[GitHub] zentol opened a new pull request #7417: [FLINK-11268][elease] Deploy multiple flink-shaded-hadoop2 artifacts

[GitBox <gi...@apache.org>]

zentol opened a new pull request #7417: [FLINK-11268][elease] Deploy multiple flink-shaded-hadoop2 artifacts
URL: https://github.com/apache/flink/pull/7417
 
 
   ## What is the purpose of the change
   
   This PR modifies the release scripts to deploy multiple `flink-shaded-hadoop2` artifacts, one for each supported hadoop version.
   
   With #7416 we only release hadoop-free Flink. Effectively this means that users have to assemble hadoop-specific distributions themselves (if they need flink-shaded-hadoop2). For convenience we should deploy various `flink-shaded-hadoop2` artifacts that can be copied into /lib, one for each hadoop version.
   
   The current `flink-shaded-hadoop2` modules however do not allow deployment of multiple artifacts for different hadoop versions, as the hadoop.version property is not used in neither the artifactId noror version. Attempting to deploy these modules would lead to them overriding each other, and a random version winning out.
   I have opted for including the hadoop.version property in the version of the artifact. We _could_ include it in the artifactId, but we'd either end up with periods in the name (e.g. `flink-shaded-hadoop2-2.4.8`) which is unusual, or require another "version-like" property to be set that doesn't contain periods.
   The downside is that `flink-shaded-hadoop2` modules now have to set `<version>1.8-SNAPSHOT-${hadoop.version}</version>`, that is they have to explicitly refer to the SNAPSHOT version. AFAIK there's no way to refer to the parents version instead (`project.version` refers to the current module).
   
   ## Brief change log
   
   * enable deployment for flink-shaded-hadoop2-uber
   * add hadoop version to flink-shaded-hadoop2(-uber) version
   * modify `deploy_staging_jars.sh` to release multiple versions of hadoop artifacts
   
   ## Verifying this change
   
   The changes to the release scripts can be verified by running the scrip and checking the uploaded artifacts. For convenience I'd remove the calls that deploy Flink itself.
   The pom changes are covered by existing E2E tests and IT cases (the yarn IT cases use flink-dist and require flink-shaded-hadoop to be in /lib).
   

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services