Skip to content

Module failover Roadmap

github-actions[bot] edited this page Aug 31, 2026 · 1 revision

Failover Module Roadmap

Current Status

Production failover runtime exists across automatic failover orchestration, disaster recovery plan execution, and queue/retry telemetry surfaces.

In Progress

  • hardening dependency-degraded and multi-step recovery edge behavior (Delivered: Q3 2026)
  • benchmark stabilization for recovery lifecycle hot paths (Delivered: Q3 2026)
  • diagnostics consistency improvements for failover queue pressure and DR failures (Delivered: Q3 2026)
  • canonical retry escalation contract helpers added (isRetryEscalationCode, toRetryTimeoutSource) in failover_api_contract.h with focused test coverage in test_failover_contract_hardening_focused.cpp (Delivered: 2026-08-09)

Planned Features

Short-term (3-6 months)

  • tighten deterministic behavior under concurrent multi-node failover storms (Target: Q4 2026) — DELIVERED 2026-08-24: topology versioning (FO-IMPL-006), consensus quorum hardening (Wave B)
  • expand regressions for fencing/quorum dependency edge scenarios (Target: Q4 2026) — DELIVERED: Wave A/C test suite (fencing, quorum persistence, concurrent promotion)
  • improve operator diagnostics for DR-step failure and retry escalation (Target: Q4 2026) — DELIVERED: runbooks, WAVE_C_CLOSURE_EVIDENCE.md

Mid-term (6-12 months)

  • re-baseline p95/p99 envelopes for failover and recovery orchestration overhead (Target: Q1 2027) — DELIVERED: benchmarks/failover/bench_failover_wave_b_gates.cpp
  • add dedicated benchmark coverage for failover manager and DR step pipelines (Target: Q1 2027) — DELIVERED: FWB-01..08
  • harden long-running reliability under repeated failover/recovery cycles (Target: Q1 2027) — DELIVERED: tests/integration/test_failover_soak_60min.cpp (SOAK-01..04)

Implementation Phases

Phase 1: Design / API Contract

  • freeze failover/recovery manager contracts for active major line (Delivered: Q3 2026)
  • define explicit error taxonomy for queue, dependency, and DR-step failure classes (Delivered: Q3 2026)
    • Contract header: include/failover/failover_api_contract.h
    • Error codes: ELECTION_TIMEOUT, SPLIT_BRAIN_DETECTED, HANDOVER_INCOMPLETE, NODE_REJOIN_FAILED, HEARTBEAT_MISSED, STATE_SYNC_TIMEOUT, QUORUM_UNAVAILABLE, INVALID_EPOCH, INTERNAL_ERROR

Phase 2: Core Implementation

  • complete hardening for queue/worker orchestration and DR-step internals (Delivered: Q4 2026)
    • Bounded stop() with documented worst-case wait per thread (failover_thread_ ≤1 s, monitoring_thread_ ≤health_check_interval)
    • attemptRecovery stats batch-updated: single lock acquisition per call instead of per iteration
    • Canonical lock order documented: failover_mutex_ → stats_mutex_ → callbacks_mutex_
    • executePlan concurrent execution guard (execution_mutex_ + try_to_lock; returns "concurrent execution rejected")
  • align dependency/fencing integration behavior to bounded runtime contracts (Delivered: Q4 2026)
    • preventSplitBrain fails closed (returns false + emitDiagnostic) when no EpochFencingManager is configured
    • emitDiagnostic() helper unifies log + event-callback emission for all fencing/quorum contract violations

Phase 3: Error Handling and Edge Cases

  • standardize fail-closed behavior for invalid plans and unsafe transition scenarios (Delivered: Q4 2026)
    • canTransition() implements real state machine table: IDLE → VERIFYING_FAILURE → CHECKING_QUORUM → STARTING_LEADER_ELECTION → (LEADER_ELECTION_IN_PROGRESS →) UPDATING_METADATA → COMPLETING_FAILOVER → IDLE; FAILED reachable from any state; back to IDLE always valid
    • transitionState() logs a warning for any transition not in the table (non-blocking, observable)
    • preventSplitBrain fails closed when fencing manager is absent (QUORUM_UNAVAILABLE diagnostic emitted)
  • unify diagnostics across queue saturation, retry, and DR-step failures (Delivered: Q4 2026)
    • emitDiagnostic(FailoverErrorCode, node_id, detail) private helper: spdlog::error + emitEvent mapping
    • Used in: preventSplitBrain (QUORUM_UNAVAILABLE), attemptRecovery exhaustion (NODE_REJOIN_FAILED)
    • QUORUM_UNAVAILABLE maps to FailoverEventType::QUORUM_CHECK_FAILED for callback consumers

Phase 4: Tests

  • expand focused regressions for queue pressure and dependency-degraded recovery scenarios (Delivered: Q3 2026)
  • extend deterministic fixture coverage for DR-step permutation and timeout cases (Delivered: Q3 2026)
  • add DR plan validation and step-isolation focused regressions (Delivered: Q3 2026)
    • Test file: tests/failover/test_failover_contract_hardening_focused.cpp
    • Test cases: FCH-01..FCH-16 (election, handover, recovery, error contract)
    • kFailoverContractSeed = 42; all tests self-contained, no external I/O
    • Test file: tests/failover/test_failover_chaos_scenarios.cpp
    • Test cases: 17 scenarios (queue saturation, pressure telemetry, concurrent access, lifecycle)
    • Test file: tests/failover/test_failover_dr_edge_scenarios.cpp
    • Test cases: DRE-01..DRE-08 (plan validation, dry-run, step hook failure, statistics)
    • kDREdgeSeed = 42; all tests self-contained, no external I/O

Phase 5: Performance and Hardening

  • lock benchmark-backed release gates for failover/recovery hot paths (Delivered: Q3 2026)
  • validate p95/p99 and throughput behavior against release baselines (Delivered: Q3 2026)
    • Benchmark file: benchmarks/failover/bench_failover_release_gates.cpp
    • Gates: FRG-01..FRG-06 (heartbeat ≤500µs, election ≤5ms, state-sync ≤200µs, health-check ≤100µs, buffer-check ≤50µs, epoch-persist ≤1ms)
    • kFailoverCanonicalSeed = 42; Repetitions(5)
  • Phase 2/3 hot-path benchmarks delivered (Delivered: Q3 2026, 2026-07-29)
    • Benchmark file: benchmarks/failover/bench_failover_phase2_phase3_gates.cpp
    • Gates: FP23-01..FP23-06 covering canTransition (≤100µs), preventSplitBrain fail-closed (≤200µs), executePlan concurrency guard (≤100µs), attemptRecovery batch stats flush (≤200µs), emitDiagnostic dispatch (≤100µs), triggerManualFailover queue-full drop (≤200µs)
    • kP23CanonicalSeed = 42; Repetitions(5); mock-only (no I/O, no threads)

Phase 6: Documentation and Acceptance

  • core failover module docs aligned to source-verifiable behavior
  • roadmap/future planning separated from historical changelog entries
  • Phase 1-6 Wave 3B Category-D closure delivered (Q3 2026)
    • Contract header, 16 focused tests (FCH-01..16), 6 release-gate benchmarks (FRG-01..06)
  • Q3 2026 status sync: roadmap validated against full module docs (2026-07-29)
    • Test evidence: chaos scenarios (17 tests PASS), contract hardening (FCH-01..16), DR edge (DRE-01..08)
    • Benchmark evidence: FRG-01..FRG-06 (bench_failover_release_gates.cpp)
  • Phase 2/3 hardening delivered (2026-07-29)
    • Test file: tests/failover/test_failover_phase2_phase3_focused.cpp
    • Test cases: P23-01..P23-08 (canTransition table, fail-closed split-brain, batch stats, concurrent DR rejection, emitDiagnostic callback)
    • kPhase23Seed = 42; all tests self-contained, no external I/O
  • Phase 5 benchmark gates for Phase 2/3 hardening delivered (2026-07-29)
    • Benchmark file: benchmarks/failover/bench_failover_phase2_phase3_gates.cpp
    • Gates: FP23-01..FP23-06 (canTransition, preventSplitBrain, concurrency guard, batch stats, emitDiagnostic, queue-full drop)
    • kP23CanonicalSeed = 42; Repetitions(5); all gates documented in benchmarks/failover/README.md

Production Readiness Checklist

  • core failover surfaces documented and source-verified
  • module-level security and failure behavior documented
  • benchmark mapping documented in performance expectations
  • dedicated failover benchmark file delivered (FRG-01..FRG-06)
  • Q3 2026 hardening, benchmark stabilization, and diagnostics items closed
  • remaining hardening tasks closed for dependency/queue/DR-step edge paths (Q4 2026)
  • Phase 2/3 hot-path benchmarks delivered (FP23-01..FP23-06) (2026-07-29)
  • release benchmark stabilization p95/p99 re-baseline complete (Q1 2027) — DELIVERED 2026-08-24: benchmarks/failover/bench_failover_wave_b_gates.cpp FWB-01..08

Known Issues and Limitations

  • runtime outcomes partially depend on external manager availability and behavior.
  • concurrent multi-node failover storm edge cases require continued hardening (Q4 2026).
  • fencing/quorum dependency edge scenarios are covered by contract tests but not yet stress-tested at scale.

Breaking Changes

No breaking failover contract planned. Any contract-breaking change requires migration notes and changelog entry before merge.

Program Execution Model — Wave Context

This module is a contributing module in the program-level Wave A → B → C → D execution model. It does not own a primary wave deliverable but must remain release_critical-green throughout all waves and must deliver Wave D operability improvements in Q1 2027. See [[../../ROADMAP.md|ROADMAP]] for the full wave model and exit criteria.

Wave D Contribution for failover

  • Deliver or validate distributed tracing, high-cardinality stress coverage, exporter reliability, and operator remediation hints as applicable to this module (Target: Q1 2027) — DELIVERED: src/failover/WAVE_D_CLOSURE_EVIDENCE.md; observability test: tests/failover/test_failover_wave_d_observability.cpp
  • Contribute to or validate long-duration soak test coverage for this module's primary paths (Target: Q1 2027) — DELIVERED: tests/integration/test_failover_soak_60min.cpp (SOAK-01..04)
  • Ensure runbook coverage for operator-critical scenarios in this module (Target: Q1 2027) — DELIVERED: docs/operability/failover_runbook_split_brain.md, failover_runbook_fencing_override.md, failover_runbook_manual_recovery.md

Cross-Wave Requirements

  • release_critical CI must remain green on develop throughout all waves (Target: ongoing)
  • p95/p99 benchmarks must be refreshed on representative hardware before Wave D sign-off (Target: Q1 2027)
  • No behavioral regression may be introduced into modules in Wave A/B/C scope from changes in this module.

Program-Level Success Criteria (contribution)

  • This module's distributed/acceleration paths fail closed (Target: Q1 2027) — VERIFIED: fencing fail-closed (FO-IMPL-003), quorum-log fail-closed (FO-IMPL-004), topology versioning (FO-IMPL-006)
  • Benchmark-backed p95/p99 baselines exist on representative hardware (Target: Q1 2027) — DELIVERED: benchmarks/failover/bench_failover_wave_b_gates.cpp (FWB-01..08)
  • Operator-critical paths have diagnostics, alerts, and runbooks (Target: Q1 2027) — DELIVERED: 3 runbooks in docs/operability/

Navigation

Home

Architecture

Governance

Modules

Developer

Clone this wiki locally