Skip to content

[SPARK] Remove hadoop-huaweicloud and its okhttp/okio dependencies #57794

Description

@anupamme

hadoop-huaweicloud (the Huawei OBS connector) has received no meaningful updates since it was added to Hadoop in 2021:

https://github.com/apache/hadoop/commits/trunk/hadoop-cloud-storage-project/hadoop-huaweicloud

Spark has zero source code, tests, or documentation that reference it; it was a pure transitive dependency bundled via the hadoop-cloud module.

It pulls in okhttp 3.12.12, which is flagged by CVE-2021-0341 (hostname verification bypass in OkHostnameVerifier). Upgrading okhttp is problematic: 4.x is Kotlin-based and drags in kotlin-stdlib and okio 2.x/3.x, creating a dependency management challenge where the net advisory count can go up rather than down depending on the version chosen.

The cleanest fix is to remove hadoop-huaweicloud entirely, which eliminates the CVE along with the entire okhttp/okio/kotlin transitive tree.

Scope of work:

  • Remove hadoop-huaweicloud dependency from hadoop-cloud/pom.xml
  • Keep the exclusion from hadoop-cloud-storage (it pulls huaweicloud transitively)
  • Remove okhttp and okio dependencies and version properties
  • Remove huaweicloud.deps.scope property from root pom.xml
  • Remove huaweicloud-provided profile
  • Remove okhttp/okio from LICENSE-binary and dependency manifest

This is not a backport candidate: Spark is not meaningfully affected on supported default paths, and the connector saw no usage in Spark's codebase.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions