voonhous opened a new issue, #19472: URL: https://github.com/apache/hudi/issues/19472
### Describe the problem The [AWS CloudWatchReporter section](https://hudi.apache.org/docs/metrics#aws-cloudwatchreporter) of the metrics page explains how to turn the reporter on and how to supply credentials, but never says which artifact provides the reporter class. The class ships in the optional `hudi-aws` module and is loaded reflectively by `MetricsReporterFactory`. `packaging/hudi-spark-bundle/pom.xml` does not shade `hudi-aws`, while `hudi-flink-bundle` and `hudi-kafka-connect-bundle` do. So a Spark user who follows the docs, sets `hoodie.metrics.reporter.type=CLOUDWATCH`, and has only the Spark bundle on the classpath cannot load `org.apache.hudi.aws.metrics.cloudwatch.CloudWatchMetricsReporter`, and the docs give no hint that a second jar is required. ### Proposed fix Add a short note to the AWS CloudWatchReporter section naming the `hudi-aws` module, the fact that not every engine bundle shades it, and the `hudi-aws-bundle` remedy. The website source lives on the `asf-site` branch, so this is a separate PR from any code change; it applies to `website/docs/metrics.md` and `website/versioned_docs/version-1.2.0/metrics.md`. ### Relationship to other issues - #15293 is the parent complaint. #19418 addresses the runtime half of it, turning the unactionable `Unable to load class` into an error that names the missing class and the bundle to add. This issue is the prevention half: a user who reads the docs first never reaches that error. - Not covered by this: #12182 and #13051 are an `ArrayIndexOutOfBoundsException` inside `CloudWatchReporter.stageMetricDatum`, where the reporter loads and then fails while publishing. #12902 reports `Unable to instantiate class`, which comes from the constructor branch of `ReflectionUtils.loadClass` and so means the class resolved but construction failed, most likely relocation skew rather than a missing artifact. All three are separate defects. -- 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]
