What is the cause of this issue?
The data in the table’s partition key column is not evenly distributed.
The LSI’s capacity is different from the table’s capacity.
The application is not implementing exponential backoff retry logic while interacting with the DynamoDB API.
The application has the IAM permission to query the DynamoDB table but not to query the LSI.
Explanations:
If the partition key’s data distribution is uneven, some partitions may experience higher request rates, leading to throttling despite the overall capacity.
The LSI shares the table’s provisioned throughput. The issue is not due to the LSI’s capacity being different.
While exponential backoff is a best practice, the error is due to provisioned throughput being exceeded, not because of a lack of retries.
The issue is related to throughput, not IAM permissions. Having permissions for the table does not affect throughput limits on LSIs.
I map out that the answer is:
The data in the table’s partition key column is not evenly distributed.