Skip to content

[Subtask] TMS: In-job MaintenanceTypeHandler pipeline skeleton #13824

Closed

Description

@lasdf1234

Describe the subtask

Add the in-job maintenance pipeline contract in plugins/tms so a single Spark Job can run an ordered chain of built-in types without inventing a second task-type enum:

  • MaintenanceTypeHandler + AbstractMaintenanceTypeHandler (sample hooks gated by metrics capability)
  • MaintenanceTypeOrchestrator (per type: sampleBefore → drift → shouldExecute → execute → sampleAfter; SKIP_TYPE / drift ends that type only)
  • MaintenanceMetricsSupport (compaction + manifest-rewrite only)
  • Four stub handlers (compaction, manifest-rewrite, snapshot-expiry, orphan-cleanup) + registry (commitChain / singleton)
  • Unit tests for registry, metrics capability, and orchestrator control flow

Out of scope: Real S0 metrics persistence / recommend gate (#13777), BuiltInJob / Spark leaf execution, commit-chain OCC / validation metrics, scheduler / OccupancyGate.

Design: §5.2 in apache/gravitino#13386.

Parent issue

#13760

Activity

  1. lasdf1234 commented on Oct 13, 2026

    @lasdf1234
    CollaboratorAuthor

    PR: #13825

  2. self-assigned this
    on Oct 14, 2026
  3. lasdf1234 commented on Oct 26, 2026

    @lasdf1234
    CollaboratorAuthor

    Closing: this subtask is no longer needed. The work is either superseded by what already landed on branch-tms-scheduler (#13885 / #13887 / #13984 / #14223), or intentionally out of scope for the current maintenance scheduler path. Not pursuing this issue further.

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

Metadata

Metadata

Assignees

Labels

2.0.0Release v2.0.0subtaskSubtasks of umbrella issue

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions