Preventing Duplicate Side Effects in Event-Driven Systems

说明与用途

在事件驱动系统中，重复交付可能导致外部操作重复。解决方案是使用业务级幂等键、唯一数据库约束、显式状态机和重试分类。

问题

Event-driven systems can deliver the same event multiple times, leading to duplicate side effects when external actions are performed without proper idempotency.

结果

By shifting from event-level to business-level idempotency and adding durable state, retries became safe, eliminating duplicate side effects and improving system reliability at scale.

工作流

• Receive external event via webhook
• Validate and persist event with stable identity
• Enqueue for processing
• Create or claim business action with unique key (account_id+automation_id+source_event_id+action_type)
• Insert action record with unique constraint
• Perform idempotent external API call using same key
• Update durable state to COMPLETED
• Log metrics and trace identifiers

证据

The real problem appears after the action has been claimed. Consider: Create action → Call external API → External API succeeds → Worker crashes → Completion state is never written.

来源

If the external API supports idempotency keys, use them. The same logical operation should always carry the same key.

来源