# AWS DynamoDB > Monitor AWS DynamoDB tables with CloudWatch Metrics Streams for latency, throttling, capacity utilization, and replication lag Source: https://last9.io/docs/integrations/databases/aws-dynamodb/ Monitor your Amazon DynamoDB tables with Last9 using CloudWatch Metrics Streams. This integration provides visibility into request latency, throttling, capacity utilization, errors, and Global Tables replication lag — with no agents to deploy. Follow the [AWS CloudWatch Metrics Stream setup guide](/docs/integrations/observability/aws-cloudwatch-metrics/) to configure the stream, then return here for DynamoDB-specific namespace and metric details. ### HTTP Endpoint URL Set the HTTP endpoint as the following. ```sh {{ .Metrics.StreamWriteURL }} ``` ### Additional Parameters Set the following keys and values in the Parameters section. The keys are case-sensitive. username: `{{ .Metrics.Username }}` password: `{{ .Metrics.WriteToken }}` ## Metric Namespace Enable `AWS/DynamoDB` namespace in your CloudWatch Metric Stream to send DynamoDB metrics to Last9. For the full list of available metrics, see the [AWS DynamoDB metrics and dimensions reference](https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/metrics-dimensions.html). --- ## Key Metrics ### Latency | Metric | Description | Dimensions | | -------------------------- | ---------------------------------------------------------------------------------------------------------------- | ------------------------ | | `SuccessfulRequestLatency` | Latency of successful requests. Monitor `avg` and `max` per operation to detect table design or capacity issues. | `TableName`, `Operation` | **Operations:** `GetItem`, `PutItem`, `UpdateItem`, `DeleteItem`, `Query`, `Scan`, `BatchGetItem`, `BatchWriteItem`, `TransactGetItems`, `TransactWriteItems` ### Throttling Throttling is the most critical DynamoDB signal. Monitor all granularities — they expose different problems. | Metric | Description | Dimensions | | ------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------- | | `ThrottledRequests` | Requests where at least one read/write event was throttled. | `TableName`, `Operation` | | `ReadThrottleEvents` | Individual read events throttled. A `BatchGetItem` of 10 items = 10 potential events. Should always be `0`. | `TableName`, `GlobalSecondaryIndexName` | | `WriteThrottleEvents` | Individual write events throttled. A `PutItem` on a table with 3 GSIs = 4 potential events. Should always be `0`. | `TableName`, `GlobalSecondaryIndexName` | | `ReadProvisionedThroughputThrottleEvents` | Reads throttled due to provisioned capacity limits. | `TableName` | | `WriteProvisionedThroughputThrottleEvents` | Writes throttled due to provisioned capacity limits. | `TableName` | | `ReadKeyRangeThroughputThrottleEvents` | Reads throttled due to **hot partition** limits. A table can throttle well below its provisioned capacity if traffic is concentrated on a few partition keys. | `TableName` | | `WriteKeyRangeThroughputThrottleEvents` | Writes throttled due to **hot partition** limits. | `TableName` | | `ReadMaxOnDemandThroughputThrottleEvents` | Reads throttled because on-demand max throughput was exceeded. | `TableName` | | `WriteMaxOnDemandThroughputThrottleEvents` | Writes throttled because on-demand max throughput was exceeded. | `TableName` | | `ReadAccountLimitThrottleEvents` | Reads throttled due to account-level capacity limits. | — | | `WriteAccountLimitThrottleEvents` | Writes throttled due to account-level capacity limits. | — | ### Capacity Utilization | Metric | Description | Dimensions | | --------------------------------------------- | ------------------------------------------------------------------------------------------------------------- | --------------------------------------- | | `ConsumedReadCapacityUnits` | Read capacity units consumed. Compare against `ProvisionedReadCapacityUnits` to track utilization percentage. | `TableName`, `GlobalSecondaryIndexName` | | `ConsumedWriteCapacityUnits` | Write capacity units consumed. | `TableName`, `GlobalSecondaryIndexName` | | `ProvisionedReadCapacityUnits` | Provisioned read capacity (published at 5-minute intervals). | `TableName`, `GlobalSecondaryIndexName` | | `ProvisionedWriteCapacityUnits` | Provisioned write capacity (published at 5-minute intervals). | `TableName`, `GlobalSecondaryIndexName` | | `OnDemandMaxReadRequestUnits` | Maximum on-demand read request units configured for the table. | `TableName` | | `OnDemandMaxWriteRequestUnits` | Maximum on-demand write request units configured for the