Server monitoring that warns you before the outage.

Most servers don't crash — they run out of something. One bash line installs the agent, 30 seconds later it's reporting, and the disk gets fixed at 90%, not at 100%.

Start free trial → View pricing
30-day free trial · up to 50 servers No credit card Linux · installs in under a minute
Report interval
30s
Metrics per server
17
Checkpoints
171+
Countries covered
70+

A /var mount at 92%

Log rotation stalls overnight on one of your servers. The next agent report crossed 90%, and the alert named the server and the mount in its first line.

02:14:03 /var crosses 90% on one serverThe 30-second report shows 92% — disk alerts are instant, no window the box: still up
02:14:03 The alert names the mount"/var at 92%" in the first line — Slack, email; not a vague "disk" alarm the box: still up
02:29 Old logs cleared — space freedThe next agent report, 30 seconds later, shows /var at 62% the box: still up
02:29:31 Auto-resolved — you're told it's overAn up notice goes to the same channels that raised the alarm the box: never fell
15 mincross → clear
The crash that never happened.02:29:31
A full /var takes the whole box down — databases, logs, sites, everything. This one peaked at 92%, because the alert arrived while there was still runway.
exact mount + actual valueinstant, per-mountup notice on recoveryCPU · RAM · load · disk — every 30 s
And without the agent? A filling disk has no symptoms until 100% — then everything on the box fails at once. The 2 a.m. climb becomes a 7 a.m. discovery, with five dead hours in between. dead box

17 metrics, one bash agent

One bash agent reports 17 metrics every 30 seconds — already charted, so you read a server's health without logging in to check.

CPU & loadUsage and 1·5·15 m load average
Memory & swapUsed, free — and swapping
Disk % — per mountEvery mount, not just /
Inode usageFull of tiny files? You'll know
17 metrics · 30 s read from the system itself · no logging in
Processes & coresCount watched, cores known
Network in / outThroughput, both directions
Host & freshnessKernel, OS, uptime — and a live pulse
Alerts fire on sustained breaches, not blips: CPU, memory, load, swap, disk and inodes per mount, process count, network throughput — and the loudest one: the server going silent. silence = alert

Alerting and history

No 3 a.m. false alarms

Paged for real problems, not momentary spikes

Uptimia fires a CPU, memory, swap, load, process-count or network alert only when every reading in your window stays over the line — a two-second spike passes in silence, a sustained problem doesn't.

  • You decide how long a problem must hold — anywhere from 1 minute to half an hour, set per metric
  • Nothing to configure to start — CPU, memory and disk are watched from the first report; add the others when you want them
  • Disk and inodes are the exception — they fire the instant a mount crosses; a full disk can't wait
Where your alerts are delivered →
CPU02:05:0030-second sample94%
CPU02:07:3030-second sample96%
CPU02:10:0030-second sample93%
10/10samples
over 90%
CPU alert — sustained02:10:01
Every reading for 5 straight minutes stayed above 90% — a real problem, not a blip that fixed itself.
SlackEmailSMS+ PagerDuty…
A two-second spike? One hot sample never survives the window — it passes in silence. Only disk and inodes fire instantly. no alert
Per-mount disk

Know exactly which disk filled up, instantly

Uptimia scopes disk-space and inode alerts to each mount, so the incident says /var — not a whole machine to hunt through. Every mount opens and resolves its own incident.

  • One incident per mount — /, /var, /data each tracked and resolved on their own
  • The mount and the value in the alert — "Disk usage above 90% on /var", at 92%
  • Per-mount overrides — hold a busy mount like /var to a stricter percentage than the rest
If nobody acknowledges an alert →
agentweb-01.caldmont.comreporting every 30 s
/rootown incident track61%
Disk alert — /var02:14:09
Disk usage above 90% on /var — at 92%. Instant, no waiting window; its own incident while /data stays green.
per-mountinstantinodes too
/datadata volumeown incident track38%
/var busier than the rest? Per-mount overrides hold one mount to a stricter line than the machine-wide default. /var → 85%
Offline detection

Three missed reports opens an incident

If a server goes completely silent — kernel panic, power, network — a separate per-minute check notices the missing reports and opens a critical incident. A dead box can't hide behind "no news is good news".

  • Flagged offline after 3 missed reports — 90 seconds of silence at the 30-second cadence
  • Auto-resolves on the next check-in — closes itself the moment the agent is back
  • No false "offline" on setup — a new server awaiting its first report is never flagged
Outside-in checks: ping and TCP port →
1
Report missed
02:40:30 · expected every 30 s
wait
30 more seconds of silence
2
Still silent
02:41:00 · second miss
wait
the third miss confirms it
3
Confirmed offline
02:41:30 · ~90 s of silence
critical
Server offline02:41:30
web-03 went completely dark — kernel panic, power or network. It auto-resolves the moment the agent reports again.
SlackEmailSMS+ PagerDuty…
Just installed the agent? A brand-new server awaiting its first report is never flagged offline. no false alarm
History & trends

A year of history, spikes included

Live numbers say what's wrong now; trends say it's been creeping up for weeks. Uptimia keeps both — 6 metric cards and 5 charts, reading the same data the alerts do.

  • 6 metric cards + 5 time-series charts — CPU, load, memory, disk, network — refreshed every 30 seconds
  • Raw 30-second detail for 24 hours — then hourly averages and maxima, so spikes survive, for a full year
  • Charts pick the right source — zoom to last hour or last year and the data switches automatically
Get these numbers emailed on a schedule →
TRENDMemory used — web-01
last 30 days · hourly data · refreshed every 30 s
30 d ago15 dnow
Creeping from 40% to 70% over a month — the kind of problem a live snapshot can't show you.
Raw · every 30 seconds 
every sample, kept 24 h
02:10:00 → cpu 93% · mem 71%
Hourly average · 1 year 
the trend line — smooth,
comparable month to month
Hourly max · 1 year 
spikes survive aggregation —
a 2-minute burst still shows a year later
Zoom anywhere — the last hour reads raw 30-second data, the last year reads hourly; charts switch source automatically. auto

Reporting in under a minute

One line to install — no packages, no runtime, and one script removes every trace.

Step 130 s

Run one install line as root

The one-liner drops a small bash script into /opt/uptimia and registers a systemd timer — or cron, if there's no systemd.

$ curl -s uptimia.com/server-agent/install.sh | bash -s -- KEY
# ✓ systemd timer uptimia-agent.timer created
Step 230 s

The agent starts reporting

Every 30 seconds it posts 17 metrics — CPU, memory, disk, load and more — authenticated with a per-server key.

Agent status
Connected — reporting every 30 s
first check-in received · 17 metrics
Server
web-01 · Ubuntu 22.04 · 4 cores
Step 3automatic

Set thresholds, get alerted

CPU, memory and disk start at 90%. A crossed threshold opens an incident and alerts your team where they already work.

#infra-alerts
Uptimia 02:14
⚠ Disk — /var at 92% on web-01
above 90%per-mount incident02:14:03
Also sent to EmailPagerDuty

Run one bash line.Know before the disk fills.

Every server, every metric, every alert channel — free for 30 days, and none of it is a paid add-on.

Start your free 30-day trial →
30 days free no credit card cancel anytime

Also included

Alerts on the channels you already use

Server alerts share one contact list with every other monitor — email, SMS, Slack, Microsoft Teams, Discord, Mattermost, Telegram, WhatsApp, PagerDuty, Twilio, Statuspage and custom webhooks.

EmailSlackTeamsPagerDutyTelegramWebhook+ more

Maintenance mode

Patch or reboot without alert debt — all alerting is silenced while metrics keep flowing.

reboot window · alerts silenced

Recovery notices

Hear when it's over, not just when it started — an "up" notice on every resolution.

✓ recovered · memory back to 58%

History that survives a downgrade

Drop a plan and a server's metrics aren't deleted — only alerting goes quiet. Upgrade and suspended servers reactivate themselves.

ingestion continues · alerts paused

One dashboard for everything

Servers sit beside your uptime, SSL, heartbeat and DNS monitors — same contacts, groups and roles, one place to look.

web-01.caldmont.comSERVER www.caldmont.comUPTIME db-backup · nightlyHEARTBEAT

Uninstall in one line

One script removes the systemd unit and /opt/uptimia entirely — no leftovers.

$ curl -fsS uptimia.com/server-agent/uninstall.sh | bash

What is server monitoring?

