Skip to content

[Draft] Enable authoritative CHK Raft planner for e2e - #2115

Merged
alex-zaitsev merged 4 commits into
refactor/chk-raftplan-shadow-20261022from
refactor/chk-raftplan-authoritative-20261022
Oct 23, 2026
Merged

alex-zaitsev merged 4 commits into
refactor/chk-raftplan-shadow-20261022from
refactor/chk-raftplan-authoritative-20261022

Conversation

@alex-zaitsev

@alex-zaitsev alex-zaitsev commented Oct 22, 2026 •

Copy link
Copy Markdown
Member

Scope

Stacked on #2114. The Raft safety planner is enabled by default on this branch, so the existing e2e harness needs no operator-installation environment-variable support.

  • PlanAdmission is authoritative for staged 1→N / 2→3 admission. The existing preliminary exclusion, ConfigMap publication immediately before STS create, rollback, and stop behavior remain.
  • CheckDisruption is authoritative inside the existing quorum wait, including after live readiness refresh.
  • Unknown admission state or inconsistent quorum counts fail closed.
  • Additional tests assert the no-env default, explicit fallback, and safe admission/disruption outcomes.

Run e2e normally

Build/deploy the operator from refactor/chk-raftplan-authoritative-20261022, then run the existing e2e workflow unchanged. The planner activates with no environment variables and no Helm/chart edits.

For optional manual rollback only, CHOP_CHK_RAFTPLAN_ENFORCE=0 restores the previous legacy decisions. The e2e suite does not need to set it.

Recommended scenarios

  1. 1→3: next joiner held while previous joiner is unready.
  2. Operator restart during first admission and ambiguous STS create.
  3. Degraded 3-member quorum during rolling updates.
  4. 2→3, 3→5, stop/resume, and scale-down regression.
  5. Confirm no membership flapping and no unplanned existing-peer restarts.

Status and limitations

  • Full e2e remains to be run by the test owner.
  • No committed Keeper Raft membership verification is added.
  • Existing membership settle delays and StatefulSet waiting behavior remain unchanged.
  • This PR stays draft until CI and e2e validate the authoritative behavior.

Dependency

Stacked on #2114 (observer and shadow parity tests).

@alex-zaitsev alex-zaitsev changed the title [Draft] Opt-in authoritative CHK Raft safety planner [Draft] Enable authoritative CHK Raft planner for e2e Oct 22, 2026
@alex-zaitsev
alex-zaitsev marked this pull request as ready for review October 23, 2026 05:51
@alex-zaitsev
alex-zaitsev merged commit b3c4d4b into refactor/chk-raftplan-shadow-20261022 Oct 23, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant