WordPress 504 Gateway Timeout: The 7 Real Fixes (2026)

A 504 is a waiting problem. The gateway knocked, PHP never answered within the timeout - usually ~30-120 seconds. Either something is genuinely too slow (a query, an import), or PHP can't answer at all (workers saturated, server overloaded). Here's how to tell which, then fix it.

1. Refresh, Then Check Scope

One reload can clear a transient spike. Then scope it: whole site timing out = server/PHP layer problem. One page or one operation (import, export, report) = that operation is too heavy - skip to fix 4. Only via Cloudflare = fix 2.

2. Rule Out the CDN

Cloudflare has a ~100-second hard timeout to the origin - if your server takes longer, Cloudflare returns its own branded 504 even when the server eventually finishes. Enable Development Mode or pause Cloudflare to test the origin directly. If direct works, either speed up the operation or move it off the web request entirely (WP-CLI, cron).

3. Deactivate Plugins

A hanging plugin holds a PHP worker until the gateway gives up. FTP: rename wp-content/plugins to plugins_off, reload. Site responds? Reactivate one at a time. Classic offenders: anything doing external API calls per pageview, backup/migration plugins mid-job, and security scanners running deep scans.

4. Find the Slow Query

Page-specific 504? Install Query Monitor on a staging copy (or briefly on production) - it lists every database query with timing. You'll usually find one query eating seconds: an unindexed meta query, a WooCommerce report over 50k orders, a "related posts" query scanning wp_posts unindexed. Fix the query, add an index, or cache the result - don't raise timeouts to accommodate it.

5. Split Heavy Operations

Imports, exports, and bulk edits that 504 aren't fixable by waiting longer - the gateway timeout is a wall. Split them: import 500 rows at a time instead of 50,000, use WP-CLI (wp import runs without HTTP timeout), or run heavy jobs as real cron tasks. This is why managed hosts offer WP-CLI - command-line PHP doesn't have a gateway waiting on it.

6. Timeouts: The Last Lever, Not the First

If the operation is legitimately long and can't be split, raise the ceilings - all of them, or the lowest still wins:

• -PHP: max_execution_time = 300 in php.ini (details in our execution time fix)
• -Nginx: proxy_read_timeout 300; (VPS only)
• -Apache: Timeout 300 in server config

On shared hosting you control none of these - which is the point: if your workload needs 5-minute web requests, it's outgrown shared hosting.

7. Server Overload and DNS

Check cPanel > Resource Usage for CPU/IOPS faults at the 504 timestamps - saturated resources timeout exactly like this. Recent DNS change? Requests may hit the old server where PHP is dead. And if clean diagnostics still show timeouts under modest traffic, that's the hosting telling you it's undersized.

Related: 502 bad gateway fix - 503 service unavailable - max execution time fix - database optimization.