SteadyNode

Nginx HTTP 504

What causes an Nginx 504 Gateway Timeout?

Understand Nginx 504 upstream timeouts and inspect slow application work, databases, and downstream dependencies.

Decode a CDN or gateway error

Paste a captured HTTP response or enter a status and headers. Only the status and response headers are sent; common credential headers are filtered first.

Supports raw headers, full HTTP responses, JSON with status/statusCode and headers, cURL -v/-i response output, and HAR exports. HAR selects a server error first. cURL commands are never executed; request headers, response bodies, and common credential headers are excluded.

Find the cause of the next outage sooner

During an outage, every second counts. SteadyNode's uptime engine translates CDN errors into likely root causes on every check.

No credit card required.

Start for free

Nginx 504: cause and troubleshooting

The upstream service did not respond to Nginx before its configured gateway timeout.

What to check next

Trace the request through application and dependency logs, reduce slow synchronous work, and tune proxy timeouts only when the longer request is expected.

Use the diagnostic code and provider headers as clues, then confirm the request against origin logs and deployment telemetry. A status code can be generated by a proxy, load balancer, function platform, or origin application; the same number does not always have the same cause across providers.

Keep a CDN error from becoming a long outage

Record the request identifier, timestamp, affected hostname, and deployment when investigating an intermittent failure. SteadyNode checks public endpoints on a schedule and translates recognized upstream errors into concise diagnostic context, helping teams identify whether a failure is at the edge or origin.

Frequently asked questions

What does Nginx 504 mean?

Nginx connected to an upstream or tried to, but did not receive a response before the configured timeout.

Should I increase the Nginx proxy timeout?

Only increase it when the workload is expected to take longer; first measure and reduce application or dependency latency.