Checks whether the maximum GPU utilization of an ECS instance meets a specified threshold within a given time range. Non-GPU instances and unmonitored instances are marked as not applicable.
Risk level
Default risk level: Low.
You can adjust the risk level.
Detection logic
-
Compliant: the maximum GPU utilization within the specified time range meets or exceeds the threshold.
-
Not applicable: the instance is not monitored by CloudMonitor or has no monitoring data.
-
Not applicable: non-GPU ECS instances.
Default time range: 7 days. Detection uses CloudMonitor data APIs and consumes the basic CloudMonitor free quota. Enable Hybrid Cloud Monitoring for reliable detection. For Hybrid Cloud Monitoring billing, see CloudMonitor billing details.
Rule details
|
Parameter |
Description |
|
Rule name |
Idle GPU detection for ECS instances |
|
Rule identifier |
|
|
Tag |
[ECS] |
|
Automatic remediation |
Not supported |
|
Rule trigger |
24-hour cycle |
|
Supported resource types |
[ACS::ECS::Instance] |
|
Input parameters |
relativeTime (default: 168) |
Remediation guide
Follow the remediation steps in Elastic GPU Service (gn/vgn/sgn series).