Uptime monitoring and status pages: 7 things to check before you rely on one
Last updated

An outage is the worst time to find out your monitoring is noisy, blind to a region or silent to your customers. Here are seven things to check before you rely on any uptime monitor and status page, with DEV3 Watchmoor as the worked example.

1. Does one flaky network hop wake you at 3 a.m.?
A single failed request is not an outage. Look for confirmation from more than one region before anyone is paged, and for damping when a monitor flaps. Watchmoor needs n of m regions to agree (default 2 of 3) and groups noisy monitors into one summary.


2. Can it check more than a home page?
Look for keyword, JSON, TCP, SSL expiry and heartbeat checks for cron jobs, not only a status code. The wizard should run the first check straight away.

3. Is the probe itself safe?
A monitor that fetches any URL a customer types can be pointed at your own network. Check that the client resolves the host once, pins that address, re-checks every redirect and refuses private and metadata addresses. Watchmoor does, and an operator blocklist adds more.
4. Does an alert reach someone who acts?
An incident should page the first step of a policy and move on if nobody acknowledges. Look for acknowledge and resolve from a phone, and for a delivery log.


5. Can customers see what you see?
A status page should be calm and fast, with uptime bars, subscribers and maintenance notices, and it should not fall over with the product it reports on. Watchmoor renders pages on the server and caches them for 30 seconds.


6. Is planned work left out of the numbers?
Maintenance windows should pause alerts and be excluded from uptime, and they should show on the page.

7. Can you send a number to a customer?
Look for monthly uptime, downtime minutes, MTTA and MTTR per monitor, exportable.

Limits
Regions are logical: checks run from your install, and extra real regions come from running the probe agent on other servers. Ping is a TCP connect, not ICMP. Email and SMS alerts are simulated until you add a mail transport and an SMS provider, and there are no voice calls. Custom domains need your own TLS. Demo monitors target fictional example hosts answered by a simulated driver. A cron entry is required, and billing runs on a simulated gateway. There are no mobile apps.
The same page after dark — Heather on near-black with the signal colours kept, so green and red mean the same thing at three in the morning.




