Guide
Why a website can be down for some people but not others
When a site works for you but not a colleague, the site itself is often fine — the path to it isn't.
Published
When a website works for you but fails for someone else — or the other way around — the server is often perfectly healthy. What differs is the route between each person and that server. Three things cause most split outages: DNS caching (different people hold different, sometimes stale, address records), regional network problems (a route or data centre near one visitor is failing while others are unaffected), and local issues (a device, browser cache, VPN, or ISP resolver). Because every visitor takes a slightly different path, “is it down?” rarely has one yes-or-no answer. The reliable way to tell is to check the site from several locations at once, instead of trusting a single browser.
The site is one thing; the path to it is many
Loading a page is a chain of hops: your browser asks a DNS resolver for the site's IP address, your network routes packets to that address, and the server answers. Two people in different cities, on different providers, with different cached data, can take almost entirely different chains to reach the same domain. If any hop on your chain is broken while someone else's chain is clean, the site is genuinely “down” for you and “up” for them — both observations are correct.
DNS caching: everyone can hold a different answer
A domain name has to be translated into an IP address, and that translation is cached at many layers — your browser, your operating system, your router, and your internet provider's resolvers. When a site moves to a new server or fails over to a backup, the new address does not reach everyone at the same moment. It propagates gradually as each cached copy expires.
How long that takes is controlled by the record's TTL (time to live). A low TTL means caches refresh quickly; a high one means some people keep using the old address for hours. During that window, visitors whose cache still points at the old, dead server see an outage, while visitors who already have the new address load the site normally. Nothing is wrong with the site — people are simply looking at two different copies of the truth.
Regional outages: down in one place, up in another
Most sites of any size are served from more than one location — a content delivery network (CDN) or multi-region hosting sends each visitor to the nearest edge. If one edge, one data centre, or the network path leading to it is failing, only the people routed through it are affected. Everyone else is served by a healthy location and never notices.
The same thing happens at the level of the internet's plumbing. A peering or backbone problem between two networks can black out one provider's customers while every other provider reaches the site fine. From the outside it looks random; underneath, it's strictly geographic.
It might be just you (or just them)
Plenty of “the site is down” reports are local. Common culprits:
- A stale browser cache or a broken cookie for that specific site.
- A failing DNS resolver, or an outage at your ISP's DNS specifically.
- A VPN, proxy, or corporate firewall routing you somewhere unusual.
- An ad-blocker or Pi-hole that is blocking part of the page.
- A wrong device clock, which can break the site's certificate check just for you.
Quick self-checks: load the site on mobile data instead of Wi-Fi, try a private/incognito window, or ask someone on a different network. If it works anywhere else, the problem is on your side of the chain, not the site's.
How to tell what is really happening
The single most useful move is to check the site from several places at once. If it responds from multiple regions but not for you, the fault is local — your network, DNS cache, or ISP. If it fails from everywhere, it's a genuine outage on the site's end, and refreshing won't help.
That is exactly what an outside check is for. To be honest about method: our check works at the HTTP layer (it actually requests the page), not by sending ICMP pings, and it runs from our servers — a vantage point outside your own network — so you can separate a real outage from a problem that is local to you. Confirming a site from several regions at once is rolling out.