Intermittent 503 responses
The website worked normally at times but returned HTTP 503 Service Unavailable responses appeared when application demand increased.
WordPress 503 Errors Caused by PHP-FPM Worker Exhaustion
A troubleshooting walkthrough involving intermittent 503 responses on a WordPress website.
The visible symptom was an unavailable website. The useful evidence appeared deeper in the PHP and traffic layers.
The symptoms
No permanent outage
Because the failure was intermittent, checking the website only once could easily miss the underlying problem.
WordPress was only the visible layer
The browser showed a website failure, but the cause required checking PHP execution and request pressure.
What the evidence showed
The 503 response was only the visible symptom. The useful evidence came from comparing PHP capacity, request pressure and application behaviour.
503 responses appeared under increased application demand
PHP execution capacity became saturated
Additional requests had insufficient worker capacity
Traffic pressure contributed to the failure pattern
The database remained available during the affected period
The failure was concentrated in the application execution layer
The failure path
A 503 error does not automatically mean the web server itself has failed.
1. Requests reached the server
The site was reachable and requests were arriving, which narrowed the investigation beyond DNS and connectivity.
2. PHP capacity became saturated
PHP-FPM had insufficient available workers to process all incoming application requests at the same time.
3. Requests waited or failed
Once PHP execution capacity was exhausted, additional requests could not be handled normally and 503 responses appeared.
4. Traffic amplified the problem
Request volume and application behaviour increased pressure on the limited PHP worker pool.
Root cause
The immediate failure was PHP-FPM worker exhaustion. A restart could temporarily restore capacity, but it would not explain why the worker pool became saturated.
Capacity, request behaviour and the reason workers remained occupied had to be considered together.
PHP-FPM capacity was insufficient during peak request pressure
Long-running or expensive requests could keep workers occupied
Traffic volume amplified pressure on the PHP worker pool
Available server resources limited practical concurrency
Application behaviour contributed to sustained worker usage
A permanent resolution required addressing the cause, not only restarting PHP
Why the problem could return
Restarting PHP-FPM can temporarily clear occupied workers, making the website appear fixed.
If the same traffic and application behaviour remain, the worker pool can become exhausted again. The useful troubleshooting question is why the workers were saturated in the first place.
Similar failures can involve PHP-FPM, web servers, resource pressure or application behaviour. See Linux server troubleshooting