HTTP 504 Gateway Timeout — Backend Response Timeout Analysis

Overview

Deep dive into HTTP 504 Gateway Timeout: diagnosing slow backends, proxy timeout tuning, and database query optimization to resolve upstream timeouts.

Key Details

• 504 means the proxy waited too long for a response from the upstream backend
• Different from 408 (client timeout) — 504 is between proxy and backend
• Default Nginx proxy_read_timeout is 60 seconds
• AWS ALB idle timeout default is 60 seconds, can be increased to 4000
• Cloudflare has a fixed 100-second timeout on free plans (cannot be changed)

Common Causes

• Backend processing a slow database query exceeding proxy timeout
• Backend server overloaded with too many concurrent requests
• Network latency between proxy and backend in different regions
• Deadlocked backend threads unable to process new requests
• Large file processing or report generation exceeding timeout
• DNS timeout resolving the backend hostname

Steps

1. 1Identify slow requests: check backend access logs for response times over 30s
2. 2Increase proxy timeout: proxy_read_timeout 300s; proxy_connect_timeout 300s; in Nginx
3. 3Optimize slow database queries: add indexes, use EXPLAIN to find full table scans
4. 4Implement async processing: return 202 Accepted and process in background
5. 5Add backend monitoring: track response times, CPU, memory, and connection pool usage
6. 6For Cloudflare 504: optimize backend to respond under 100s or upgrade plan for longer timeouts

Tags

More in 5xx Server Error

Frequently Asked Questions