Skip to content

HIVE-29777: Iceberg: Add GCS, ADLS, and OSS Hadoop mapping for REST vended credentials - #6655

Open
difin wants to merge 1 commit into
apache:masterfrom
difin:gcs_adls_oss_mappings_for_vended_creds
Open

HIVE-29777: Iceberg: Add GCS, ADLS, and OSS Hadoop mapping for REST vended credentials#6655
difin wants to merge 1 commit into
apache:masterfrom
difin:gcs_adls_oss_mappings_for_vended_creds

Conversation

@difin

@difin difin commented Jul 28, 2026

Copy link
Copy Markdown
Contributor

What changes were proposed in this pull request?

This PR extends Iceberg REST vended credential propagation beyond S3 to GCS, ADLS, and OSS. It refactors IcebergVendedCredentialUtil into provider-specific Hadoop mappers that translate Iceberg StorageCredential config keys into the correct fs.gs.*, fs.azure.*, and fs.oss.* properties. New GCS/ADLS/OSS secret keys are added to hive.conf.hidden.list so they travel via the secure Credentials channel rather than plain job properties. Unit tests cover the new mappings and round-trip behavior. Integration tests are left for follow-up work.

Why are the changes needed?

The original vended-credential work only propagated Hadoop mappings for S3. Iceberg REST catalogs can vend credentials for GCS, ADLS, and OSS as well, but without equivalent mapping those jobs cannot configure the corresponding Hadoop FileSystem connectors on Tez/LLAP execution paths. This change closes that gap so multi-cloud Iceberg deployments can use vended credentials consistently across providers.

Does this PR introduce any user-facing change?

Yes, for deployments using Iceberg REST catalogs with vended credentials on GCS, ADLS, or OSS. Those workloads can now propagate storage credentials to Hadoop FileSystem paths on Tez/LLAP, which was not supported before. Users on S3-only setups or without vended credentials should see no change in behavior.

How was this patch tested?

Unit tests in TestIcebergVendedCredentialUtil were expanded to cover GCS, ADLS, and OSS Hadoop property mapping, secret routing into jobSecrets, mapper selection by URI scheme and config-key fallback, and Credentials-channel round-trip restore. Existing S3 coverage was kept and refactored to use the new mapper layout. No integration coverage was added for GCS/ADLS/OSS in this PR; that is planned as follow-up work.

@difin
difin force-pushed the gcs_adls_oss_mappings_for_vended_creds branch from 166f643 to 254c757 Compare July 28, 2026 20:51
@difin difin changed the title Gcs adls oss mappings for vended creds HIVE-29777: Iceberg: Add GCS, ADLS, and OSS Hadoop mapping for REST vended credentials Jul 28, 2026
@sonarqubecloud

Copy link
Copy Markdown

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants