Skip to content

FUTURE_ENHANCEMENTS.md

Latest commit

 

History

History
111 lines (85 loc) · 5.29 KB

FUTURE_ENHANCEMENTS.md

File metadata and controls

111 lines (85 loc) · 5.29 KB
Raw

Hinweis: Vage Eintraege ohne messbares Ziel, Interface-Spezifikation oder Teststrategie mit <!-- TODO: add measurable target, interface spec, test strategy --> markieren.

Transaction Module - Future Enhancements

Scope

  • Reliability and safety hardening of transaction lifecycle, isolation, and distributed coordination paths.
  • Operational and observability hardening for SAGA, OCC, and coordinator-driven flows.
  • Performance and resilience hardening for high-concurrency transaction workloads.
  • Completed: Stub #279 — Distributed Transaction Manager Phase-2 fail-closed behavior (Q3 2026) — all remote participants now receive COMMIT/ABORT decisions with fail-fast validation

Design Constraints

  • Transaction state transitions must remain deterministic and fail-safe under partial failures (Target: ongoing)
  • Isolation and conflict semantics must remain correctness-first under concurrency pressure (Target: ongoing)
  • Distributed coordinator decisions must be durable and recoverable before external commit effects (Target: Q4 2026)
  • Compensation logic must remain idempotent and replay-safe across retries/recovery (Target: Q4 2026)
  • Public transaction APIs remain additive-only in active major versions (Target: ongoing)

Required Interfaces

Interface Consumer Notes
TransactionManager / Transaction APIs query/server/runtime layers core lifecycle and isolation behavior
DistributedTransactionManager distributed coordinator paths prepare/commit/abort and recovery semantics
SAGAOrchestrator / distributed SAGA interfaces orchestration/business flows compensation ordering and remote step behavior
LockManager and conflict paths concurrency control lock acquisition and contention safety
TransactionBatcher ingest/write-heavy paths batching throughput and bounded latency
TransactionAuditor audit/compliance operators append/query/export traceability

Implementation Notes

Lifecycle and Isolation Hardening

Priority: High Target: Q3-Q4 2026

  • Strengthen edge-case coverage for begin/prepare/commit/abort transitions.
  • Harden isolation and conflict handling under mixed high-contention workloads.
  • Extend deterministic error-path coverage for timeout and rollback behavior.

Distributed Coordination Hardening

Priority: High Target: Q4 2026

  • Expand in-doubt, recovery, and participant-liveness regression packs.
  • Tighten coordinator durability guarantees and replay safety.
  • Improve observability for distributed transition timelines.

SAGA and Compensation Hardening

Priority: Medium Target: Q4 2026

  • Validate compensation idempotency under retry storms and partial remote failures.
  • Expand orchestration DAG and dependency-failure test coverage.
  • Harden manual intervention and recovery paths.

Throughput and Operations Hardening

Priority: Medium Target: Q1 2027

  • Re-baseline throughput/tail-latency envelopes for transaction-heavy profiles.
  • Keep audit and batching overhead bounded under peak concurrency.
  • Expand benchmark evidence for production guardrail tuning.

Test Strategy

  • Focused transaction lifecycle/isolation regression suites.
  • Distributed coordinator and failure-injection matrix for prepare/commit/abort/recovery.
  • SAGA compensation and replay-safety regression matrix.
  • Performance regressions for throughput, tail latency, and contention envelopes.

Performance Targets

  • Maintain stable transaction tail-latency envelopes under representative mixed workloads.
  • Keep throughput regressions inside release budget thresholds.
  • Keep batching/audit overhead bounded under sustained peak concurrency.

Security / Reliability

  • Fail closed on invalid transaction states and unsafe transition preconditions.
  • Preserve deterministic timeout/cancellation/rollback behavior.
  • Prevent unbounded memory growth in queues and coordination structures.

Risk Backlog

Risk 1: Distributed in-doubt reconciliation drift

Severity: High Signal: inconsistent participant/coordinator completion states after faults. Mitigation: stronger recovery validation, replay guards, and transition telemetry.

Risk 2: Compensation divergence under repeated retries

Severity: Medium Signal: repeated compensation attempts produce inconsistent outcomes. Mitigation: stricter idempotency checks and replay-safety regression packs.

Risk 3: Contention-driven tail-latency spikes

Severity: Medium Signal: p99 latency grows sharply in mixed isolation workloads. Mitigation: contention diagnostics, bounded queue behavior, and tuning guardrails.

Adoption Scenarios

Scenario A: Correctness-first lane

  • Prioritize isolation, state-transition safety, and deterministic recovery behavior.
  • Promote only after full lifecycle and failure-injection gate pass.

Scenario B: Throughput-first lane

  • Prioritize batching and coordinator throughput hardening with safety parity.
  • Promote only after benchmark and regression gate pass.

Scenario C: Distributed-first lane

  • Prioritize multi-participant reliability and compensation/replay robustness.
  • Promote only after distributed fault-matrix gate pass.