A website never announces its own outage. It just goes quiet, and the first person to notice is usually a customer in the middle of placing an order. Website uptime monitoring flips that order: the system notices first, you fix it, and the customer never knows anything happened.
What is website uptime monitoring?
Uptime monitoring means an external system calls your website on a short cycle, behaving exactly like a real visitor, and records the outcome. If the site fails to respond or returns the wrong content, the system opens an incident and alerts whoever is on call.
Every check result is stored as uptime history. That history answers questions gut feeling never can: how many times did the site go down this month, how long did each outage last, and is your real uptime 99.9 percent or closer to 98.
Why not just wait for customers to report it?
Because the cost of downtime is not spread evenly. The first minute is nearly free. Once a customer becomes your monitoring system, you lose far more than repair time:
- You lose the sale on the spot. A buyer who hits an error page rarely retries the same day.
- You lose trust that outlasts the outage. One “hey, is your site down?” message from a client costs more reputation than hours of silent downtime.
- You lose the data to improve. Without check history you cannot see that incidents cluster around a certain hour or follow a certain deploy.
Watching your website 24/7 by machine is cheaper than every alternative, including the most popular one: hoping.
What kinds of checks does AgentWatch run?
A URL stays alive thanks to several layers working together, so AgentWatch offers four check types to point at whichever layer breaks:
| Check type | What it verifies | What it catches |
|---|---|---|
| HTTP | Status code and a keyword in the response body | Broken deploys, 500 errors, blank pages, hung backends |
| TCP | Whether a port is open and accepting connections | Databases, mail servers and internal services gone silent |
| Ping | Whether the host answers ICMP | Machine powered off, network trouble, VPS suspended by the provider |
| SSL | Certificate expiry date | Forgotten renewals, broken auto-renewal |
The keyword match on HTTP checks deserves a highlight. Plenty of sites die in the “returns 200 but shows an error page” style. Require a string that must appear in the response, your brand name in the footer works well, and this whole class of silent failure gets caught too.
How short should the check interval be?
The interval decides how early you find out. At 5 minutes, the worst case means the site has been down for nearly 5 minutes before the next check runs. At 30 seconds, that gap effectively vanishes.
Practical guidance by site type:
- Landing pages, blogs, brochure sites: 1 to 5 minutes is plenty.
- E-commerce, payment flows, integrated APIs: aim for 30 seconds to 1 minute.
- Low-traffic internal tools: 5 minutes, and prioritize the SSL and TCP checks.
On AgentWatch the minimum interval is 5 minutes on Free, 1 minute on Starter, and 30 seconds from the Team plan up.
Are external checks alone enough?
Enough to know the site is alive, not enough to know why it died, and definitely not enough to see trouble building up. An external check cannot see a disk at 2 percent free space, memory draining away or a container restarting itself every 10 minutes.
That is why AgentWatch pairs uptime checks with an agent that runs inside the server: CPU, RAM, disk, service status and container health appear on the same dashboard as your check results. When the site goes down, you see the inside context immediately instead of starting to guess. The server monitoring page goes deeper on that half of the picture.
How long does 24/7 setup take?
About 3 minutes, with nothing to install:
- Create a free account with your email.
- Add an HTTP check with your site URL, pick an interval and a keyword if you want one.
- Connect Zalo alerts or email so incidents come find you.
From then on every check is recorded, incidents open and close automatically, and you can publish a public status page on its own URL to share uptime with your customers.
What does uptime monitoring cost?
The Free plan includes 10 service checks at a 5-minute interval with 7 days of history, enough for a few websites and their key APIs. Paid plans start at 49,000 VND per month with a 1-minute interval, while Team at 149,000 VND and Business at 399,000 VND run 30-second intervals with 100 and 300 checks respectively. Full details live on the pricing table, with the product overview on the homepage.
During the beta, every new account gets the Business plan free. Your website is running right now, and the only real question is: if it stops, who finds out first, you or your customer?