Multi-node schema abort — follower kill and crash recovery

XMLWordPrintableJSON

    • Type: Task
    • Resolution: Unresolved
    • Priority: Major - P3
    • None
    • Affects Version/s: None
    • Component/s: None
    • Storage Engines - Foundations
    • 386.3
    • SE Foundations - 2026-08-04
    • 3

      Extends WT-17974 with KILL_FOLLOWER crash recovery. Validates that both the leader and follower homes recover correctly after the follower is killed.

      Add KILL_FOLLOWER mode to the test from WT-17974:

      • Parent sends SIGKILL to the follower after at least one checkpoint pickup completes.
      • Leader continues running until the parent kills it after the follower exits.

      Verification:

      • Leader home: open with disaggregated=(lose_all_my_data=true), query last_disaggregated_schema_epoch as the epoch cutoff, replay leader per-thread record files up to the cutoff as in WT-17972.
      • Follower home: open as leader for recovery so the page log checkpoint populates last_disaggregated_schema_epoch, then replay follower per-thread record files up to the cutoff using the same verification logic.

      Deliverables:

      • Extend test/csuite/schema_disagg_leader_abort/ from WT-17974 with KILL_FOLLOWER kill mode and dual-home verification in main.c and verify.c.
      • Evergreen stress variant in test/evergreen_disagg.yml with extended run time.

            Assignee:
            [DO NOT USE] Backlog - Storage Engines Team
            Reporter:
            Jie Chen
            Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

              Created:
              Updated: