We currently have some really slow responses in our backend, but for quick-fixing we would need to increase the timeout period above 60 seconds. We thought we can configure this in the Host section of the config for the specific backend (first byte: 120000ms, connection: 1000ms, between bytes: 10000ms), but it doesn’t seem to have an effect. We get a 503 Backend Read error exactly after 60 seconds. Doing the request against the backend directly works and yields a response. Any tips?
For HTTP requests this should be working for you as configured, but this sounds like very long held connections over TLS. Fastly uses a separate tool for terminating TLS in our Points of Presence. This has a hard time out of 60 seconds for connections to ensure that these cannot be exhausted, maintaining a reliable service for all our customers.
If this is affecting all your requests, I’d recommend configuring a failure page for requests which are timing out while you look into the causes. If the slow requests are a limited subset of your requests like report generation, I’d recommend that the could bypass the CDN as reports are commonly user-specific and unlikely to see much benefit of our caching.
If you need a more detailed reply, please contact email@example.com and we can dig into your specific needs.
Hej Jonathan! Thanks for the answer above! We already had contact via firstname.lastname@example.org and we solved the case by the workaround you mentioned, hitting the backend directly for those sort of requests.
Thanks for all the help!