503 Service Unavailable

The short version: A 503 Service Unavailable means the server is up but cannot take your request right now, typically because it is overloaded, restarting, or in maintenance. It is almost always temporary and on the website's side. Visitors should wait and reload. Owners should check server load, recent deploys, plugins, the CDN, and the logs.

Key takeaways

• A 503 means the server is reachable but will not handle your request just now. The usual reasons: overload, a restart, throttling, or maintenance. RFC 9110 frames it as a temporary overload or scheduled downtime.
• A 502 or 504 is a broken or missing reply. A 503 is the server answering correctly that it will not serve you yet, which makes it deliberate or self-aware rather than a fault.
• A healthy 503 carries a Retry-After header. This is the SEO-safe way to be down, because Google reads 503 as temporary and comes back later instead of dropping the page the way a 200 error or 404 can.
• A 503 from a CDN like Cloudflare almost always means no healthy origin passed the health check, not that Cloudflare broke. Bypass the edge and hit the origin to tell the two apart.
• On the owner side the usual suspects are an exhausted worker pool (PHP-FPM pm.max_children, or Gunicorn/Puma/Unicorn workers), a crash-looping app, an active rate limit, or a leftover maintenance flag such as WordPress's .maintenance file.

Error type

HTTP 5xx server error

Whose side

Almost always the website

Fix difficulty

Moderate (website owner)

Common cause

Server overload or maintenance

What does 503 Service Unavailable mean?

A 503 Service Unavailable is an HTTP status code that means the server got your request but is not in a state to handle it right now. The machine is reachable and running. The application behind it, though, is overloaded, restarting, throttling traffic, or deliberately switched off for maintenance.

The server is choosing to refuse rather than failing to be reached, so a 503 is nearly always something on the website's side, not your browser, device, or connection. A well-behaved 503 includes a Retry-After header telling clients how long to wait. As a visitor there is not much to do beyond waiting and reloading.

503

HTTP status code

5xx

Server-side error class

Retry-After

Header that tells clients when to return

How the 503 Service Unavailable error appears

The wording changes depending on your browser, device, or server. Here is how this error commonly shows up:

example.com

503

Service Unavailable

The server is temporarily unable to service your request.

• 503 Service Unavailable
• 503 Service Temporarily Unavailable
• HTTP Error 503
• Error 503 Service Unavailable
• HTTP Server Error 503
• 503 Backend fetch failed (Varnish)

503 vs 500, 502 and 429

All four are server-side responses, but a 503 specifically means the server is reachable yet temporarily refusing the request, which sets it apart from a crash, a bad gateway reply, or a per-client rate limit.

Cloudflare edge errors that can look like a 503

When a CDN sits in front, the exact 5xx code tells you whether the edge or the origin failed, often before you open a log.

What causes 503 Service Unavailable?

• A traffic spike has overloaded the server, so it is shedding requests because CPU, memory, or connection limits are maxed out.
• The website is in maintenance mode for a deploy or update and returns 503 on purpose (WordPress writes a .maintenance file during updates; many frameworks have an equivalent mode).
• The application server or process pool has crashed or is restarting, leaving the web server with nothing healthy to forward requests to.
• The PHP-FPM or worker pool is exhausted: pm.max_children reached, say, or every Gunicorn, Puma, or Unicorn worker busy, so new requests queue or get rejected.
• Rate limiting or DDoS protection is throttling traffic right now, answering 503 to anything above the allowed threshold.
• A CDN or reverse proxy such as Cloudflare, Fastly, or Nginx cannot reach a healthy origin (no backend passed the load balancer health check) and serves its own 503.
• Autoscaling has lagged behind demand, so too few healthy instances sit behind the load balancer to carry the current load.

How to find the cause fast

1. Open your host's status page and confirm whether scheduled maintenance is running.
2. In your hosting dashboard, look at server load: CPU, memory, and the connection or worker limits.
3. Read the web server and application logs for the underlying error (/var/log/nginx/error.log, wp-content/debug.log).
4. Bypass the CDN or proxy and hit the origin directly. That tells you whether the 503 is coming from the edge or from your server.

