gguptp opened a new pull request, #262:
URL: https://github.com/apache/flink-connector-aws/pull/262

   ## Purpose of the change
   
   Updates the AWS SDK v2 to 2.59.2. This version adds account-based endpoints 
(ABEs) for DynamoDB Streams, so the DynamoDB Streams source sends requests to 
`<account-id>.streams-ddb.<region>.amazonaws.com` instead of the regional 
endpoint. The SDK takes the account ID from the stream ARN (`DescribeStream`, 
`GetShardIterator`) or from the shard iterator (`GetRecords`). Users can turn 
ABEs off with the SDK setting `account_id_endpoint_mode=disabled` (for example 
`AWS_ACCOUNT_ID_ENDPOINT_MODE=disabled`).
   
   Other changes that come with the new SDK version:
   
   - **Default retry mode.** SDK 2.58.0 changed the default retry mode from 
`legacy` to `standard`, and the `standard` and `adaptive` modes now use retry 
behavior 2.1 
([changelog](https://github.com/aws/aws-sdk-java-v2/blob/master/changelogs/2.58.x-CHANGELOG.md#2580-2026-10-27)).
 Clients that use the SDK default retry strategy, for example the DynamoDB sink 
and the DynamoDB Streams split readers, now use standard retries. 
`AWSClientUtilTest` now expects `DefaultStandardRetryStrategy`. Users can keep 
the previous behavior with `AWS_RETRY_MODE=legacy`.
   - **NOTICE files.** The SDK `eventstream` artifact moved from 
`software.amazon.eventstream` to `software.amazon.awssdk`, so the 
`software.amazon.awssdk:*` shade include now bundles it. The NOTICE files of 
the three SQL connectors now list `software.amazon.awssdk:eventstream:2.59.2`. 
I checked that its classes are in each shaded jar.
   
   ## Verifying this change
   
   This change is already covered by existing tests:
   
   - Unit tests of all modules pass on JDK 17 (784 tests), and the 
`flink-connector-aws-base` and `flink-connector-dynamodb` tests pass on JDK 11 
with `-P java11-target`. I did not run the Docker-based ITCases locally.
   - `mvn clean install -Pcheck-convergence` passes, and the CI license check 
(`exec:java@check-license`) passes.
   
   I also ran a job against real AWS (us-east-1) with the connector built from 
this branch. The job writes 2,000 items with `DynamoDbSink` and reads their 
`INSERT` events back with `DynamoDbStreamsSource` from `TRIM_HORIZON`. An SDK 
interceptor logged the host of each call.
   
   | Run | Result | Hosts |
   |---|---|---|
   | Local MiniCluster, non-blocking `GetRecords` | 2000/2000 items read | All 
DynamoDB Streams calls and `BatchWriteItem` went to the ABEs 
(`<account-id>.streams-ddb...`, `<account-id>.ddb...`) |
   | Local MiniCluster, blocking `GetRecords` | 2000/2000 items read | Same as 
above |
   | Amazon Managed Service for Apache Flink (Flink 2.2 runtime), non-blocking 
| 2000/2000 items read | All DynamoDB Streams calls went to the ABE. 
`BatchWriteItem` went to the regional `dynamodb.us-east-1.amazonaws.com` |
   
   On Managed Service for Apache Flink, `BatchWriteItem` has no table ARN 
parameter, so the SDK takes the account ID from the credentials. The 
credentials there did not supply one, so the SDK used the regional endpoint. 
This is the existing DynamoDB behavior and this PR does not change it. The 
DynamoDB Streams calls do not depend on the credentials, because the account ID 
comes from the stream ARN or the shard iterator.
   
   ## Significant changes
   
   - [x] Dependencies have been added or upgraded
   - [ ] Public API has been changed (Public API is any class annotated with 
`@Public(Evolving)`)
   - [ ] Serializers have been changed
   - [ ] New feature has been introduced
     - If yes, how is this documented? (not applicable / docs / JavaDocs / not 
documented)
   
   ---
   
   ##### Was generative AI tooling used to co-author this PR?
   
   - [X] Yes (please specify the tool below)
   
   Generated-by: Claude Code (Claude Opus 5.5)
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to