Preventing Overselling in High‑Concurrency PHP E‑commerce: Test Cases and Solutions

The article investigates how to avoid product inventory overselling in a PHP e‑commerce system when handling high concurrency, using a product with initial stock of 15 and 30 total requests (10 concurrent per batch) as the test scenario.

Test 1 – Plain Order : Executed with ab -n 30 -c 10 http://xxxxx.top/code/the_limit/add_order.php. Result: 15 successful stock deductions, inventory dropped to -7, and 8 requests correctly reported insufficient stock, indicating the plain method allows negative inventory.

Test 2 – Unsigned Mode : Executed with ab -n 30 -c 10 http://xxxxx.top/code/the_limit/unsigned.php. Result: 15 successful deductions, inventory reached -6, and 9 insufficient‑stock detections; adding FOR UPDATE lock on the query had little effect.

Test 3 – MySQL Transaction (Row Lock) : Executed with ab -n 30 -c 10 http://xxxxx.top/code/the_limit/ACID.php. Result: 15 successful deductions, inventory never became negative, and 15 correctly reported insufficient stock; the transaction handled 3000 requests with 1000 concurrent threads without failure.

Test 4 – Non‑Blocking File Exclusive Lock : Compared blocking and non‑blocking file locks; both prevented negative inventory, but performance ranked as Transaction > Blocking Lock > Non‑Blocking Lock.

Test 5 – Redis Queue with Optimistic Lock (code excerpt):

<?php
$redis = new Redis();
$redis->connect("127.0.0.1", 6379);
$redis->auth('PASSWORD');
$redis->watch('sales'); // optimistic lock
$sales = $redis->get('sales');
$n = 15;
if ($sales >= $n) {
    insertLog('库存为0 秒杀失败');
    exit('秒杀结束');
}
$redis->multi();
$redis->incr('sales');
$res = $redis->exec(); // 1 on success, 0 on failure
if ($res) {
    // proceed with MySQL update
    $con = new mysqli('localhost','root','PASSWORD','dev');
    $product_id = 1;
    $buy_num = 1;
    sleep(1);
    $sql = "update products set store=store-$buy_num where id=$product_id";
    if (mysqli_query($con, $sql)) {
        echo "秒杀完成";
        insertLog('恭喜 秒杀成功');
    }
} else {
    insertLog('抱歉 秒杀失败');
    exit('抢购失败');
}
function db(){
    global $con;
    $con = new mysqli('localhost','root','WOrd1234.*','dev');
    if(!$con){ echo "数据库连接失败"; }
}
function insertLog($content){
    global $con;
    $sql = "INSERT INTO `order_log` (content) values('$content')";
    mysqli_query($con,$sql);
}
?>

The corresponding load test command was

ab -n 30 -c 10 http://xxxxxx.top/code/the_limit/optimistic_redis_lock.php

. This approach kept inventory from going negative and demonstrated the best performance among all methods.

Final Conclusion : For high‑concurrency flash‑sale scenarios, using Redis with an optimistic lock provides the most reliable and fastest solution to prevent overselling.