HTTP 503 Service Unavailable: who is sending it?

What the message looks like

limiting requests, excess: 20.640 by zone "perip", client: 198.51.100.7, server: example.com, request: "GET /api/ HTTP/1.1"

Common causes

• limit_req or limit_conn kicks in – nginx answers with 503 by default.
• Application in maintenance mode, e.g. Nextcloud with maintenance => true.
• Load balancer (HAProxy, cloud LB) has no healthy backend server.
• PHP-FPM or app workers saturated, queue overflowing.
• Service not ready yet after an update or restart.

Check

1. Check headers: Server and Retry-After show who answers
   curl -sSI https://example.com/ | head -n 5Copy
2. Search error.log for rate limit hits
   grep -E 'limiting (requests|connections)' /var/log/nginx/error.log | tail -n 5Copy
3. Find limits and fixed 503 responses in the config
   nginx -T 2>/dev/null | grep -nE 'limit_req|limit_conn|return 503'Copy
4. Show failed services
   systemctl --failedCopy

Fix

• For rate limits set burst and nodelay sensibly and return the right code with limit_req_status 429;.
• End the application's maintenance mode, e.g. occ maintenance:mode --off.
• Check the load balancer health checks: path, port and expected status code.
• Match worker count to RAM and load, e.g. pm.max_children for PHP-FPM.
• Restart the service and check the journal for why it was not ready.

FAQ

What is the difference between 502 and 503?

502 means the backend answered invalidly or not at all. 503 means the service is temporarily unavailable on purpose or due to overload.

Does a 503 hurt Google rankings?

Not short term – 503 is meant for exactly this. If it lasts for days, Google drops pages from the index. Send a Retry-After header during maintenance.