Overview
Establish the affected hostnames, regions and times before adjusting timeouts or retries. Without scope and samples, configuration changes can hide the actual cause.
Capture a reproducible sample
Capture the time and time zone, URL, method, status, error details and any correlation identifier. Record recent deployment, certificate and access-control changes.
Check the connection chain in order
Check origin DNS and ports, then firewall access, the TLS chain and hostname. The origin Host header and SNI must match the intended virtual host configuration.
- Test the same origin endpoint from a comparable environment
- Use a health-check path with meaningful responses
- Do not leave certificate verification disabled after diagnosis
Separate slowness from application rejection
After connecting, the application may still time out, deny access, rate-limit or route incorrectly. Correlate origin and edge records to determine whether the request arrived and where it ended.
Retries and fallback must respect application semantics
Retries carry different risks for reads and writes. Check whether repetition could create duplicate orders or actions before setting retry conditions, limits and failover behavior.