CloudWatch Dashboard


1. Dashboard Rows#

Row CloudWatch Metrics Statistic Display
Overview RequestCount Sum request rate by load balancer
Overview HealthyHostCount, UnHealthyHostCount Min / Max target group health
Overview HTTPCode_Target_5XX_Count, HTTPCode_ELB_5XX_Count Sum error count and error rate
Overview TargetResponseTime p95 / p99 latency time series
Drilldown HTTPCode_Target_2XX/4XX/5XX_Count Sum status code split
Drilldown TargetConnectionErrorCount Sum target connection failure
Capacity ActiveConnectionCount, NewConnectionCount Sum connection pressure
Capacity ProcessedBytes Sum bandwidth trend
Debug RejectedConnectionCount Sum ALB capacity / connection rejection

2. Notes#

ALB dashboard 要按 load balancer 和 target group 过滤。
如果服务 behind ALB，应用错误率优先看 Target 5xx，不要只看 ELB 5xx。

3. Dimensions#

ALB metrics 不是只有一个维度。CloudWatch 里常见维度包括:

LoadBalancer:
    整个 ALB 入口。
    用来看总 RequestCount、ELB 5xx、connection、processed bytes。

TargetGroup:
    backend target group。
    用来看某个服务的 target health、target 5xx、target latency。

AvailabilityZone:
    AZ 维度。
    用来看是否只有某个 AZ 异常。

TargetGroup + LoadBalancer:
    服务视角最常用。
    ECS service behind ALB 时，优先用这个维度展示 request / error / latency。

选择规则:

Dashboard for ALB platform:
    use LoadBalancer dimension

Dashboard for ECS service behind ALB:
    use TargetGroup + LoadBalancer dimension

Incident drilldown:
    add AvailabilityZone when suspecting single-AZ issue
Backward YACE Dashboard