Many of the outages I’ve been called to fix could have been prevented with an alert a few days earlier. A disk that filled up slowly with logs until the database couldn’t write. Memory usage creeping up until PHP processes started crashing. An SSL certificate whose renewal had been failing silently for weeks. A backup job that stopped running months before anyone needed it.
Monitoring turns these silent failures into early warnings. It’s not glamorous, but it’s one of the most valuable things you can set up for any server running WordPress. Here’s what I monitor, how and what makes alerts genuinely useful rather than noise.
Two kinds of monitoring
External monitoring
Checks your site from outside, the way visitors experience it. Is it up? Is it fast? Is the SSL certificate valid? External monitoring catches problems regardless of their cause, including DNS failures, network issues and complete server outages, where internal monitoring can’t report anything.
Internal monitoring
Watches the server itself: CPU, memory, disk, processes and services. It tells you why something is going wrong, and often warns you before visitors notice.
You need both. External monitoring tells you when something is broken. Internal monitoring tells you what’s about to break and why.
External checks to set up
Uptime
Check your site every minute or few minutes from several locations. Don’t just check that the server responds; check that the page contains expected content, like a word from your homepage or footer. A server returning a blank page or an error page with a 200 status should count as down.
Key pages
Monitor more than the homepage:
- A page that isn’t heavily cached, to detect PHP or database problems that cached pages hide.
- The login page.
- For stores, the cart or checkout.
- Important forms or API endpoints.
Response time
Track how long pages take to respond. A gradual slowdown is an early warning of resource problems, database growth or a misbehaving plugin.
SSL certificate expiry
Alert well before certificates expire, such as two or three weeks ahead. Automatic renewal usually works, until it doesn’t.
Domain expiry
Alert a month or more before the domain expires.
DNS
Some monitoring services check that DNS records still point where they should, catching accidental changes or hijacking.
Internal metrics to watch
Disk space
Disk space is the most common cause of mysterious failures. When a disk fills up, databases stop writing, logs stop recording, uploads fail and sessions break, often with confusing error messages.
Watch:
- Overall disk usage: alert at around 80 percent and again at 90 percent.
- Inodes: the number of files a filesystem can hold. Millions of small files, like cache files or sessions, can exhaust inodes even when space remains.
- Growth rate: a disk growing faster than usual may indicate runaway logs, backups stored locally or a malware infection creating files.
| du -sh /var/log /var/www/* | sort -h</code> |
|---|
Memory
Watch memory usage and swap. Heavy swapping slows everything dramatically. If the kernel’s out-of-memory killer starts terminating processes, it often kills the database or PHP, causing outages.
| dmesg -T | grep -i "killed process"</code> |
|---|
Alert when available memory stays low for an extended period or swap usage climbs significantly.
CPU and load
High CPU usage for short bursts is normal. Sustained high usage suggests traffic spikes, bot attacks, inefficient code, runaway processes or cryptocurrency mining malware. Load average compared with the number of CPU cores gives a quick sense of whether the server is overloaded.
PHP-FPM
Watch for:
- Workers reaching pm.max_children, which means requests are queuing.
- Slow requests in the PHP-FPM slow log.
- PHP-FPM restarts or crashes.
The PHP-FPM status page shows active, idle and queued processes, and many monitoring tools can read it.
Database
Watch:
- Whether the database service is running.
- Connection counts compared with the maximum.
- Slow query counts.
- Replication status, if you use replicas.
Web server
Watch error rates, especially 5xx responses. A spike in 502 or 504 errors usually points to PHP or upstream problems.
Services
Make sure critical services are running: Nginx or Apache, PHP-FPM, the database, Redis and any queue workers. Monitoring tools can check processes and restart them if needed, though repeated restarts mean something needs fixing.
Application-level checks
Beyond the server:
- Backups: alert if a backup job fails or the latest backup is older than expected.
- Scheduled tasks: alert if WordPress cron or Action Scheduler stops running. Heartbeat monitoring, where your cron job pings a URL each time it runs, works well.
- Security: alerts for file changes, new administrator accounts and malware detection.
- Email delivery: monitoring transactional email bounces and failures.
Tools
You don’t need an enterprise monitoring platform to monitor a few WordPress servers well.
- External uptime services offer free and paid plans for uptime, response time, SSL and domain checks, with alerts by email, SMS and chat apps.
- Lightweight server agents collect and visualise CPU, memory, disk, processes and service metrics with minimal setup, often with built-in alerts.
- Your hosting provider’s monitoring, often available in the control panel, covers basic resource usage.
- Self-hosted monitoring stacks, combining metrics collection, dashboards and alerting, suit people managing many servers who want full control.
- Simple scripts with cron can check disk space and send alerts, a perfectly good start for a single server.
A minimal disk alert script:
| USAGE=$(df / –output=pcent | tail -1 | tr -dc '0-9') |
|---|---|---|
| echo "Disk usage on $(hostname) is ${USAGE}%" | mail -s "Disk alert: $(hostname)" [email protected] |
Make sure the server can actually send email, or use a chat webhook instead.
Setting useful alerts
Monitoring is only valuable if alerts lead to action. Too many alerts, and people stop paying attention. Too few, and problems go unnoticed.
Alert on symptoms that matter
Prioritise alerts that indicate visitor impact or imminent failure: site down, errors rising, disk nearly full, certificate about to expire, backups failing.
Use sensible thresholds
Avoid alerting on brief spikes. Alert when a condition persists, like CPU above 90 percent for ten minutes, or memory low for fifteen minutes.
Use severity levels
- Critical: site down, disk almost full, database stopped. Send to a channel that wakes someone up if needed.
- Warning: disk at 80 percent, certificate expiring in two weeks, slow responses. Send to email or a chat channel reviewed daily.
Route alerts to the right place
Send alerts where someone will see them, like a dedicated chat channel or phone notifications for critical issues. An alert sent to an unmonitored inbox protects nothing.
Review and tune
When an alert fires without a real problem, adjust its threshold. When a problem happens without an alert, add one.
Dashboards and history
Historical graphs are as useful as alerts. They show trends, like gradual memory growth or disk usage climbing every month, and help diagnose incidents after the fact. When something goes wrong, being able to see exactly when memory started rising or when traffic spiked makes root cause analysis much faster.
Logs as monitoring
Logs complement metrics. Error logs, PHP-FPM logs and access logs explain what metrics show. Make sure logs are rotated, so they don’t fill the disk, and kept long enough to investigate incidents.
Start small
If you’re not monitoring anything today, don’t try to set up everything at once. Start with external uptime monitoring with a content check, SSL and domain expiry alerts and a disk space alert. That takes less than an hour and covers the problems that most often take sites offline. Add memory, PHP-FPM, database and backup monitoring over the following weeks as you learn what normal looks like for your server.
Review alerts after incidents
After every outage or near miss, ask whether monitoring caught it, how quickly and whether the alert reached the right person. Then adjust. Monitoring improves fastest when it’s tuned against real events rather than guesses.
Keep monitoring documented
List what’s monitored, where alerts go and who responds. When people change roles or go on holiday, that list makes sure alerts still reach someone who can act.
Frequently asked questions
Do I need server monitoring on managed hosting?
Your host monitors the infrastructure, but external uptime monitoring is still valuable, and you should know where to see resource usage for your account.
What’s the most important thing to monitor?
External uptime with content checks, plus disk space. Together, they catch a large share of real-world problems.
How often should uptime checks run?
Every one to five minutes is typical. Business-critical sites benefit from one-minute checks from multiple locations.
How do I avoid alert fatigue?
Alert only on conditions that need action, use thresholds that ignore brief spikes, separate critical alerts from warnings and tune alerts regularly.
Final thoughts
Monitoring is how you hear about problems while they’re still small. Combine external checks for uptime, key pages, response time, SSL and domain expiry with internal monitoring of disk, memory, CPU, PHP-FPM, the database and critical services. Add checks for backups and scheduled tasks, set thresholds that ignore noise and route alerts to people who will act. A little setup now saves a lot of late-night firefighting later.




