Multi-node schema abort — leader kill and follower step-up

XMLWordPrintableJSON

    • Type: Task
    • Resolution: Unresolved
    • Priority: Major - P3
    • None
    • Affects Version/s: None
    • Component/s: None
    • Storage Engines - Foundations
    • 386.293
    • SE Foundations - 2026-08-04
    • 3

      Extends WT-17975 with crash recovery coverage for the leader kill scenario, where the follower detects the leader's death and steps up.

      Add KILL_LEADER mode to the test from WT-17975:

      • Parent sends SIGKILL to the leader.
      • Follower detects pipe EOF, steps up: reconfigures to leader role, sets stable timestamp, takes a checkpoint, and writes a sentinel file.
      • Parent waits for the sentinel, lets the stepped-up follower run briefly as the new leader, then sends SIGKILL.
      • Verify the stepped-up follower home using the same epoch-cutoff logic from WT-17972.

      Deliverables:

      • Extend test/csuite/schema_disagg_leader_abort/ from WT-17975 with KILL_LEADER kill mode and step-up logic in workload.c and main.c.
      • Evergreen task entry in test/evergreen_disagg.yml covering the KILL_LEADER mode, including a stress variant.

            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: