Network monitors sit on their own screen, each row showing the protocol, the host and port, the last response time, and whether the service is currently answering. A database, a cache, a mail server, and a DNS resolver can all sit side by side, and a red row points you at the one piece of infrastructure that's actually down.

01 — Four kinds of check, one monitor type#
A network monitor has a type, and each type answers a different question about a service.
TCP is the workhorse. Open a connection to a host and port: accepted means up, refused or timed out means down. Postgres on 5432, Redis on 6379, SMTP on 587, SSH on 22. Most of what you'll monitor lives here.
UDP is weaker by design. Because the protocol is connectionless, the check mostly confirms the socket opened, and only reports down when it can't be created at all. Don't over-trust it.
DNS monitoring answers a different question: does your domain resolve, and does it resolve correctly? Pick a record type (A, AAAA, MX, CNAME, TXT, or NS), optionally set an expected value, and WatchCron flags a mismatch as down. That catches hijacks, botched migrations, and expired domains that a plain "does it resolve" check sails past.
ICMP sends a single ping. Reply means up, non-zero means down. Simple, but plenty of hosts and firewalls drop ping entirely, so a blocked ping reads as down even when the service is fine. Use TCP where you can.
Configuration: host, port (required for TCP and UDP), DNS record type and optional expected value, timeout from 1 to 60 seconds. Check interval depends on your plan: 5 minutes minimum on Free, 1 minute on Starter, 30 seconds on Pro and Business. Defaults are 10-second timeout, 5-minute interval.
02 — What each protocol checks, and when it counts as down#
Protocol | What it checks | Counts as down when |
|---|---|---|
TCP | A connection to | Connection refused, or it times out |
UDP | A UDP socket to | The socket can't be created (a weak signal — see limits) |
DNS | A record of the chosen type resolves, and matches the expected value if set | No record returns, or the resolved value doesn't match what you expected |
ICMP | A single ping gets a reply | The ping returns a non-zero result |
TCP covers 90% of what teams actually monitor. If you're not sure which type to pick, start there.
03 — Catch DNS that resolves to the wrong place#
Resolving is not the same as resolving correctly. Set an expected value on a DNS monitor, and WatchCron compares the record it gets back against the one you told it to expect, flagging the monitor down when they don't match. That catches the failure modes a plain "does it resolve" check sails past: a DNS hijack pointing your domain at someone else's server, a migration that updated the wrong record, or an expired domain quietly handed to a parking page. The resolved value is saved with the check, so you can see exactly what it returned instead.
04 — What's tracked, and what you'll see#
Response time in milliseconds gets recorded on every check, regardless of protocol. On a monitor's detail view that becomes a 24-hour response-time chart, a 30-day uptime history, and the last fifty checks shown as bars, alongside a panel with the monitor's type, host, port, interval, and timeout.

05 — What teams check with port monitoring#
The services that actually take a site down are rarely the website itself. It's almost always something behind it:
Databases on their native ports, like PostgreSQL on
5432, MySQL on3306, or MongoDB on27017.Caches and queues that the app leans on, such as Redis on
6379, Memcached on11211, or a RabbitMQ broker.If mail has to keep flowing, an SMTP server on
587or465and an IMAP server on993are worth a TCP check each.SSH on
22, so you know a box is reachable before you need to get into it.DNS resolvers and your own records, checked against an expected value so a mispointed domain surfaces immediately.
06 — Where alerts go#
Network monitors share the same alert routing as every other check type. On Free you get 3 channels (pick from email, Slack, Telegram, webhooks, or any other supported type); Starter raises that to 10, Pro to 25, and Business removes the cap. SMS opens on Pro. Voice calls and WhatsApp are Business-only, for the infrastructure where a missed page costs real money. Every alert can be acknowledged straight from the notification without logging in, and any monitor can be paused during maintenance so it doesn't wake the team at 3 AM.
07 — What port monitoring won't do#
A connection test is a blunt instrument on purpose, and it's worth knowing where the bluntness shows.
A TCP check confirms the port accepts a connection, not that the service behind it is healthy. A database that accepts connections but rejects every login still reads as up, so pair a TCP check with an application-level check when "reachable" and "working" can diverge.
UDP is unreliable by nature, not by implementation. Because there's no connection to establish, the check mostly confirms the socket opened, and it reports down only when the socket can't be created. Don't lean on a UDP monitor for anything you'd want to be paged about.
ICMP is a single ping with no packet-loss analysis, and plenty of hosts and firewalls drop ping entirely. A blocked ping reads as down even when the service is perfectly fine, so use TCP against a real port where you can.
A few more edges worth naming: checks run from a single location with no multi-region confirmation, and there's no retry before an alert, so the first failed check notifies you, which can get noisy on a flaky network. Response time is tracked but not something you can alert on with a threshold yet. And a DNS check compares the first record returned, so a domain with several A records is judged on the first one. We'll add round-robin support when the demand is there. For now, the honest answer: port monitoring tells you a door is open. Whether anything useful is behind it is a different question.
For services that do speak HTTP, uptime monitoring checks the full response, status code, and certificate rather than just the open port. For scheduled jobs that report in on their own, that's cron job monitoring. If you're mapping out which layer each tool covers, the comparison of cron, server, and APM monitoring lays it out, and setup details are in the docs.
Point WatchCron at a host and port, know the moment a database, cache, or DNS record stops answering. Free plan: up to 5 port monitors, 3 alert channels, 5-minute intervals. No credit card. Paid plans raise every limit and drop the interval to 30 seconds.08 — FAQFrequently asked#
Uptime monitoring checks the full HTTP response of a website or API, including status code, content, and certificate. Port monitoring checks whether an infrastructure service, like a database or cache, is accepting connections at all. They sit at different layers and pair well: one watches the front door, the other watches the services behind it.
Four: TCP, UDP, DNS, and ICMP (ping). TCP is the most reliable and the one to default to.
Yes. Set an expected value on a DNS monitor, and WatchCron flags it down when the resolved record doesn't match, which catches DNS hijacks, botched migrations, and expired domains.
It confirms the port is accepting connections. It won't catch a database that's up but rejecting logins or returning errors, so pair it with an application-level check when you need that distinction.
UDP is connectionless, so "up" is a weak signal that mostly means the socket opened. ICMP is a single ping that many firewalls block outright. Where you can, monitor a real TCP port instead.