How to Fix Server Uptime: Monitor and Reach 99.9%+

Uptime can't be judged from one snapshot. A site that loads during an audit may still go down for hours each month. This guide shows you how to set up continuous external monitoring, read the results, fix the most common causes of downtime, and decide whether you need a more reliable host.

By Updated ⏱ 45 minutes to set up, then 30 days of data🎯 Intermediate● High impactTechnical
Illustration: Server uptime fix guide
TECHNICAL · FIX GUIDE Server uptime IMPACTEFFORT TIME45 minutes to set up, then 30 days of data LEVELIntermediate

Why it matters

When Googlebot repeatedly gets connection errors or 5xx responses, it slows its crawling. If the outage lasts, pages can drop out of the index. AI crawlers and answer engines that fetch your pages during an outage get nothing to cite.

Visitors and paid traffic land on error pages, so ad spend is wasted and checkouts fail. 9% uptime still allows about 43 minutes of downtime a month. 99% allows over 7 hours, which customers and crawlers will notice.

✕BeforeNo monitoring in place; the site wasdown 6 hours last month and the ownerlearned about it from a customer✓AfterAn external monitor checkshttps://example.com/healthz everyminute from several regions, sends

What you need

Admin access to your hosting control panel or server (SSH for VPS/dedicated)
An account with an external uptime monitor (UptimeRobot, Better Stack, Pingdom, or self-hosted Uptime Kuma on a separate server)
Google Search Console access for your domain
Access to DNS/CDN settings if you use Cloudflare or similar

Step-by-step fix

1

Create a health-check URL that tests the full stack

Monitoring only the homepage can miss failures when a CDN serves a cached copy, and a static file only proves the web server is running. Add a small endpoint that queries the database and returns 200 OK when healthy or 503 when not. Make sure it is never cached, and keep it out of search results with a noindex header or a robots.txt disallow.

php
<?php
// /healthz.php on example.com
header('Cache-Control: no-store');
header('X-Robots-Tag: noindex');
try {
  $pdo = new PDO('mysql:host=localhost;dbname=example', 'monitor_user', 'CHANGE_ME', [PDO::ATTR_TIMEOUT => 3]);
  $pdo->query('SELECT 1');
  http_response_code(200);
  echo 'OK';
} catch (Exception $e) {
  http_response_code(503);
  echo 'DB DOWN';
}
Pro tip: Give the monitor a read-only database user with no table privileges. SELECT 1 doesn't need them.
2

Add external HTTP monitors from multiple regions

In your monitoring tool, create an HTTP(S) monitor for https://example.com/healthz and a second one for https://example.com/. Configure them like this:

  • Check interval: 1 minute, if your plan allows it.
  • Timeout: 10–30 seconds.
  • Keyword check: set it to OK on the health URL, so a 200 response showing an error page still counts as down.
  • Confirmation: require failure from at least 2 locations before marking the site down, to avoid false alarms.

Also enable SSL certificate expiry and domain expiry alerts. Expired certificates and domains are a common cause of avoidable outages.

Pro tip: Never host your monitor on the same server or network you are monitoring. If the server dies, the monitor dies with it and stays silent.
3

Route alerts to people who will act on them

Add at least two alert channels, such as email plus SMS, Slack, Microsoft Teams or a phone call. Set a short escalation: alert immediately, then repeat every 15 minutes until someone acknowledges. Turn on recovery notifications so you can record how long each outage lasted.

com) from your monitoring tool.

Pro tip: Add your hosting provider's status page RSS feed or alerts to the same channel. That way you can tell a host-wide incident from a problem on your own server.
4

Check Google's view in Search Console

In Google Search Console, go to Settings → Crawl stats → Open report, then click Host status. Look at Server connectivity. A red or warning icon means Googlebot hit server errors or timeouts in the last 90 days.

Then open By response and check the share of 5xx responses. Compare the dates with your monitor's incident log.

Pro tip: For planned maintenance, return 503 with a Retry-After header instead of 200 or 404. Google treats 503 as temporary.
5

Find the cause of each outage in the server logs

Run these commands on a VPS or dedicated server for each incident window. You are looking for four things:

  • crashed services
  • out-of-memory (OOM) kills
  • disk-full errors
  • traffic spikes, such as bots or attacks, that line up with the downtime

On shared hosting, open Metrics → Errors and Resource Usage in cPanel (or your host's equivalent) and look for CPU, entry-process or memory limit hits.

text
uptime
free -m
df -h
sudo dmesg -T | grep -i -E 'killed process|out of memory'
sudo journalctl -u nginx -u php8.3-fpm -u mysql --since "2026-01-10 02:00" --until "2026-01-10 04:00"
sudo tail -n 200 /var/log/nginx/error.log
sudo awk '{print $1}' /var/log/nginx/access.log | sort | uniq -c | sort -rn | head
Pro tip: If repeated OOM kills point to PHP-FPM or MySQL, lower pm.max_children or innodb_buffer_pool_size to fit your RAM before you pay for a bigger server.
6

Make services restart and serve stale content on failure

Configure systemd to restart crashed services automatically. Then let Nginx (or your CDN) serve a cached copy of pages when the backend returns errors or times out. These two changes turn many short crashes into outages visitors never see.

nginx
# sudo systemctl edit php8.3-fpm  (repeat for mysql, your app service)
[Service]
Restart=on-failure
RestartSec=5s

# /etc/nginx/conf.d/cache.conf
proxy_cache_path /var/cache/nginx levels=1:2 keys_zone=site:10m max_size=1g inactive=24h;

server {
  server_name example.com;
  location / {
    proxy_cache site;
    proxy_cache_valid 200 301 10m;
    proxy_cache_use_stale error timeout updating http_500 http_502 http_503 http_504;
    proxy_cache_background_update on;
    proxy_pass http://127.0.0.1:8080;
  }
}
Pro tip: After any change, run sudo nginx -t && sudo systemctl reload nginx so a typo doesn't take the site down.
7

Move hosts if you're still below 99.9% after 30 days

If the monitor shows less than 99.9% over a full month and the outages trace back to the host, plan a migration. Typical host-side causes are network incidents, overloaded shared servers and unannounced maintenance. When choosing a new host:

  • Pick one with a written SLA of 99.9% or higher that pays service credits.
  • Prefer managed or cloud hosting with automatic failover.
  • Lower your DNS TTL to 300 seconds a day before the switch.
  • Run old and new servers side by side until DNS has propagated.
Pro tip: An SLA only gives you money back after an outage. Your own 30-day monitor history is the evidence you need to claim credits or justify the move.
8

Review the uptime report monthly

Once the fixes are live, let the monitor run for at least 30 days. Then export the uptime report and check three things:

  • the overall percentage
  • the number of incidents
  • the longest single outage

The target is 99.9% or higher, with no outage longer than about 15 minutes. Repeat this review every month and after any hosting, plugin or deployment change.

Pro tip: Also track average response time. A steady rise usually comes before resource-limit outages.

Platform-specific notes

WordPress

Jetpack includes a free Downtime Monitor under Jetpack → Settings → Security, but it checks only about every 5 minutes. Pair it with an external 1-minute monitor. On shared hosting, most WordPress downtime comes from plugin-heavy pages hitting CPU or entry-process limits. A page-cache plugin or a move to managed WordPress hosting usually solves it.

Shopify

Shopify runs the hosting, so you can't change servers. Still monitor your storefront and checkout URLs to catch DNS, app-script or theme problems. Check shopifystatus.com during incidents to confirm whether an outage is platform-wide.

Wix / Webflow / Squarespace

Uptime depends on the platform, so moving hosts means moving platforms. Monitoring mainly catches custom-domain DNS errors and SSL failures. During outages, check the vendor status page (for example status.webflow.com) and keep your DNS records exactly as the platform documents them.

Cloudflare

Enable caching with Always Online (Caching → Configuration) so cached pages keep serving if your origin goes down. Pro plans and above also offer Traffic → Health Checks for your origin. Watch for 52x errors (520–526): they mean Cloudflare could reach no healthy origin, so the fix is on your server, not at Cloudflare.

Next.js (Vercel or self-hosted)

On Vercel, add a route such as app/healthz/route.ts that returns a 200 response and checks your database. Use export const dynamic = 'force-dynamic' so it is never statically cached. Self-hosted Node apps should run under a process manager (systemd with Restart=on-failure, or PM2) behind Nginx with proxy_cache_use_stale.

Check that it worked

✅ How to verify

  • Stop your app or database service briefly during a quiet period. Confirm the monitor marks the site down and alerts arrive on every channel within 1–2 minutes, then confirm the recovery alert.
  • Run curl -I https://example.com/healthz and confirm it returns HTTP/2 200, Cache-Control: no-store and X-Robots-Tag: noindex.
  • After 30 days, check that the monitor's uptime report shows 99.9% or higher with no long single outages.
  • In Search Console under Settings → Crawl stats → Host status, confirm Server connectivity shows a green check for recent weeks.

⚠️ Common mistakes

  • Running the uptime monitor on the same server or provider as the site. A full outage then produces no alert.
  • Monitoring only a cached homepage behind a CDN. The CDN keeps returning 200 while the origin, checkout and logins are down.
  • Treating the host's advertised '99.9% uptime' as fact without measuring it, or not knowing the SLA excludes 'scheduled maintenance'.
  • Returning 200 or 404 for maintenance pages instead of 503 with Retry-After. This can make Google index the maintenance page or drop URLs.

Is this still an issue on your site? Run a free audit — ZORO checks this and 294 other SEO, AI-visibility, accessibility and security signals, with fixes for your platform.

Check my site free

FAQ

What is a good server uptime percentage?

Aim for at least 99.9%, which allows about 43 minutes of downtime per month. Business-critical and e-commerce sites should target 99.95–99.99%, or roughly 22 to 4 minutes a month. Below 99.5% (about 3.6 hours a month), outages will affect both visitors and crawling.

Does website downtime hurt SEO?

Short, occasional outages have little lasting effect, especially if your server returns 503. Repeated or multi-day outages make Googlebot slow its crawling, and pages that keep returning errors can be dropped from the index until the site is stable again.

Why can't a single site audit measure uptime?

Uptime is a percentage over time. One request only shows the site was up at that moment. You need continuous checks, ideally every 1–5 minutes from several locations for at least 30 days, to calculate a meaningful figure.

How often should an uptime monitor check my site?

Every 1 minute is best for revenue-generating sites. A 5-minute interval is acceptable for small sites, but it can miss short outages and inflates the measured uptime. Always require confirmation from more than one location to avoid false alarms.