The most preventable outage: a lapsed domain registration

At midnight the website went down. So did the API. So did email, and the status page you would have used to tell anyone about it. All at once, to the second. That detail is the whole diagnosis: individual services do not fail in perfect unison. A domain does.

The registration had lapsed. Auto-renew had been on for years, which is precisely why nobody worried about it. The card on file expired in March. The renewal failed in April. The registrar sent three notices to the address on the account, which forwarded to a distribution list retired in a reorg. After the grace period, the registrar took the name back.

Why the monitor did not catch it

Here is the cruel part. Once a domain lapses, the registrar usually points it at a parking page, and a parking page returns a perfectly cheerful 200. An uptime check aimed at your hostname can report the site as up while the domain no longer belongs to you. If the monitor resolves fresh it gets NXDOMAIN instead, which plenty of checks treat as a transient blip and retry past. Either way, the one date that mattered was never being watched, because it does not live in any system you operate.

How the team loses the first hour

The symptoms point everywhere except the cause. Everything down at once reads like a catastrophic infrastructure failure, so the team checks the cloud provider, the DNS provider, the load balancer, the recent changes. Someone eventually runs a WHOIS and sees the status flags: the domain is in redemption. The realisation lands with a particular kind of dread, because at that point there is nothing to deploy.

Recovery is not in your hands either. The domain may be redeemable, often for a premium restoration fee, sometimes only after a holding period. Until the registrar acts, you wait. A problem that took one missed payment to create can take days to undo.

It is the most preventable outage there is, and it still takes down serious teams every year.

What monitoring changes

The fix is unglamorous and total: track the registrar expiry date itself, from WHOIS, independent of whatever auto-renew claims, with alerts far enough out that a lapse is a calendar nuisance rather than an emergency. And the date should be owned by the organisation, not by whoever happened to register the domain and has since changed teams.

Crowswatch tracks domain and registrar expiry alongside the rest of your domain health, so the most avoidable outage in infrastructure stays a reminder you act on, not an outage you have to explain.