📄 中文摘要
在处理 AI 作业时，锁的释放通常在 finally 块中进行，以确保资源的正确管理。然而，在某些情况下，这种做法可能导致不必要的成本。作者在工作队列中遇到的问题是，AI 调用可能会失败，导致用户在重试时再次触发相同的作业。通过在失败时释放锁，反而可能使得系统在高负载下频繁调用 AI，增加了成本。因此，考虑使用 Redis 的 TTL 机制，可以在一定时间内限制重试频率，从而降低资源浪费和成本。此方法提供了一种新的思路来优化 AI 作业的处理。
泄漏锁的案例：将 Redis TTL 作为昂贵 AI 作业的故障冷却时间
📄 English Summary
The Case for Leaky Locks: Redis TTL as Failure Cooldown for Expensive AI Jobs
Releasing locks in a finally block is a common practice to ensure proper resource management, but it can lead to unnecessary costs in certain scenarios. The author encountered an issue where AI calls could fail, resulting in users retrying the same job, which triggered expensive AI evaluations repeatedly. By releasing the lock upon failure, the system could end up calling the AI model too frequently under high load, increasing costs. Implementing a TTL mechanism with Redis could help limit the retry frequency, reducing resource waste and expenses. This approach offers a new perspective on optimizing the handling of AI jobs.