- Home
- Monitoring
- UDP port monitoring
UDP monitoring for DNS, NTP and other services that answer datagrams
Tenium sends a probe datagram to your host and port and expects any reply. It speaks real DNS on port 53 and real NTP on port 123, and works for other services that answer anything at all.
- From $7 a month on Starter
- Free plan: 30 monitors for HTTP(S), keyword, ping and port
- No credit card for Free
- Amsterdam19msreply
- London28msreply
- ProbeDNS A query
- Reply19 ms
- Port53
Illustrative example
Check it now, in your browser
No account. UDP port monitoring: type a target and see roughly what it looks at. Nothing you type is stored.
A browser cannot open raw TCP, UDP or ICMP connections, and the public check API only speaks HTTP(S). So instead of a fake result, here is the command a monitor automates: it runs it on a schedule from several regions and confirms failures.
What a udp port monitor looks at
Protocol-aware probes
Port 53 gets a genuine DNS A query, port 123 a genuine NTP client packet. Any other port gets a short CRLF datagram.
Any reply is up
The check passes when a datagram comes back before the timeout. The content of the reply is not compared against anything.
Closed port detection
If the host answers with ICMP port unreachable, the check fails at once with "UDP port closed" instead of waiting for the timeout.
Silence is down
No answer within the timeout (capped at 15 seconds) is a failure, so a service that never replies to unknown packets cannot be monitored this way.
UDP port monitoring: what an alert looks like
The same event arrives as an e-mail, a Telegram message or a Slack card. This is the wording the channels send; the host, times and numbers are made up.
ns1.clienta.com:53
ns1.clienta.com:53 · UDP port closed · confirmed from Amsterdam (AMS), London (LON)
| Error | UDP port closed |
|---|---|
| Confirmed by | Amsterdam (AMS), London (LON) |
| Since | 14:02:11 UTC |
| Target | ns1.clienta.com:53 |
| Next step | Check the server / hosting panel. Acknowledge so the team knows you are on it. |
AcknowledgeOpen incident
Tenium · checked from Amsterdam · 14:02 UTC
Tenium alerts
🔴 DOWN · ns1.clienta.com:53 · UDP port closed · confirmed from Amsterdam (AMS), London (LON)
reply /ack or acknowledge · open incident
TTenium APP 2:02 PM
DOWN · ns1.clienta.com:53
ns1.clienta.com:53 · UDP port closed · confirmed from Amsterdam (AMS), London (LON)
- Error: UDP port closed
- Confirmed by: Amsterdam (AMS), London (LON)
- Target: ns1.clienta.com:53
AcknowledgeOpen incident
Check the server / hosting panel. Acknowledge so the team knows you are on it. · checked from Amsterdam
UDP port monitoring: e-mail, Discord and Google Chat work on every plan; Slack, Telegram, Teams and Mattermost start on Starter. There are no SMS or voice-call alerts.
Response time with a threshold
Every check stores how long the response took. UDP port monitoring: drag the threshold and see when a slow-response alert would have fired.
How fast you find out
Pick a plan to see its check interval and how long a failure takes to reach you, including the confirmation round.
UDP port monitoring: a failure lands at a random moment between two checks, so on average you wait half an interval, and the worst case is a whole interval plus the confirmation window. A single failing region is not enough: all two regions must agree before an incident opens. The other regions are asked straight away, and we allow up to 60 seconds for the answers.
How to set up udp port monitoring in 30 seconds
Enter the target
Add the host and set the UDP port number, and Tenium fills in a sensible timeout and default confirmation settings.
Pick an interval
Fastest check interval: 30 s on Pro and Agency, 1 min on Starter, 5 min on Free. Slower intervals are available on every plan.
Choose alert contacts
Choose who is told. Email, Discord and Google Chat work on every plan, Slack, Telegram, Teams and Mattermost start on Starter, webhooks on Pro.
Save
The first check runs within seconds. If a check ever fails, other regions re-check it before an incident opens, so one bad route does not page you.
What each error means
This is the text an incident shows in its Error line, the usual cause, and the first thing to look at.
CLOSEDUDP port closed
The host sent back ICMP port unreachable, so the check fails at once instead of waiting for the timeout.
Shown as: UDP port closed
SILENCENo reply within 5 s
UDP gives no handshake, so silence is the only signal. A service that never answers unknown packets cannot be monitored this way.
Shown as: No reply within 5 s
ENOTFOUNDDNS lookup failed for ns1.clienta.com
The hostname does not resolve. Fix that first.
Shown as: DNS lookup failed for ns1.clienta.com
NO_PORTNo port configured
A UDP monitor needs a port number.
Shown as: No port configured
UDP port monitoring by plan
| Free $0/mo | Starter $7/mo | Pro $29/mo | Agency $59/mo | |
|---|---|---|---|---|
| Available | — | ✓ | ✓ | ✓ |
| Monitors | 30 | 100 | 300 | 1000 |
| Fastest interval | 5 min | 1 min | 30 s | 30 s |
| Regions | 1 | 2 | 2 | 2 |
| Custom headers | — | — | ✓ | ✓ |
| Response-time alerts | — | — | ✓ | ✓ |
| Escalation to a second contact | — | — | ✓ | ✓ |
| Status pages | 1 | 3 | 10 | Unlimited |
This check starts on the Starter plan. HTTP(S), keyword, ping and port checks are on every plan.
Other monitor types
UDP is silent by default
TCP has a handshake, so "connected" is a clear signal. UDP has nothing like it: a datagram goes out and, unless the service chooses to answer, nothing comes back. A monitor therefore has only one positive signal, a reply, and two ambiguous outcomes. Tenium's UDP monitor sends one probe datagram from the checker, waits for the timeout and classifies what happened. It is honest about what that can prove, and this page is equally honest about the services it cannot cover.
The three outcomes
Every UDP check ends in one of these states:
- A datagram comes back before the timeout: up. The response time is how long it took, and the size of the reply is saved in the details. The content is not inspected.
- The target host replies with an ICMP "port unreachable": the check fails immediately with "UDP port closed". This is the one negative signal UDP offers, and it arrives quickly.
- Nothing comes back before the timeout: failure after waiting. Silence is indistinguishable from a dropped packet, a firewall, or a service that ignores unknown input, so it is treated as down. The timeout you configure is capped at 15 seconds.
The probes: real DNS, real NTP, and a polite poke for the rest
Tenium chooses the payload by port. On 53 it sends a standard DNS query asking for the A record of example.com, so a working resolver or authoritative server replies; even a REFUSED answer counts as proof of life. On 123 it sends a 48-byte NTP client request, which any NTP server answers. On 161 it sends a short text string that is not a valid SNMP request, and on every other port a CRLF datagram. You cannot supply your own payload or check the reply against an expected string.
That is a real limit, and it decides where the monitor fits. It is excellent for the two protocols above and for any service that answers or rejects arbitrary input. It is a poor fit for protocols that demand a precise handshake before they will say a word.
Services that will look down even when healthy
Several well-known UDP services are designed not to answer strangers. WireGuard ignores any packet that is not a valid authenticated handshake, which is a security feature. OpenVPN configurations with TLS authentication drop unauthenticated packets the same way. Syslog receivers on port 514 only listen. Game servers usually expect their own query format: Source-engine games want a specific information request, Minecraft Bedrock a RakNet ping, voice servers their own protocol. SNMP agents drop packets that are not valid SNMP, and the generic probe is not. QUIC on 443 expects a proper initial packet.
Do not add these as UDP monitors and then wonder about the permanent red. Use the alternatives below.
Better options for silent services
The reliable pattern is to monitor from the inside out. For a WireGuard endpoint, run a small script on a client that checks the latest handshake time and calls a cron heartbeat URL only when the tunnel is healthy; if the ping stops, you are alerted. For a syslog collector, a minutely job that checks whether the log file has grown in the last five minutes and then calls a heartbeat catches a dead receiver. For game servers, many expose a TCP query or RCON port, a status web page or a plugin that reports player counts; use a port monitor or a keyword monitor on that. In all cases a ping monitor on the host covers the machine itself.
Firewalls, plans and testing
Allow inbound UDP from the addresses on the checker IP list on the port you monitor. Stateful firewalls handle the reply automatically. Private addresses are refused before any packet is sent. UDP monitoring starts on the Starter plan and counts as one monitor against your allowance of 100, 300 or 1000.
One quirk: the "test check" button in the editor is disabled for UDP monitors, so you save first and read the first result on the monitor page. If the first run says "No UDP reply", try the same probe from a shell to make sure the service answers at all: dig @your-resolver example.com A for port 53, or ntpdate -q host for port 123. If those work from outside your network, the monitor will too.
UDP port monitoring questions
Can I send my own payload and expect a specific reply?
Not today. The probe is chosen by port and any reply is accepted. If you depend on a protocol-specific handshake, monitor through a TCP port, a status page or a heartbeat instead.
Why is my DNS server marked up even though it refuses queries?
A REFUSED or SERVFAIL response still proves the server is alive and answering UDP 53. To check that names resolve correctly, add a DNS monitor with expected values.
Can I monitor a Minecraft or other game server?
Java Edition servers listen on TCP, so a port monitor works. Bedrock and most Steam-based games use UDP with their own query formats, which the generic probe does not speak.
What timeout should I use?
Three to five seconds suits resolvers and NTP. Longer values only delay the verdict, and anything above 15 seconds is reduced to 15 anyway.
Try udp port monitoring on Starter
This check starts on the Starter plan at $7 a month. The Free plan, no card needed, covers 30 monitors with HTTP(S), keyword, ping and port checks.