A retail company is developing a web analytics application to track user clickstream data in real time. The website events are sent to an Amazon Kinesis Data Stream. An AWS Lambda function is configured as the consumer to process these records and write aggregated metrics to a database. During peak traffic, the developer notices a high rate of ProvisionedThroughputExceededException errors on a single shard, even though the overall stream throughput is well below the limit. Additionally, the Lambda function occasionally terminates before completing the processing of its batch. Which action should the developer take to resolve these issues?
- AModify the producer to use a static string as the partition key to consolidate records on a single shard, and configure the Lambda function to run inside a private VPC subnet without a NAT gateway.
- Modify the producer to use a high-entropy partition key such as session_id instead of a static value, and increase the Lambda function's timeout or decrease its batch size.Cevap
- CKeep the static partition key and increase the Kinesis retention period, while reducing the Lambda timeout to 3 seconds to force fast retries on timed-out batches.
- DModify the producer to use session_id as the partition key, and configure the Lambda execution role's trust policy to grant write access to the target database.
Cevap
Modify the producer to use a high-entropy partition key such as session_id instead of a static value, and increase the Lambda function's timeout or decrease its batch size.
Using a high-entropy partition key such as session_id ensures that data is evenly distributed across all shards in the stream, preventing a single shard from becoming a bottleneck and throwing ProvisionedThroughputExceededException errors. Adjusting the Lambda function's timeout or decreasing the batch size ensures the function has sufficient time to complete processing before the execution is terminated.
Adım Adım Çözüm
Anahtar Kavram
Even shard distribution using high-entropy partition keys in Amazon Kinesis Data Streams, combined with aligning Lambda batch size and timeout settings.