A client’s site is down: what to do, in what order
Last updated: 9 November 2026
The phone rings: “the site is not working”. Without a method, you rush onto the server at random. With a method, you fix it faster, and avoid making things worse.
1. Confirm the outage
Open the site over 4G, not from your wifi. The problem may be on the client’s side. If external monitoring is running, look at what it says: since when, which error. It is the client’s first question and you will already have the answer.
2. Diagnose, from the simplest to the rarest
- Has the certificate expired?
- Is the domain name still active and pointing to the right place?
- Is the server responding? Disk full, memory saturated, process stopped.
- Is the container or service started?
- Was there a deployment or an update just before?
In most cases, the cause is in this list. A full disk is surprisingly common: logs grow and nobody looks.
3. Restore first, understand later
If a rollback is possible, do it. Hunting for the root cause while the site is down costs more than restoring and then investigating with a clear head.
4. Warn the client
A short message: the site is unavailable, you are on it, estimated time. Then a second one when it is fixed. A client forgives an outage; they do not forgive silence.
5. Write down what happened
Three lines: what happened, what caused it, what you changed so it does not come back. In StatOps, you write it on the incident; the client reads it on their status page and finds it again in the month’s report. If the same incident comes back, the tool reminds you what was done last time.