In our production environment using Traefik v3.1.1, we have observed occasional DNS resolution timeouts (error: dial tcp: lookup internal domain name: i/o timeout), which occur more frequently in environments with poor network connectivity and happen almost daily.
We understand that net/http performs DNS resolution for each new HTTP connection (without DNS caching), so network fluctuations may cause DNS resolution timeouts. However, network issues are often challenging to troubleshoot.
We noticed that fasthttp supports DNS caching, and according to official fasthttp test data, it is significantly faster than net/http. This seems like a potential solution for our scenario.
We also noted that the official PR at Introduce a fast proxy mode to improve HTTP/1.1 performances with backends by kevinpollet · Pull Request #11122 · traefik/traefik · GitHub introduced fasthttp. We would like to inquire whether the Traefik team has conducted load testing for this change, what performance improvements have been observed for HTTP/1.1 proxy scenarios compared to net/http after adopting fasthttp, and whether specific load testing data is available for reference.