Lock

You might say

Two requests decrement the last item concurrently. Lock the shared record so one proceeds and the other waits and rechecks.

A lock is a concurrency-control mechanism that limits simultaneous access to a shared resource.When two requests update one inventory record, the lock owner checks and decrements before release; the waiter then rechecks, so only one succeeds. The lock must cover the correct resource. Inconsistent acquisition order can deadlock, and an in-process lock does not automatically protect another service instance.

Know first

Thread

How two inventory requests own, wait for, and release a lock

Inventory record lockstock_9 = 1

Request Areads and decrements

Request Breads concurrently

final stockmay become -1

EvidenceTwo workers enter the critical section

The shared record has no access order.

How a lock orders access to a shared record

Problem and key mechanism: Both requests can read stock 1 and decrement it, causing overselling; lock the shared inventory record so one request checks and updates while the other waits and then rechecks.

Verify the waiter rechecks: Concurrent testing yields one success, one out-of-stock result, and final stock 0. Cross-instance scope and deadlock still need handling.

A second scenario: clinic scheduling: When two staff members schedule the same clinic room and time slot, the lock should cover that shared slot. After the first releases it, the waiter rechecks so two appointments cannot occupy the same slot.

You can say this to an AI Agent

Lock the inventory check and update for the same shared record. Record owner, waiter, release, and timeout. Test two concurrent requests so only one succeeds and the waiter rechecks after acquiring the lock and returns out of stock. If the service has multiple instances, do not rely only on separate in-process locks.

Recommended tool

Web fine-tuning tool · Selector Pro

Select page elements for precise AI edits

Click a button, text, or card, note the change, then copy its exact location to AI.

Open Selector Pro website

Only one item remains and two backend instances decrement it concurrently. Which lock decision is reliable?

A. Use a lock that protects the same database record, then make the waiter recheck stockB. Create one in-process lock per instance and assume the instances now exclude each otherC. Let both requests decrement first, then acquire a lock to record what happened