Server monitoring tools that watch every host from the outside
UptimePilot pings your servers, opens the ports your services listen on and listens for heartbeats from backups and scheduled jobs. It checks from US East and EU Central every minute (every 30 seconds on Team and above), and it alerts the right person by email, SMS, Slack, Microsoft Teams or webhook only after a second region confirms the problem. Nothing to install on the host.
Server monitoring that answers one question first
Is the server reachable, and are its services answering? A server can be powered on and still useless: the web service has crashed, the mail daemon stopped accepting connections, the database refuses new clients after hitting its connection limit, or a firewall change closed SSH. Checks from outside the network catch all of these, because they test the path a real client takes.
UptimePilot combines three check types for servers, all included in every plan.
- Ping (ICMP) to confirm the host is reachable and to measure round trip time and packet loss.
- TCP port checks to confirm a service accepts connections on its port, with a connect timeout you set.
- Heartbeat monitors for backups, scheduled jobs and workers that run on the server and report in by calling a unique URL.
Server uptime monitoring with ping and TCP ports
A ping check sends ICMP echo requests and records whether replies come back and how long they take. It tells you the host and the network path are up. Many hosts block ICMP on purpose, so ping alone is not enough. A TCP port check opens a connection to a port and records how long the handshake takes. If the connection is refused or times out, the service is down even when the host still answers ping.
Typical port checks that teams set up on day one look like this.
- Web on 80 and 443, next to an HTTP check from website monitoring for the status code and keyword.
- Mail on 25, 465, 587 and 993, so a stopped mail service does not go unnoticed for a weekend.
- Databases on 3306, 5432 or 6379, reachable only from allowed addresses, so you check the ports that are meant to be public.
- SSH on 22 or your custom port, to know remote access still works before you need it during an incident.
For a deeper look at ports and network paths, see the network uptime monitor page.
Linux server monitoring and Windows server monitoring without an agent
Because every check runs from outside, UptimePilot works the same way for Linux and Windows servers, virtual machines, bare metal and appliances. You do not install an agent, open an inbound management port or give UptimePilot credentials. You add an IP address or hostname, choose the check type and the port, and the first result arrives within a minute. Windows hosts usually block ping by default, so a TCP check on 3389, 443 or the port of your application is the reliable signal there.
What UptimePilot does not measure
UptimePilot is not a metrics agent. It does not collect CPU load, RAM usage, disk space or process lists from inside the server. If you need those graphs, keep the tool that collects them, and use UptimePilot as the independent outside view that still works when the server itself, or its local monitoring, has stopped.
How server monitoring software should decide a server is down
A single failed probe is not an outage. Packets get lost, a route flaps for a few seconds. UptimePilot marks a monitor down only after a second region confirms the failure, which removes most false alarms caused by one network path. You set the timeout per check (for a TCP connect, 5 seconds is a sensible start; for a host across an ocean, 10 seconds) and the number of failed checks in a row before an alert.
When a check fails and is confirmed, UptimePilot records an incident with its start time, the cause (timeout, connection refused, packet loss above your threshold) and the region results. The alert goes to the channel you chose. If nobody acknowledges it within the number of minutes you set, escalation alerts the next person. When the check passes again, a recovery alert closes the incident with its duration. Read more on website downtime tracker alerts and escalation.
- Maintenance windows pause alerts while you patch and reboot, so the reboot does not page anyone.
- Groups keep servers per client, project or environment, with their own alert rules.
- A status page can show servers as components, so customers see "Mail delayed" instead of opening tickets.
Heartbeats for backups and scheduled jobs on the server
The worst server problems are silent. A backup script exits with an error every night and nobody reads the log. A heartbeat monitor flips the check around: the job calls its unique URL when it finishes, and UptimePilot alerts you when the call does not arrive on schedule plus a grace period. The full setup is on the cron job monitoring page.
Which plan fits your servers
Every plan includes ping, TCP port, HTTP, keyword, API, SSL certificate, domain expiry and heartbeat checks, multi-region confirmation, maintenance windows and escalation. Plans differ by limits.
- Starter, $49 a month or $24 a month billed yearly ($288 a year). 50 monitors, checks every minute, 10 heartbeat monitors, 30 SMS alerts a month, 3 team members, 90 days of history.
- Team, $99 a month or $49 a month billed yearly ($588 a year). 200 monitors, checks every 30 seconds, 50 heartbeat monitors, 150 SMS alerts a month, 10 team members, 1 year of history, monthly SLA PDF.
- Business, $249 a month or $124 a month billed yearly ($1,488 a year). 1,000 monitors, 250 heartbeat monitors, 500 SMS alerts a month, 25 team members, 2 years of history, white-label client reports and REST API. A good fit for MSPs.
- Enterprise, $599 a month or $299 a month billed yearly ($3,588 a year). 3,000 monitors, 1,000 heartbeat monitors, 1,500 SMS alerts a month, 100 team members, 3 years of history, SSO (SAML) and audit log.
Compare every limit on uptime monitoring pricing. Yearly billing saves 50 percent, and you cancel anytime from your account.
01
Do I need to install anything on my servers?
No. UptimePilot checks from outside, with ping and TCP connections from two regions. The only optional change on the server is a curl call at the end of a job, if you use heartbeat monitors.
02
Can UptimePilot show CPU, memory or disk usage?
No. UptimePilot checks reachability and services from outside and does not collect metrics from inside the host. It complements a metrics tool by working even when the server or that tool is down.
03
My server blocks ping. What should I use?
Use a TCP port check on a port the server serves, such as 443 or 22. It proves the service accepts connections, which is a stronger signal than ping anyway.
04
Can I monitor database and mail ports?
Yes. Any TCP port that is reachable from the internet can be checked, including mail and database ports. If a database is restricted to certain addresses, check the application port that depends on it instead.
05
How fast will I hear about a server that went down?
Checks run every minute on Starter and every 30 seconds on Team and above. After a second region confirms the failure, the alert goes out immediately to the channels you chose.
Start monitoring your servers in a few minutes
Add a hostname or IP address, choose ping or a TCP port, pick who gets alerted. The first result arrives within a minute.