Metrics for system monitoring event-triggered tasks are measured at the scaling group level. A metric's value is the average value from all instances in the scaling group and is updated when the instance count changes. This topic describes the supported monitoring metrics for these tasks, including metrics for ECS, Elastic Container Instance (ECI), and Application Load Balancer (ALB).
ECS metrics
The following table describes the metrics of Elastic Compute Service (ECS) instances that are supported by event-triggered tasks of the system monitoring type.
Collection source | Metric (Auto Scaling console) | Metric (Auto Scaling API) | Unit | Statistical period | Supported network |
ECS instances | (ECS) CPU Utilization | CpuUtilization | % |
| VPC and classic network |
(ECS) ECS Concurrent Connections | ConcurrentConnections | Count | VPC and classic network | ||
(ECS) Internal Network Outbound Traffic | IntranetTx | Kb/min | VPC and classic network | ||
(ECS) Internal Network Inbound Traffic | IntranetRx | Kb/min | VPC and classic network | ||
(ECS) Public Network Outbound Traffic | VpcInternetTx | Kb/min | VPC and classic network | ||
(ECS) Public Network Inbound Traffic | VpcInternetRx | Kb/min | VPC and classic network | ||
(ECS) System Disk Read (BPS) | SystemDiskReadBps | Byte/s | VPC and classic network | ||
(ECS) System Disk Write (BPS) | SystemDiskWriteBps | Byte/s | VPC and classic network | ||
(ECS) System Disk Read (IOPS) | SystemDiskReadOps | Count/s | VPC and classic network | ||
(ECS) System Disk Write (IOPS) | SystemDiskWriteOps | Count/s | VPC and classic network | ||
CloudMonitor | (Agent) CPU Utilization | CpuUtilizationAgent | % |
| VPC and classic network |
(Agent) GPU Utilization | GpuUtilizationAgent | % | VPC and classic network | ||
(Agent) Idle GPU Memory Percentage | GpuMemoryFreeUtilizationAgent | % | VPC and classic network | ||
(Agent) GPU Memory Usage | GpuMemoryUtilizationAgent | % | VPC and classic network | ||
(Agent) Memory | MemoryUtilization | % | VPC and classic network | ||
(Agent) Average System Load | LoadAverage | None | VPC and classic network | ||
(Agent) Total TCP Connections | TcpConnection | Count | VPC and classic network | ||
(Agent) Established TCP Connections | TcpConnection | Count | VPC and classic network | ||
(Agent) Internal Network Packets Sent | PackagesNetOut | Count/s | VPC and classic network | ||
(Agent) Internal Network Packets Received | PackagesNetIn | Count/s | VPC and classic network | ||
(Agent) Public Network Packets Sent | PackagesNetOut | Count/s | Classic network | ||
(Agent) Public Network Packets Received | PackagesNetIn | Count/s | Classic network |
If you select a metric collected by the CloudMonitor agent, Auto Scaling installs the agent on all instances in the scaling group. Auto Scaling also enables the Automatically Install CloudMonitor Agent on Newly Purchased ECS Instances option in the CloudMonitor console to ensure monitoring data is collected from new ECS instances.
If you select GPU metrics, ensure that GPU drivers are installed on the instances. For more information, see Installation guideline for NVIDIA Tesla and GRID drivers.
Elastic Container Instance metrics
The following table describes the metrics of elastic container instances that are supported by event-triggered tasks of the system monitoring type.
Collection source | Metric (Auto Scaling console) | Metric (Auto Scaling API) | Unit | Statistical period | Supported network |
CloudMonitor | CPU Utilization | EciPodCpuUtilization | % |
| VPC |
Memory | EciPodMemoryUtilization | % |
ALB metrics
If you associate a scaling group of the ECS type or Elastic Container Instance type with an Application Load Balancer (ALB) instance, you can use the (ALB) QPS per Backend Server metric to monitor the resource usage in the scaling group.
Collection source | Metric (Auto Scaling console) | Metric (Auto Scaling API) | Unit | Statistical period | Supported network |
ALB server group | (ALB) QPS per Backend Server | LoadBalancerRealServerAverageQps | Count/s | 15 seconds, 1 minute, 2 minutes, 5 minutes, and 15 minutes Note The 15-second statistical period is available only for scaling groups of the ECS type. | VPC |