Terminal

curl -sI https://example.comHTTP/1.1 503 Service UnavailableRetry-After: 120# expected when healthy: HTTP/1.1 200 OK

How 503 Service Unavailable looks from the outside

A 503 is one of the rare outages a server announces about itself, so an external check sees a clean HTTP 503 instead of a connection timeout. When the response carries a Retry-After header, a well-behaved monitor can back off rather than pounding a server that is already struggling. Keep an eye on one pattern in particular: 503s that bunch up around your deploy windows. That is usually maintenance mode or a crash loop, not real traffic overload.

To confirm the exact code a URL returns, or to re-test several at once after a fix, run them through the free bulk URL status checker.

How to fix 503 Service Unavailable

If you are a visitor

1. Give it a minute, then reload. A 503 is often temporary and clears once the server recovers or maintenance wraps up. If the response carries a Retry-After value, that is roughly how long to wait.
2. Reload the page instead of resubmitting a form, so you do not fire off a duplicate action once the website is back.
3. Open the website in a private window or a different browser. That rules out a stale cached error page.
4. See whether it is down for everyone with Is Website Down? If it is down for all, the fault is on their end and waiting is your only move.
5. Bother with your own connection only if other websites are failing too. A 503 almost never points at your network, since the server clearly answered you.

If you run the website

1. First, make sure you did not leave maintenance mode on. On WordPress, an interrupted update can leave a stray .maintenance file in the website root; deleting it usually brings the website back.
2. Look at server resources and load. If a spike is exhausting CPU, memory, or connections, add capacity or scale out so healthy instances keep up, and check that autoscaling thresholds fire early enough.
3. Confirm the application is actually running. Check that PHP-FPM, your app server, or your container is up, restart it if it crashed, then watch that it stays up instead of crash-looping.
4. If requests are queueing, size the worker pool to your traffic and CPU (PHP-FPM pm.max_children, or the worker count for Gunicorn, Puma, or Unicorn) and fix any slow requests tying workers up.
5. Revisit your rate limiting and DDoS rules. Throttling real traffic? Loosen the limits or allowlist the affected clients. Under attack? Leave the protection on and mitigate at the edge.
6. When a load balancer or CDN sits in front, check its origin health checks. A 503 from Cloudflare, an AWS ELB, or Nginx upstream usually means no backend passed the check, so fix or replace the unhealthy origins.
7. Send a correct Retry-After header with intentional 503s (maintenance or throttling) so clients, crawlers, and monitors know it is temporary and when to retry.
8. Read the application and web server logs for the root cause: out-of-memory kills, fatal errors, exhausted database connections, or a plugin, extension, or deploy you just introduced.

Still not fixed? Next steps

• When the 503 lines up with your deploy windows, it is maintenance mode or a crash loop rather than real overload. Look for a stuck maintenance flag, then restart and watch whether the app actually stays up.
• Requests piling up under load? Size the worker pool to your traffic and CPU, and chase down the slow requests that pin workers. Throwing hardware at it alone rarely fixes the queue.
• A 503 from a load balancer or CDN means you check origin health. No backend passing the check tells you every instance is unhealthy, so repair or replace the origins.
• If real users are getting throttled, loosen the rate limit or allowlist them. If it is an attack, leave the protection on and absorb it at the edge instead of the origin.

Code & configuration

Copy-paste starting points. Replace example.com and the paths with your own, and test changes on staging before production.

Return a correct Retry-After with intentional 503s (Nginx)

error_page 503 @maintenance;
location @maintenance {
    add_header Retry-After 3600 always;
    return 503;
}

Clear a stuck WordPress maintenance state

# an interrupted update leaves this file in the website root
rm /var/www/html/.maintenance

How to prevent 503 Service Unavailable

A 503 often lasts only seconds or minutes, which is exactly why it slips through until a customer reports it. Pulsetic checks your website and key endpoints from multiple locations as often as every 30 seconds, and alerts you by email, SMS, voice call, Slack, Discord, and more the moment they start returning 503s or stop responding. It will not tell you which worker pool filled up, but it will tell you quickly that something is failing, and when it recovers, so you can move before your visitors do.

Learn how Pulsetic's uptime monitoring detects this from the outside, across 15+ locations.