Troubleshooting

Website Uptime Troubleshooting System

An outage is a layered problem: DNS, TLS, network, server, application. Diagnose them in order and the cause is usually obvious within minutes. Diagnose them randomly and the same outage burns an hour. The system below is the layered version — built to be run under pressure.

Layer 1 — Is it actually down?

Confirm from an external probe before touching anything. Local DNS cache, captive portal, or ISP-side routing can fake an outage. If the site is up from outside, the investigation moves to your network, not theirs.

Layer 2 — DNS

Fresh lookup from a clean resolver. If the answer is wrong or missing, that's the root cause. If the answer is correct, DNS is cleared as a suspect — move on without revisiting.

Layer 3 — TLS

Check the handshake and expiry. An expired or misconfigured cert is browser-level fatal — the site is effectively down even if the server responds.

Layer 4 — Network and server

Reach the server on port 443 from multiple regions. Failure here points to a routing, firewall, or server-up/down issue. Success here means the problem is in the application layer above.

Layer 5 — Application

5xx responses, slow responses, or specific endpoint failures live here. This is where logs and APM take over from network diagnostics. The point of the layered system is to arrive here only when layers 1–4 are cleanly ruled out.

Recommended next steps

Diagnostic workflow

1
Confirm uptime from outside
Independent reachability check.
Run check
2
Check DNS and TLS in parallel
Cover layers 2 and 3 quickly.
Run DNS lookup
3
Promote to continuous monitoring
So the next outage fires before users notice.
Open monitors
4
Read the uptime monitoring guide
Build the continuous coverage that prevents repeats.
Read the guide

Frequently asked questions

Where should I start when a site is down?

Confirm from outside your network first. If the site is up from elsewhere, the problem is local or regional.

How do I tell DNS vs server problems apart?

If a fresh DNS lookup returns the right record but the server doesn't respond, it's a server/network issue, not DNS.

What if the cert is expired?

Browsers refuse the connection entirely. The server may be 'up' in the technical sense, but the site is effectively down for all users.

What's a 'partial outage'?

Some users can reach the site, others can't — usually DNS propagation, CDN routing, or a single failed region. Multi-region monitoring catches these cleanly.

Should the troubleshooting system be documented?

Yes. Tribal knowledge is the most expensive thing in an outage — a written, layered system saves minutes that translate directly to revenue.