table. | `TableName` | | `AccountProvisionedReadCapacityUtilization` | Percentage of provisioned read capacity used across the account. | — | | `AccountProvisionedWriteCapacityUtilization` | Percentage of provisioned write capacity used across the account. | — | | `MaxProvisionedTableReadCapacityUtilization` | Utilization of the single most-read table or GSI in the account. | — | | `MaxProvisionedTableWriteCapacityUtilization` | Utilization of the single most-written table or GSI in the account. | — | ### Errors | Metric | Description | Dimensions | | -------------------------------- | ------------------------------------------------------------------------------------------------------------------- | ------------------------ | | `SystemErrors` | HTTP 500 internal errors from DynamoDB. Should always be `0`. | `TableName`, `Operation` | | `UserErrors` | HTTP 400 errors from client-side problems (bad parameters, missing tables, auth failures). Account-level aggregate. | — | | `ConditionalCheckFailedRequests` | Failed conditional writes where the condition expression evaluated to false. | `TableName` | | `TransactionConflict` | Item-level requests rejected due to concurrent transactional conflicts on the same items. | `TableName` | ### Returned Data | Metric | Description | Dimensions | | ------------------- | ------------------------------------------------------------------------------------------------------------------------- | ------------------------ | | `ReturnedItemCount` | Items returned by `Query`, `Scan`, or `ExecuteStatement`. A low ratio versus items scanned indicates inefficient queries. | `TableName`, `Operation` | ### TTL | Metric | Description | Dimensions | | ---------------------------- | --------------------------------------------------- | ----------- | | `TimeToLiveDeletedItemCount` | Items deleted by TTL expiration in the time period. | `TableName` | ### Global Tables (Replication) | Metric | Description | Dimensions | | ------------------------- | ------------------------------------------------------------------------------------------------ | ------------------------------ | | `ReplicationLatency` | Time between an item appearing in one region's stream and arriving in another region's replica. | `TableName`, `ReceivingRegion` | | `PendingReplicationCount` | Item updates written to one replica but not yet replicated to another. Global Tables v2017 only. | `TableName`, `ReceivingRegion` | ### Kinesis Data Streams | Metric | Description | Dimensions | | ------------------------------- | ------------------------------------------------------------------------- | --------------------------------- | | `AgeOfOldestUnreplicatedRecord` | Age in milliseconds of the oldest record waiting for Kinesis replication. | `TableName`, `DelegatedOperation` | | `FailedToReplicateRecordCount` | Records DynamoDB failed to replicate to the Kinesis data stream. | `TableName`, `DelegatedOperation` | | `ThrottledPutRecordCount` | Records throttled by Kinesis due to insufficient stream capacity. | `TableName`, `DelegatedOperation` | --- ## Recommended Alerts | Signal | Condition | Severity | | ----------------------------------------- | ------------------------------------------------------------------ | -------- | | Read throttle events | `ReadThrottleEvents > 0` | Warning | | Write throttle events | `WriteThrottleEvents > 0` | Warning | | Hot partition — read | `ReadKeyRangeThroughputThrottleEvents > 0` | Critical | | Hot partition — write | `WriteKeyRangeThroughputThrottleEvents > 0` | Critical | | System errors | `SystemErrors > 0` | Critical | | Consumed read capacity ≥ 80% provisioned | `ConsumedReadCapacityUnits / ProvisionedReadCapacityUnits > 0.8` | Warning | | Consumed write capacity ≥ 80% provisioned | `ConsumedWriteCapacityUnits / ProvisionedWriteCapacityUnits > 0.8` | Warning | | Global Tables replication lag | `ReplicationLatency > 60000` ms | Warning | | Transaction conflicts | `TransactionConflict > 0` anomaly | Warning | :::tip Alert on **consumed/provisioned capacity ratio** at 80% rather than on throttle events directly. This gives you lead time to scale before requests start failing. ::: :::note `ReadThrottleEvents` and `WriteThrottleEvents` count individual events within batch operations, not just requests. A single `BatchWriteItem` writing to a table with 3 GSIs can generate up to 4 throttle events. Use these alongside `ThrottledRequests` for complete throttle visibility. ::: --- ## Troubleshooting Please get in touch with us on [Discord](https://discord.com/invite/Q3p2EEucx9) or [Email](mailto:support@last9.io) if you have any questions.