Websites go down for a short list of reasons: a server or hosting failure, an overload from traffic, a bad code or plugin change, a DNS or expired-domain problem, an attack, or a failure in a service the site depends on. Most outages trace back to a change someone made or a limit someone did not see coming. The fastest way to find yours is to work out whether the site is down for everyone or only for you, then read the status code and check DNS, the certificate and the server in that order.
This guide walks through each cause, the symptom it usually produces, and the first check to run.
TL;DR
- Downtime means a site is unreachable, unresponsive, or too degraded to finish a task, not only a blank page.
- The usual suspects are human error (bad deploys, config changes), hosting or hardware failure, traffic spikes, DNS and domain lapses, plugin or CMS bugs, and attacks.
- The HTTP status code and the error type (timeout, DNS failure, certificate error) point to the layer that broke.
- Stabilize first, then investigate the root cause.
- Monitoring from outside your own network tells you about an outage before your users do.
Table of Contents
- What counts as website downtime
- Is it down for everyone or just you?
- The most common causes of website downtime
- What causes a website to time out
- Symptom-to-cause quick reference
- A short troubleshooting routine
- Our perspective: catch it before your users do
- Detecting downtime sooner
- Start monitoring your site with Pingfloat
- FAQ
- Sources
What counts as website downtime
Downtime is any period when a site is unreachable, unresponsive, or degraded so far that visitors cannot complete what they came to do. Gatling's overview of downtime causes makes the same point: it covers full outages and functional failures. A checkout that returns errors while the homepage loads fine is downtime for the people trying to buy.
It also helps to separate two kinds. Planned downtime is maintenance you schedule and announce. Unplanned downtime is everything else: infrastructure problems, traffic surges, software bugs and outside attacks, according to the same source. This article is about the second kind.
Is it down for everyone or just you?
Before you chase a cause, find out where the failure sits. Five minutes here saves an hour of wrong guesses.
- Test from another network. Open the site on your phone with Wi-Fi off. If it loads, the problem is your network, VPN, browser cache or local DNS cache.
- Test from outside your location. A check run from a remote server shows what visitors in other places see. A website uptime monitor checks from outside your own machine, so a local glitch does not fool you.
- Read the response. A status code means the server answered. A timeout or a name-resolution error means the request never got that far.
- Trace the path. If requests die partway, a traceroute shows the hop where packets stop, which separates a routing problem from a dead server.
- Look at the certificate. A browser warning can look like an outage. Run the domain through an SSL checker to see whether the certificate has expired or the chain is broken.
If the site fails everywhere, move on to the causes below. If it fails for some regions only, suspect DNS, a CDN or routing.
The most common causes of website downtime
Human error and bad deployments
People break sites more often than hardware does. Gatling's write-up cites human error as involved in 43% of unplanned downtime events, and it describes it as hard to detect and fix. The typical forms are a faulty release, a config change pushed to production, a deleted database table, or a firewall rule that blocks real traffic.
Symptom: the outage starts right after a deploy or change. Often a 500 error, a white screen, or a broken feature on one route.
First check: compare the failure time with your deploy log. If they line up, roll back before you debug. Google's SRE guidance on troubleshooting says that during a major outage you should stabilize the system first, for example by diverting traffic or disabling subsystems, and then investigate the root cause. The 500 Internal Server Error reference lists what typically sits behind that response.

Server, hosting and hardware failures
Small sites often run on one server, which makes that server a single point of failure. If it crashes, runs out of disk, or its host has a problem, nothing else answers. Plesk's guidance on downtime points to hardware as a leading cause of downtime for small and midsize businesses, and notes that database errors can leave a site loading partial or blank pages, or not at all.
Symptom: total unreachability, connection refused, or intermittent timeouts. Possibly a 502 or 504 if a proxy sits in front.
First check: log into your host's dashboard and look for incident notices, then check CPU, memory, disk and network on the machine. Google's own performance guidance suggests investigating CPU, network, memory and disk I/O bottlenecks, and says CPU that stays above 80% deserves a fix. If you are working out which resource is the limit, this walkthrough on cutting time to first byte covers common server-side bottlenecks.
Traffic spikes and overload
A site that handles 200 visitors an hour can fall over at 20,000. TechTarget points to a sudden surge exceeding available bandwidth as a classic crash trigger, and gives the example of Office Depot's site going down for about three hours on Cyber Monday in 2021. Product launches, a viral post, a sale email or a crawler gone wild can all do it.
Shared infrastructure adds a twist. Cloudflare explains that in multitenancy, a spike at one site can cause downtime for unrelated sites on the same hardware.
Symptom: slow pages first, then timeouts and 503 errors as load climbs. MDN describes 503 as the server not being ready to handle a request, commonly because of maintenance or overload. See our 503 Service Unavailable guide for fixes.
First check: compare traffic graphs against the failure time. Prevention means monitoring server performance, putting a CDN in front to reduce load, and scaling resources when needed, as the web.dev guidance above recommends. Google's SRE book adds that capacity planning should be paired with realistic performance testing, because it is hard to predict which resource runs out first without it.

DNS problems
DNS translates your domain into an IP address. When it fails, the server can be perfectly healthy and still unreachable. Misconfigured records, a bad nameserver change, a lapsed DNS provider or a propagation delay all produce the same result: visitors get a name-resolution error.
DNS can also fail at the provider level. In 2016 a DDoS attack on the DNS provider Dyn took down services including The New York Times, BBC, CNN, Fox News and The Guardian.
Symptom: "server not found" errors, one region fine and another broken, or an outage right after a nameserver or record change.
First check: query your records directly. Google's crawler documentation recommends inspecting DNS records with dig and analyzing network errors with tcpdump or Wireshark. Our DNS troubleshooting guide covers five commands that find most faults, and the step-by-step fix for a DNS misconfiguration takes you through the repair.
Expired domains and certificates
This is the most avoidable cause on the list. When a domain registration expires, the registrar severs the link between the domain name and its IP address, and Cloudflare notes that users can no longer reach the site through that domain. An expired TLS certificate does something similar in practice: browsers throw a full-page warning, and most visitors leave.
Symptom: the site vanishes or shows a security warning on a date that, in hindsight, was entirely predictable.
First check: look up the registration expiry and the certificate expiry. Turn on auto-renewal for both, and set a calendar reminder as a backstop for the contact email that renewal notices go to.
Plugin, CMS and code errors
Broken code takes sites down quietly. TechTarget lists broken code, plugin errors and update problems among the usual reasons websites crash, and Plesk notes that incompatible CMS plugins can cause outages. A WordPress update that conflicts with a plugin is the classic example, but any dependency upgrade can do it.
Symptom: a 500 error or white screen after an update, often only on certain pages.
First check: read the server error log, then disable the newest plugin or revert the newest dependency. Keep a staging copy so updates get tested before visitors see them.
Cyberattacks
DDoS attacks flood infrastructure with malicious traffic, overwhelming servers and exhausting bandwidth, as Cloudflare's guide to preventing downtime explains. Hacks and malware can also take a site offline, either because the attacker breaks it or because your host suspends it.
Symptom: a sudden flood of requests from many addresses, or unexplained file changes and a host suspension notice.
First check: look at request logs for an abnormal pattern and ask your host or CDN about mitigation. Basic hygiene helps too: CISA's guidance recommends cybersecurity basics such as multi-factor authentication. For background on why security and trust go together, see this piece on website security and customer trust.
Network and ISP problems
Sometimes your server is fine and the path to it is not. A routing failure or an ISP outage can make a site unreachable for a whole region. The visible sign is that some visitors fail while others do not, which is why checking from several locations matters.
Third-party dependencies
Your site may depend on a payment gateway, an auth service, a CDN, an API or a database host. When one of them stalls, requests pile up behind it and your own servers can follow it down. Google's SRE book calls this pattern a cascading failure.
What causes a website to time out
A timeout means the client gave up waiting. The server may be overloaded, a database query may be hanging, a proxy may be waiting on a slow upstream (see our 504 Gateway Timeout guide), or a firewall may be silently dropping packets. There is a client-side angle too. AWS's reliability guidance says clients should set explicit connection and request timeouts rather than rely on defaults, which can be infinite or higher than your service goals allow, and it warns that setting them too low creates artificial failures.
Symptom-to-cause quick reference
| What you see | Likely layer | First thing to check |
|---|---|---|
| Name not found, DNS error | DNS or expired domain | Domain expiry, nameservers, records |
| Browser certificate warning | TLS certificate | Certificate expiry and chain |
| Connection timed out | Server, firewall or network | Server load, firewall rules, traceroute |
| 500 Internal Server Error | Application code or config | Error log, latest deploy or plugin |
| 502 Bad Gateway | Proxy or upstream server | Upstream health, 502 guide |
| 503 Service Unavailable | Overload or maintenance | Traffic, capacity, maintenance mode |
| 504 Gateway Timeout | Slow upstream | Slow queries, upstream response time |
A short troubleshooting routine
- Confirm scope. Everyone, one region, or only you.
- Stabilize. Roll back the last change, shed load, or switch on a maintenance page, then dig.
- Identify the layer. DNS, network, TLS, server or application, using the table above.
- Fix and verify. Retest from outside your network, not only from your own browser.
- Write down what happened. One paragraph on cause, detection time and fix. Repeat causes show where to invest.
Our perspective: catch it before your users do
Most downtime is not exotic. It is a lapsed renewal, a rushed deploy or an overloaded box, and the damage grows with the time between the failure and the moment someone notices. That is why we built Pingfloat as a set of tools that sit close together: an uptime monitor, an SSL checker, traceroute, an HTTP header and status reference, and utilities for the small checks you run in the middle of an incident.
Our view is that monitoring should be the first thing you set up, not the last. Core tools work with no installation or account, so you can check a site in seconds, and signing up adds scheduled monitoring and saved results when you want ongoing coverage. Pair the monitor with a sensible alert setup (our guides on downtime alert checklists and alert thresholds and routing show how) and an outage becomes a notification, not a customer complaint.
— Andreas
Detecting downtime sooner
Prevention and early detection work together. Practical steps:
- Monitor from outside. Your servers cannot report that they are unreachable. External checks can.
- Alert on latency, not only on up or down. Google's performance guidance suggests alerting on response-time latency, especially at the 95th or 99th percentile, since slowness usually comes before failure.
- Track expiry dates. Domain and certificate renewals are predictable, so there is no excuse for them to surprise you.
- Test capacity before the campaign. Load-test ahead of launches and sales instead of discovering limits live.
- Remove single points of failure where the budget allows: a CDN, a second DNS provider, a replica database.
- Keep a rollback path for every deploy.
If you want a step-by-step setup, our walkthrough on uptime monitoring in 15 minutes is written for developers.
Start monitoring your site with Pingfloat
Pingfloat's Website Uptime Monitor checks whether your site is reachable from outside your own network, and you can pair it with the SSL checker and traceroute when something looks off. The tools are free to use, and the rest of the suite is on the Pingfloat home page.

FAQ
Why do websites randomly go down?
Usually it is not random. Common triggers are a traffic surge, a change someone deployed, a plugin update, a hosting or hardware fault, or a DNS or renewal lapse. Intermittent outages often point to overload or an unhealthy dependency, so check logs and traffic graphs around the failure time.
Why is my website down right now?
First check whether it is down for everyone: load it on a different network and test it from an external monitor. If it fails everywhere, look at the status code or error type to find the layer: DNS, certificate, server or application. If it only fails for you, clear your DNS cache and check your VPN or ISP.
How do I prevent website downtime?
Monitor server performance, put a CDN in front of the site, and scale resources before demand arrives. Add alerts for latency and for domain and certificate expiry, test releases on staging, and keep a rollback plan. These steps cover the most frequent causes.
How do I fix a website that is down?
Stabilize first: roll back the latest change, shed load or enable a maintenance page. Then find the failing layer with the status code, a DNS lookup, a certificate check and the server logs. Retest from outside your network once you apply a fix.
Sources
- 10 Website downtime causes and how to prevent them (Gatling)
- 7 reasons why websites crash (TechTarget)
- How The New York Times handled election-night traffic (Data Center Knowledge)
- Website downtime: what causes it (Plesk)
- How to keep a website from going down (Cloudflare)
- 503 Service Unavailable (MDN)
- Set client timeouts (AWS Well-Architected)
- Troubleshooting methodology (Google SRE)
- Cascading failures (Google SRE)
- Debug network and DNS errors (Google Search Central)
- Fix an overloaded server (web.dev)
- Cybersecurity best practices (CISA)
