Focus on Client-Side Errors in a Monitoring System

A server can report normal latency while some users cannot reach it. Client and external observations help reveal failures before the request enters the service’s own monitoring boundary.

Client-side errors

HTTP error counts are useful server-side signals, but they only cover requests for which a response was produced and recorded. Also inspect timeouts and user-visible correctness; a problem need not appear as HTTP 500.

A drop in server traffic can suggest a reachability problem, but demand also varies naturally. A failure affecting one provider or region may disappear in fleet totals. Compare client reports and external probes by region, network, and time to narrow the affected population.

Several factors can prevent clients from reaching the server, including:

• DNS resolution failures.

• Routing failures between the client and the service provider.

• Third-party infrastructure failures (e.g., middleboxes or CDNs).