In this guide · 8 sections
Short answer
Set seven alerts: site or service down, disk above 80%, CPU or memory sustained above 70-80%, SSL certificate expiring within 14 days, domain renewal, backup did not run and unusual failed logins. Send them to two named people, and test each alert once.
The seven alerts worth setting
Start with checks that map to real failures. Each one below catches a problem that has a clear fix.
| Alert | What it catches | Starting threshold |
|---|---|---|
| Uptime (HTTP or port check from outside) | Site, app or server unreachable | Down for 2 consecutive checks |
| Disk usage | Logs, uploads or backups filling the disk | Above 80%, then urgent at 90% |
| CPU and memory | Runaway process, traffic spike, undersized server | Above 70-80% for 10 to 30 minutes |
| SSL certificate expiry | Auto-renewal that silently failed | 14 days before expiry |
| Domain expiry | Lapsed card or old registrar email | 30 days before expiry |
| Backup ran and succeeded | Backup job stopped or failed | No success signal within 26 hours of a daily job |
| Failed logins | Password guessing against SSH, admin panels or RDP | Sharp rise above the normal daily level |
DigitalOcean's alert set-up guide recommends 70% as a starting threshold for CPU, memory and disk, and load averages near the server's vCPU count. It also lets you require the condition to last 5, 10 or 30 minutes, or an hour, before alerting, which cuts noise from brief spikes.
Outside-in checks vs inside metrics
There are two kinds of monitoring, and small businesses need both.
Uptime checks (outside-in)
Is it reachable?Free to about US$12/month for a few checks
- A service outside your server requests the site or port every 1 to 5 minutes
- Catches outages, DNS mistakes and expired certificates
- Works even if the server itself is down
- Cannot warn you that a disk is about to fill
Best for: Websites, booking pages, APIs and email servers.
Server metrics (inside)
Is it healthy?Often free from your cloud provider
- An agent on the server reports CPU, memory, disk and load
- Warns days or weeks before a failure
- Shows trends for capacity planning
- Silent if the server or agent is dead, so pair with uptime
Best for: VPS, cloud servers and on-site servers.
Our comparison of uptime monitoring tools for websites goes deeper on the outside-in side.
Tools and prices (as of October 2026)
Prices below are from each provider's page, checked on 21 October 2026, in US dollars.
| Tool | What it does | Price |
|---|---|---|
| UptimeRobot Free | 50 monitors, 5-minute interval, HTTP, port, ping and keyword checks | US$0; described as for hobby and non-profit projects |
| UptimeRobot Solo | 10 monitors, 60-second interval, all monitor types including SSL and DNS | US$12/month billed annually (US$13 monthly) |
| UptimeRobot Team | 100 monitors, 30-second interval, 3 seats, webhooks | US$39/month billed annually (US$46 monthly) |
| DigitalOcean Monitoring | CPU, memory, disk, load and bandwidth alerts for Droplets by email or Slack | No additional cost |
| DigitalOcean Uptime | URL or IP checks with latency and SSL alerts | US$1 per check per month; 1 check free |
| Healthchecks.io | Alerts when a scheduled job such as a backup does not check in | Free for 20 jobs; Business US$20/month |
AWS, Azure and Google Cloud have their own metrics and alerting, usually with free allowances. For a single server, your cloud provider's built-in alerts plus one outside uptime service is usually enough.
Monitoring backups: the alert most people miss
Backup software normally emails when a job fails. It does not email when the job never runs: a cron entry deleted during an upgrade, an expired cloud credential, a server rebuilt without the backup script. That silent failure is the most common way small businesses discover they have no recent backup.
The fix is a 'dead man's switch'. The backup job pings a monitoring URL when it finishes successfully; if the ping does not arrive on schedule, you get an alert. Healthchecks.io and similar tools (including heartbeat monitors in UptimeRobot's paid plans) work this way.
Who gets the alerts, and how
An alert sent to a shared inbox nobody reads is the same as no alert. Decide before the first incident who receives what.
- Two named people per alert, so holidays and sick days are covered.
- Urgent alerts by phone push, SMS or a chat app; warnings by email. UptimeRobot sells SMS and voice credits separately, and supports Slack, Teams and Telegram on paid plans.
- WhatsApp alerts usually need a third-party service or webhook; check the cost and that numbers are kept up to date.
- A first step written down for each alert: who to call, where the login is kept, and how to restart the service.
- A status page for customer-facing systems, so staff can see whether a problem is already known.
Keep alerts quiet enough to trust
Monthly alert hygiene
Monitoring tells you something is wrong; it does not patch or fix anything. Our monthly server maintenance checklist covers the updates and checks that prevent many alerts in the first place, and our guide to website maintenance costs in Singapore shows what different levels of support cost.
Do it yourself or get help?
If you are comfortable with your cloud provider's console, setting the seven alerts takes an afternoon. The harder part is deciding who responds, and acting on alerts consistently month after month.
Our Monitor plan (S$29/month) watches your website, SSL certificate, domain and email reputation with uptime checks every 5 minutes and a monthly report. For servers, Server Protection Care (S$149/month per server) applies monthly updates, checks that backups and protections are still working, runs a monthly restore test with proof, and tracks certificate and domain renewals. IT Sorted is a Singapore-based team of developers with over 10 years of experience; you pay upfront, send the brief in writing and follow progress on your order page.
Frequently asked questions
Start with seven checks: outside uptime, disk usage, CPU and memory, SSL certificate expiry, domain expiry, whether the backup ran successfully, and failed login attempts. These cover most real-world failures. Add more only when you have a specific problem to watch for.
A common approach is a warning at 80% and an urgent alert at 90%. DigitalOcean suggests 70% as a starting threshold for CPU, memory and disk alerts. Tune the level after a few weeks, based on how quickly your disk usage normally grows.
Yes. Many cloud providers, including DigitalOcean, include resource alerts at no extra cost. UptimeRobot's free plan offers 50 monitors at 5-minute intervals, though it describes the plan as for hobby and non-profit projects. Healthchecks.io monitors up to 20 scheduled jobs for free.
Use a heartbeat or 'dead man's switch' check. The backup job pings a monitoring URL each time it completes; if no ping arrives on schedule, the service alerts you. This catches jobs that silently stopped, which a failure email alone cannot.
Some monitoring tools offer WhatsApp alerts directly, such as Healthchecks.io on its paid plans with WhatsApp credits, while others need a webhook to a third-party service. SMS, push notifications, Slack, Teams and Telegram are more widely supported. Always have a second channel in case one fails.