CloudWatch Dashboard


1. Dashboard Rows#

Row CloudWatch Metrics Statistic Display
Overview ApproximateAgeOfOldestMessage Maximum stat and time series by queue
Overview ApproximateNumberOfMessagesVisible Average time series by queue
Overview ApproximateNumberOfMessagesNotVisible Average time series by queue
Overview DLQ ApproximateNumberOfMessagesVisible Average stat by DLQ
Drilldown NumberOfMessagesSent, NumberOfMessagesDeleted Sum producer vs consumer throughput
Drilldown NumberOfMessagesReceived, NumberOfEmptyReceives Sum polling behavior
Capacity visible backlog / delete rate metric math drain signal
Debug ApproximateNumberOfMessagesDelayed Average delay queue investigation

2. Notes#

DLQ 是独立 queue，需要单独选择或用 queue naming convention 过滤。
Approximate metrics 适合趋势和告警，不适合精确计数审计。

3. ops/s Unit#

ops/s 是 operations per second。

在 SQS dashboard 里，它表示 message-related metric event 的每秒速率，不是业务交易数：

sent:
    NumberOfMessagesSent / period_seconds

received:
    NumberOfMessagesReceived / period_seconds

deleted:
    NumberOfMessagesDeleted / period_seconds

empty receive:
    NumberOfEmptyReceives / period_seconds

CloudWatch 里的这些 metric 通常取 Statistic: Sum：

Sum:
    当前 period 内的总次数

ops/s:
    Sum / period_seconds

例子:

period = 300
NumberOfMessagesDeleted Sum = 1500

ops/s = 1500 / 300 = 5

所以 dashboard 里如果你看到 ops/s，本质上是在把 CloudWatch 的 period 总数换算成每秒速率，方便比较不同时间窗口下的吞吐。

Backward YACE Dashboard