Skip to content

Native Iceberg errors print data and delete file paths with their credentials #7559

Open

Description

@dwsmith1983

Describe the bug

Found while working on #7128. Several native Iceberg errors print a data or delete file path exactly as the catalog recorded it:

  • iceberg_location_scoped.rs: "Invalid S3 path {path}", "S3 path {path} has no bucket"
  • iceberg_scan.rs: "Failed to stat delete file '{path}'" and the size check after it
  • iceberg_write.rs: "{path} is not the data file this output was made for", "... on the output of data file {path}"
  • planner.rs: "Delete file '{file_path}' has a negative {what}"

A location with a password in its userinfo or a SAS token or signature in its query string would end up in task failures and logs. Catalog locations rarely carry credentials, so this is low risk, but it is the same leak #7128 fixes for Parquet, Delta and object store paths.

Steps to reproduce

Each message formats the path with {path} as is, so a location like s3://alice:SECRET@bucket/t/delete.parquet?sig=TOKEN reaching any of them is printed with SECRET and TOKEN.

Expected behavior

These messages show the path without its password, username or query, using the URL redaction helpers #7127 adds, as the #7128 pull requests do for the other paths.

Additional context

Not part of the four #7128 pull requests, which cover object store setup, the Parquet writer, native Delta and the JVM rejected path reason.

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

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions