Skip to content

[SPIKE][DO NOT MERGE] Multi-backend providers for Iceberg catalogs (POC)#10893

Open
markhoerth wants to merge 3 commits intoapache:mainfrom
markhoerth:spike/multi-backend-credentials
Open

[SPIKE][DO NOT MERGE] Multi-backend providers for Iceberg catalogs (POC)#10893
markhoerth wants to merge 3 commits intoapache:mainfrom
markhoerth:spike/multi-backend-credentials

Conversation

@markhoerth
Copy link
Copy Markdown
Collaborator

@markhoerth markhoerth commented Apr 28, 2026

POC for multi-backend Iceberg catalog support. Not for merge.

Validates GCS + S3 cross-cloud queries from a single Gravitino IRC
catalog via Trino and Spark (transcript in PRD appendix).

Three changes:

  • new: MultiSchemeFileIO — server-side FileIO dispatch by URI scheme
  • IcebergCatalogUtil: programmatically sets io-impl=MultiSchemeFileIO
    when GCS credentials are configured
  • CatalogWrapperForREST: removes io-impl from catalogPropertiesToClientKeys
    (Gravitino-internal class; must not leak to clients)

Opening as draft for @roryqi review per Slack thread.

markhoerth and others added 3 commits April 17, 2026 15:07
Adds MultiSchemeFileIO for server-side FileIO dispatch by URI scheme,
complementing existing client-side credential dispatch. Validates
GCS + S3 cross-cloud queries from a single Gravitino IRC catalog
via Trino and Spark.

POC only. Not for merge. See PRD for design details.
@roryqi
Copy link
Copy Markdown
Contributor

roryqi commented Apr 29, 2026

Thanks for your draft code. Could I refactor code and continue the work?

@roryqi
Copy link
Copy Markdown
Contributor

roryqi commented Apr 29, 2026

I drafted a new PR. If io-impl is unset, we can infer the impl from the location. #10904

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants