Skip to content

Multi-dataset atomic commit primitive #6668

@ragnorc

Description

@ragnorc

CommitBuilder::execute(transaction) is per-dataset. There is no public Rust primitive to atomically commit a set of transactions across multiple Lance datasets, even though the namespace specification and the generated REST client already define one.

Any system that maintains an invariant across two or more Lance datasets — e.g., a join-key relationship, a parent/child reference, or a manifest that names which version of each underlying dataset is currently "published" — must coordinate the per-dataset commits at the application layer. The standard workaround is a single coordinating dataset whose atomic commit serves as the linearization point for everything else. That works, but it leaves a window between the per-dataset commits and the coordinator commit during which partial success is observable.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions