Most outages start on the server long before a visitor notices. The disk fills with logs over a week, memory leaks until the process is killed, a kernel update reboots the machine at night and a service does not come back. An uptime check sees the result. Server monitoring sees it coming.
Uptimecheck tracks server resources with a small agent for Linux, macOS (Intel and Apple silicon) and Windows, and alerts you when a value crosses the threshold you set. Read keeping your website running with server status monitoring for the background.
What the agent reports
- CPU usage, with the 1-minute load average on Linux and macOS.
- Memory usage.
- Disk usage for every disk, with the fullest one charted.
- Network traffic in and out.
- Reboots, so an unplanned restart does not go unnoticed.
Install in one line
Copy the install command from the server’s page and run it on the server. The agent is a single binary that runs as a service and reports once a minute. Releases are signed, and with automatic updates turned on, the agent installs new versions itself, only when the signature checks out.
Set thresholds that match your servers
A database server at 80 percent memory is normal. A web server at 80 percent memory may be leaking. Set CPU, memory and disk limits per server, and how long a value must stay above the limit before an incident opens, from 1 to 30 minutes. Typical starting points:
- Disk above 85 or 90 percent.
- CPU above 90 percent for 5 minutes or more, so a short spike during a deploy does not page anyone.
- Memory above 90 percent on a server that normally sits well below.
- Any reboot outside a maintenance window.
If reports stop arriving for longer than you allow, from 2 to 15 minutes, the server counts as down. Our guide to server alert thresholds goes deeper.
A typical night
At 01:40 the disk on web-01 passes 92 percent. The nightly backup will need more space than that. Uptimecheck sends the alert to the on-call channel, someone rotates the logs, and the backup runs on time. Without the alert, the backup fails, the cron job monitor goes missing at 02:05, and by morning the web server is out of space too.
Servers next to everything else
Uptime checks tell you a site stopped answering. Server metrics tell you why, or warn you before it happens. When a website monitor fails, the first question is usually whether the server is overloaded, out of memory or out of disk. Put servers, uptime monitors and cron jobs in the same project, and the answer is one click away instead of an SSH session at three in the morning.
For agencies and hosting providers
Group servers per client or per environment in projects, and give each team access to the workspace with the right role. A disk that grows two percent a week is a conversation to have with the client before it becomes an outage.
Alerts go to email, Slack, Microsoft Teams, Discord, Google Chat, Mattermost, Telegram, PagerDuty, ntfy, Pushover, webhooks and push notifications on every plan. Start includes 5 servers, Professional 25 and Business 150.
Plans start at €9 per month billed yearly or €12 billed monthly, excl. VAT. Compare plans on the pricing page, or start a 14-day free trial without a card.