Server monitoring is the continuous tracking of a server's health — CPU, memory, disk, load, network and processes — so you're alerted the moment a resource runs short or the machine goes offline. Uptimia's Linux server monitoring does it with a small agent that reports 17 metrics every 30 seconds and raises an incident when a metric crosses your threshold.

The agent

How does the agent work?

Your Linux server
reads /proc · df
~17 metrics · every 30 s
Uptimia
charts + alerts

A bash script on a systemd timer POSTs one small JSON document, authenticated with a per-server key — only bash and curl needed.

Alert logic

Sustained vs. instant alerts

CPU · load · memory
whole window over
then alert
Disk · inodes
alert instantly

a spike passes in silence — only a sustained breach pages you; a full mount can't wait, so disk fires at once

The vitals

Which server metrics should you monitor?

The signals that predict real incidents — and how Uptimia watches each. CPU, memory and disk are watched from the first report; the rest are yours to switch on.

Every monitor type in one account →
MetricDefault thresholdHow Uptimia alerts
CPU usage90%sustained · 5-min window
Memory90%sustained · 5-min window
Disk · per mount90%instant, per mount
Load averageoff by defaultsustained · window
Swapoff by defaultsustained · window
Process countoff by defaultsustained · window
Network throughputoff by defaultsustained · window
Server offline3 missed reportscritical · ~90 s

Server monitoring FAQ

01What is server monitoring?+
The continuous tracking of a server's health — CPU, memory, disk, load, network and processes — so you're alerted when a resource runs short or the machine goes offline. Uptimia's small agent reports 17 metrics every 30 seconds and opens an incident when a metric crosses your threshold.
02How does Uptimia server monitoring work?+
A one-line install drops a small bash script on your server and registers a systemd timer (or cron). Every 30 seconds it reads /proc and df and posts 17 metrics — CPU, memory, swap, disk per mount, inodes, load, network, process count and OS info — authenticated with a per-server key. You get 6 metric cards, 5 time-series charts and threshold alerts — server monitoring and alerting from one small agent.
03Which operating systems are supported?+
The one-line installer targets Linux — the agent reads /proc and df, so it runs on standard distributions (Ubuntu, Debian, RHEL, Alma and so on). macOS and Windows PowerShell agents also exist and report the same metrics, set up by hand (launchd on macOS, a Scheduled Task on Windows) rather than by the one-liner. There is no BSD agent — for those hosts, use Uptimia's outside-in checks (ping, TCP port, HTTP).
04Can I change how often the agent reports?+
No — the 30-second cadence is fixed. What you control is alerting: each metric's threshold, and how long a problem must hold before anyone is paged.
05Will a brief CPU spike page me at 3 a.m.?+
No. CPU, memory, swap, load, process-count and network alerts only fire when every reading in your window (default 5 minutes) stays over the threshold — a momentary spike passes in silence. The exceptions are disk and inodes, which fire instantly: a full disk can't wait for a window.
06How quickly do I know if a server goes down?+
If a server stops reporting entirely, a per-minute check flags it offline after 3 missed reports — 90 seconds of silence — and opens a critical incident that auto-resolves when the agent checks back in. Three missed reports confirm it's really gone; a new server awaiting its first report is never falsely flagged.
07Does the agent read my logs or run commands?+
No. It's server resource monitoring only — the payload is CPU, memory, disk, network, load, uptime, process count and basic OS/kernel/hostname info. No logs, no file contents, no remote commands — a one-way report of numbers.
08How much history do I keep?+
Raw 30-second samples are kept for 24 hours. Beyond that, Uptimia stores hourly averages and maxima — spikes survive in the max — for a full year. Charts switch source automatically by range.
09Can I monitor Docker or Kubernetes containers?+
The agent reports host-level metrics — the whole machine's /proc and df, not per-container stats. Right for the servers your containers run on; no container-scoped collection today.
10Is server monitoring a paid add-on?+
No. Every paid plan includes server monitoring alongside uptime, SSL, transaction, DNS and heartbeat monitoring — never as a paid add-on. Plans differ only in how many servers are included. The free plan doesn't include server monitors; the 30-day free trial does (up to 50 servers), no credit card needed.

Start monitoring your servers today.

Run one bash line — and be the first to know when a server runs hot, fills up, or goes dark.

Monitor your servers →
30-day free trial Up to 50 servers No credit card Bash + curl · uninstall in one line
Server monitoring is available on the 30-day trial and every paid plan, next to every other monitor type.