs of specified priorities in a specific resource quota. If multiple rules match, the strictest rule applies.
Resource spec template: If resource spec templates are configured for a resource quota, users must select a template when they create DSW instances or DLC jobs in the quota.
DLC
Maximum run duration: Set the maximum time a job can run. Jobs that exceed this limit are terminated automatically.
Job priority: Limit the highest priority of jobs that each role or member can submit.
Allow public resource usage: If this option is turned off, users can't create on-demand DLC jobs in the current workspace. This prevents unexpected charges.
-
Maximum wait duration: Set timeout thresholds for job wait time, queue time, and environment preparation time within a resource quota. A notification is sent when any threshold is exceeded.
NoteTo receive timeout alerts, also configure notification rules for DLC jobs on the Event Notification Configuration page. Otherwise, this setting alone doesn't trigger any alerts.
DSW
-
Auto-shutdown policy:
Shutdown policy: Automatically shut down instances based on their runtime or resource utilization (such as idle duration, CPU utilization, memory utilization, or GPU utilization) to reduce costs.
Exclusion policy: Exclude DSW instances from automatic shutdown by instance priority or instance name.
-
Environment save policy: Instances with cloud disks as system disks automatically save their environments by default, so no configuration is required. For instances that don't use cloud disks, the administrator can enable Save environment image before auto-shutdown as needed.
Large environment images may take a long time to save or fail to save, which reduces the shutdown success rate and slows down resource release.
Enable auto-migration for abnormal node instances: When an underlying Lingjun node fails, the system automatically shuts down and restarts the instance so that the node can self-heal. The environment image is saved during the restart, but running processes can't be restored.
Instance priority: Limit the highest priority of instances that each role or member can create.
Allow public resource usage: If this option is turned off, users can't create on-demand DSW instances in the current workspace. This prevents unexpected charges.
Maximum wait duration: Set timeout thresholds for resource usage. The default threshold is 30 minutes. If the queue time or environment preparation time exceeds the threshold, the instance fails and stops automatically.
EAS
Service priority: Configure resource usage priorities by workspace role or member. Valid values range from 1 to 9, where 1 is the lowest priority and 9 is the highest.
Allow public resource usage: If this option is turned off, users can't create on-demand EAS services in the current workspace. This prevents unexpected charges.