AWS amazondynamodb documentation change
Summary
Added notes clarifying how HTTP 500 errors from AWS FIS experiments are counted in both SystemErrors and FaultInjectionServiceInducedErrors metrics. Added documentation for new 'Source' dimension in ConsumedWriteCapacityUnits metric to distinguish customer writes from global table replication writes.
Security assessment
Changes provide operational clarity about metric behavior during fault injection experiments and capacity monitoring. No security vulnerability is addressed; changes enhance monitoring capabilities without introducing security features.
Diff
diff --git a/amazondynamodb/latest/developerguide/metrics-dimensions.md b/amazondynamodb/latest/developerguide/metrics-dimensions.md index 70b933532..5b8303455 100644 --- a//amazondynamodb/latest/developerguide/metrics-dimensions.md +++ b//amazondynamodb/latest/developerguide/metrics-dimensions.md @@ -861,0 +862,4 @@ When DynamoDB returns a system error (HTTP 500), most AWS SDKs automatically per +###### Note + +The HTTP 500 errors that AWS FIS injects during a fault injection experiment are also counted in `SystemErrors`, in addition to being counted in `FaultInjectionServiceInducedErrors`. As a result, alarms on `SystemErrors` fire while an experiment is running. To distinguish experiment-induced errors from organic service errors, compare `SystemErrors` with `FaultInjectionServiceInducedErrors` for the same period. + @@ -865,0 +870,4 @@ Dimensions: `TableName, Operation` +###### Note + +DynamoDB publishes `SystemErrors` only on the `TableName` and `Operation` dimension pair. It does not publish this metric on `TableName` alone. A CloudWatch alarm that specifies only the `TableName` dimension never matches data and remains in the `INSUFFICIENT_DATA` state, so specify both `TableName` and `Operation`. + @@ -1197,0 +1206,4 @@ The requests to DynamoDB that generate a simulated HTTP 500 status code during t +###### Note + +These errors are a subset of the `SystemErrors` metric: each error that AWS FIS injects is counted in both `FaultInjectionServiceInducedErrors` and `SystemErrors`. Use this metric to identify how much of `SystemErrors` is due to a fault injection experiment rather than to organic service errors. + @@ -1228,0 +1241,2 @@ The metrics for DynamoDB are qualified by the values for the account, table name + * Source + @@ -1322,0 +1337,16 @@ This dimension limits the data to a particular AWS region. It is used with metri +### Source + +DynamoDB emits this dimension for the `ConsumedWriteCapacityUnits` metric and lets you distinguish what drives the write capacity consumption. It can have one of the following values: + +`Customer` + + +Write capacity consumed by writes that you make directly to the table (including writes to a replica table in a global table). + +`GlobalTable` + + +Write capacity consumed by global table replication writes that DynamoDB applies to a replica table to keep the replicas in sync. + +Use this dimension to separate your own write consumption from global table replication consumption, for example when planning capacity. +