All requests
ShippedPublic request — anyone can read this thread

Tell me when a client site is down — and when someone quietly changed it

RDRyan DelgadoWeb Agency Partner ·

We are four people looking after 43 sites we mostly did not build. Two things keep going wrong, and we learn about both of them from someone else.

The first is that a site goes down. A certificate expires, a host runs out of disk, a plugin update takes the checkout with it. The client notices, calls us, and by then it has been down long enough for them to have counted the hours.

The second is quieter and worse. Somebody with edit access changes a page — a price, a phone number, a legal line we wrote with a lawyer — and nobody tells us. We find it weeks later, by accident, usually while looking for something else.

What we actually need

  • An availability check per site: is it answering, is it answering fast, is the certificate about to expire.
  • A change check per page: what is on this page that was not here yesterday.
  • One alert stream for both. We do not want to run two tools and reconcile them.
  • A way to mute the noise — the parts of a page that change on every load and mean nothing.

Why it matters

Both failures cost us the same thing, which is the client's belief that we are watching. A site that was down for six hours on a Sunday is a refund conversation. A price that was wrong for three weeks because someone edited it and did not say is a harder one, because we are the people who were supposed to notice.

What we do today

A free uptime pinger that emails one address, and nothing at all for changes except memory. When a client asks "did this always say that?", the honest answer is that we do not know.

DetectorAutomationIntegrations
Progress
  1. Filed
  2. In review
  3. Approved
  4. In progress
  5. Shipped
96 people need this too
4 commentsSorted by oldest
RD

Ryan Delgado, Web Agency Partner ·
Scale, so it is clear what we are asking for: 43 sites, three of them with a checkout, the rest brochure sites on four different CMSes we do not control. Six people have edit access on some of them and none of those people work for us.

IN

Incapon ·
These are two checks that feel like one to you, and we think you are right. One week, timeboxed, ending on the 5th: an availability check on a short interval, a content check on a slow one, and a single alert stream out of both so you are not subscribed to two products. The part we will measure rather than promise is the noise — a brochure site with a rotating hero will look "changed" every hour unless we can tell layout drift from an edit.

RD

Ryan Delgado, Web Agency Partner ·
Two days in and the down alerts already paid for it — a cert expired on a Saturday and we knew before the client did. The change alerts were the mess you predicted: one site fired every run because of a cookie banner that writes a new id into the DOM.

IN

Incapon ·
Completed, one week from the call. Availability checks every minute from two regions, so a single failed probe does not page anyone; content checks hourly with the noisy regions of a page excludable by selector, which is what fixed your cookie banner. Both land in the same alert, tagged as "down" or "changed", and both go out on Slack as well as email. One limit: a site behind a login is out of scope for now — we check what a visitor can see.