CloudWatch Dashboard


1. Scope#

这个方案是 Grafana 直接使用 CloudWatch datasource，不经过 YACE / VictoriaMetrics。

datasource:
    CloudWatch

use when:
    dashboard 数量少
    查询频率低
    不需要 PromQL / vmalert / 长期低成本保存

2. Dashboard Rows#

Row CloudWatch Metrics Statistic Display
Overview StatusCheckFailed Maximum stat, 0 healthy / 1 failed
Overview CPUUtilization Average time series by InstanceId
Overview NetworkIn, NetworkOut Sum time series by InstanceId
Drilldown StatusCheckFailed_Instance, StatusCheckFailed_System Maximum time series by InstanceId
Drilldown VolumeReadOps, VolumeWriteOps Sum time series by VolumeId
Drilldown VolumeReadBytes, VolumeWriteBytes Sum time series by VolumeId
Capacity CPUCreditBalance Average time series for T family
Capacity CPUSurplusCreditsCharged Sum time series for T Unlimited cost
Capacity BurstBalance Average time series by VolumeId
Debug VolumeQueueLength Average time series by VolumeId

3. Variables#

region:
    CloudWatch region

InstanceId:
    EC2 instance dimension

VolumeId:
    EBS volume dimension

4. Notes#

EC2 CloudWatch 默认没有 memory / filesystem 指标。
如果要在 CloudWatch datasource 里展示 memory / disk，需要 CloudWatch Agent 写入 CWAgent namespace。

如果只看 EC2 / EBS 原生指标，CloudWatch datasource 足够。
如果要和 app metrics、node_exporter、vmalert 统一，使用 YACE + VictoriaMetrics。
Backward YACE Dashboard