Skip to content

vfs: L0 TXID gaps cannot self-heal #1459

Description

@corylanou

Problem

The VFS build compacts through a bare Compactor with no SourceGapHandler (vfs.go:170) and keeps no local LTX files — dirty state is discarded after a successful upload (vfs.go:1913).

If a provider loses an acknowledged L0 object, the gap detection from #1155 stops compaction at the gap (preventing a non-contiguous destination level), but nothing can heal it: there is no local copy to re-upload.

Scope

Tracking issue. Recovery likely requires a snapshot-based path similar to the DB-side fallback discussed in the local-file-missing issue, but VFS-specific design is needed first. No PR planned until the approach is settled.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions