From e7e319a5b80084fd2c3946f02f85a2b69331e198 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Vi=CC=81ctor=20Falco=CC=81n?= <victoor89@gmail.com>
Date: Sun, 5 Jul 2026 11:14:01 +0200
Subject: [PATCH] fix(queue): raise retry_after above the longest job timeout
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The database queue connection used retry_after=90s while several jobs
declare timeouts far above it (up to 600s). When a job outlives its
reservation window the queue hands it to a second worker; a tries=1 job
then fails with MaxAttemptsExceededException and re-runs its side
effects (double model billing, reset progress) — exactly what
CategorizeUncategorizedTransactionsJob's tries=1 was meant to prevent.

Raise the default to 900s (env-overridable via DB_QUEUE_RETRY_AFTER) so
it exceeds the longest job timeout, and add a guard test asserting every
app/Jobs $timeout stays below retry_after so a new long job cannot
silently reintroduce the bug.

Fixes PHP-LARAVEL-2D
---
 config/queue.php                  |  7 +++++-
 tests/Feature/QueueConfigTest.php | 37 +++++++++++++++++++++++++++++++
 2 files changed, 43 insertions(+), 1 deletion(-)
 create mode 100644 tests/Feature/QueueConfigTest.php

diff --git a/config/queue.php b/config/queue.php
index 4fe599aa..cfa5fbc2 100644
--- a/config/queue.php
+++ b/config/queue.php
@@ -40,7 +40,12 @@ return [
             'connection' => env('DB_QUEUE_CONNECTION'),
             'table' => env('DB_QUEUE_TABLE', 'jobs'),
             'queue' => env('DB_QUEUE', 'default'),
-            'retry_after' => (int) env('DB_QUEUE_RETRY_AFTER', 90),
+            // retry_after must exceed the longest job $timeout on this connection
+            // (currently 600s). Otherwise a long job outlives its reservation, the
+            // queue hands it to a second worker, and a tries=1 job dies with
+            // MaxAttemptsExceededException (and re-runs its side effects). The
+            // worker --timeout must sit between the longest timeout and this value.
+            'retry_after' => (int) env('DB_QUEUE_RETRY_AFTER', 900),
             'after_commit' => true,
         ],
 
diff --git a/tests/Feature/QueueConfigTest.php b/tests/Feature/QueueConfigTest.php
new file mode 100644
index 00000000..e510cc45
--- /dev/null
+++ b/tests/Feature/QueueConfigTest.php
@@ -0,0 +1,37 @@
+<?php
+
+use Illuminate\Support\Facades\File;
+use Illuminate\Support\Str;
+
+/**
+ * The database queue driver releases a job back to the queue once its
+ * reservation is older than retry_after. If a job runs longer than that, a
+ * second worker picks it up: a tries=1 job then dies with
+ * MaxAttemptsExceededException and re-runs its side effects. Guard the
+ * invariant so a new long-running job can never silently reintroduce it.
+ */
+test('every queued job timeout stays below the database queue retry_after', function () {
+    $retryAfter = (int) config('queue.connections.database.retry_after');
+
+    $offenders = collect(File::allFiles(app_path('Jobs')))
+        ->map(function ($file): string {
+            $relative = Str::of($file->getPathname())
+                ->after(app_path().DIRECTORY_SEPARATOR)
+                ->replace(DIRECTORY_SEPARATOR, '\\')
+                ->replaceLast('.php', '');
+
+            return 'App\\'.$relative;
+        })
+        ->filter(fn (string $class): bool => class_exists($class))
+        ->mapWithKeys(function (string $class): array {
+            $timeout = (new ReflectionClass($class))->getDefaultProperties()['timeout'] ?? null;
+
+            return [$class => $timeout];
+        })
+        ->filter(fn ($timeout): bool => is_int($timeout) && $timeout >= $retryAfter);
+
+    expect($offenders->all())->toBe(
+        [],
+        'These jobs have a $timeout >= retry_after ('.$retryAfter.'s); raise DB_QUEUE_RETRY_AFTER above the longest job timeout.'
+    );
+});