4 things writing an HTTP prober in Rust taught me

Short summary

The author built an uptime prober in Rust and dropped reqwest for raw hyper to expose timing details a normal HTTP client hides. Four key lessons emerged: disable connection pooling so cold-path timing is measured accurately, make errors carry per-phase timing data so failures are diagnosable, block SSRF by filtering resolved IPs rather than hostnames, and downcast errors to produce exact failure reasons instead of matching on message strings. Each lesson is broadly applicable beyond monitoring.

• •Disable connection pooling when timing requests so cold-path costs aren't hidden
• •Enrich errors with per-phase timing (DNS, TCP, TLS) so failures remain diagnosable
• •Block SSRF by filtering resolved IPs, not hostnames, to defeat DNS rebinding

Generated with AI, which can make mistakes.