What is monitoring? Explained with smoke alarms
Monitoring constantly measures how your app behaves, and alerts call a person when something breaks. Explained with smoke alarms that go off before a fire spreads.
- Author
Umair Mateen- Published
- Updated
- Reading time
- 5 min
- Category
- DevOps
In one sentence: Monitoring means constantly measuring how your website or app is behaving, and alerts are the automatic warnings that call a person when those measurements show something is broken, or about to break, so it can be fixed before many customers notice.
The everyday picture
Think about the smoke alarms in an office building. Nobody stands in every room all night sniffing the air. Small sensors on the ceiling do the watching, every minute of every day.
Most of the time they stay quiet. When smoke passes a certain level, the alarm sounds and someone is called. A small fire in a bin gets put out before it spreads to the whole floor.
Software works the same way. Your website or app is the building. Monitoring is the sensors. An alert is the alarm that sounds and calls the right person.
Monitoring measures your app all the time, like smoke sensors on a ceiling. When a rule says the numbers are serious, an alert calls the person on duty, who fixes the problem before it spreads.
How it works, step by step
- Your system runs (the building is open for business). Customers visit your website, place orders and log in, all day and night.
- Monitoring measures it constantly (the sensors on the ceiling). Software collects numbers such as how many visitors arrive, how long each page takes to load and how many requests fail. Google's engineering guide on running large systems calls four of these the "golden signals": speed, traffic, errors and how full the system is.
- A rule decides when it is serious (the smoke level that sets off the alarm). Someone writes a rule like "if more than a few orders fail in five minutes, raise an alert". Below that line, the numbers are simply recorded.
- The alert goes out (the alarm sounds). The warning is sent by email, to a chat tool or to an on-call phone app. Tools such as Prometheus's Alertmanager group many related warnings into one message, so one problem does not cause a hundred phone calls.
- A person responds (someone comes to put out the fire). The person on call checks what customers are seeing, fixes the cause and the numbers return to normal.
Try it · Try the dial
Turn up the failures and watch the alarm
Drag the dial and watch what changes.
Below the line. The numbers are simply recorded: a few failures can happen on any busy day.
Where the picture stops
- A smoke alarm only knows about smoke. Monitoring watches many different things at once, from page speed to failed payments, and keeps a history of them. That history shows slow trends, like a system that gets a little fuller every week.
- Smoke usually means trouble. A number on its own may not. A busy morning can look like a problem. Good alert rules are tuned so they only fire when customers are really affected, otherwise people start ignoring them.
- Monitoring can also test from the outside. Some checks act like a visitor knocking on the front door to see if anyone answers. A ceiling sensor cannot do that.
Why it matters for your business
- You hear about problems before your customers tell you. A broken checkout found in minutes costs far less than one found the next morning through complaints.
- Problems get fixed faster. Alerts point the team to what is broken, and the recorded numbers help them see why.
- Fewer false alarms means a team that still listens. Google's guide warns that when alerts fire too often, people start to skim past them and can miss a real one. It says "Every page should be actionable", meaning each call to a person should need them to do something.
- You can plan ahead. Trends show when you are running out of room, so you can grow before things slow down.
An example
Imagine a small online shop that sells kitchen tools. Its website records how many payments succeed and fail every minute.
One evening, a change at its payment provider starts rejecting card payments. Within a few minutes, failed payments pass the line set in the alert rule. The developer on call gets one message on her phone, not dozens, because related warnings are grouped together.
She sees the failures started at the same time as the provider's change, switches to the backup payment option and tells the provider. A handful of customers were affected, not a whole evening of orders. Without monitoring, the shop might only have noticed the next day, when sales looked oddly low.
Frequently asked questions
What is the difference between monitoring and alerting?
Monitoring is collecting and recording the numbers all the time. Alerting is the part that watches those numbers and calls a person when a rule says something needs attention now.
What should a small business monitor first?
Start with what customers feel: is the website up, how fast does it load and are important actions like logins and payments failing? Google's guide suggests alerting on these symptoms rather than on every internal detail.
What is alert fatigue?
It is what happens when people get so many warnings, many of them false or unimportant, that they stop paying attention. The fix is fewer, clearer alerts that always need a real response.
Do I need someone awake 24 hours a day?
Not always. Many teams send only serious alerts to an on-call phone at night and leave smaller issues as tickets for the next working day.
Related explainers
- What is scaling? Explained with checkout lanes at rush hour
- What are backups? Explained with a spare key at a friend's house
- What is cloud hosting? Explained with a serviced office
The DevDrift sets up cloud hosting with monitoring and alerts that reach the right person: see how.
Sources
Could you explain this to a colleague now?
Mark it and posts that mention it will show a tick. Saved in this browser only.