Guides · Website · Outage
My Website Is Down: What to Check in 10 Minutes
Test the site from another network, note the exact message and check the domain and the certificate. These checks are often enough to tell who needs to step in.
The short answer
Start by opening the site from a phone on 4G or 5G, with Wi-Fi turned off. If it works, the problem probably comes from your network, your device or its cache. If it stays unreachable, note the address you tested and the exact message displayed.
A number like 403, 404, 500, 502 or 503 means a server answered. The problem is then most often in the site, its configuration or its hosting. A message like “server not found,” “connection timed out” or “connection refused” means the request stops earlier: domain name, DNS setting, network or an unavailable server.
Next, check the expiration date of the domain name and of the security certificate. Do not reinstall anything at random. Send your host or provider the exact message, the time you noticed it, the result of the test on another network and the latest changes you know of.
Rule out a problem limited to your network
Open the site from a phone on 4G or 5G with Wi-Fi off, then from a second browser. If it works in either test, leave the site alone: restart the router first, check the device’s clock and clear only the cache of the browser concerned.
If the site stays unreachable from several networks, copy the exact message. A numeric code proves a server answered. A message without a code can point to a domain, a DNS setting, a certificate, a network or an unavailable server. The table below helps you choose which lead to check.
Find the right contact
The message displayed does not always give the exact cause, but it tells you where to look and who can act. A parked-domain page, for example, shows that a server does answer — it is just no longer the one hosting your site.
| What you see | Did a server answer? | Most likely cause | Who fixes it |
|---|---|---|---|
| 500 — Internal Server Error | Yes | Error in the site’s code: incompatible plugin, invalid configuration file or rewrite rule | Your provider |
| 502 — Bad Gateway | Yes, an intermediate server | The application server behind it is down | The host on shared hosting, your provider elsewhere |
| 503 — Service Unavailable | Yes | Scheduled maintenance, or the hosting plan’s limit has been reached | The host or your provider |
| 403 — Forbidden | Yes | File permissions, or a web application firewall blocking the visitor | The host or your provider |
| 404 on every page except the home page | Yes | Lost rewrite rules (common with WordPress) | Your provider |
| “Parked domain” page or the host’s default page | Yes, but it is not your site | The domain no longer points to your hosting | The host and the registrar |
| “Account suspended” page | Yes | Unpaid invoice, resource overage, flagged content | The host, and only the host |
| “Server not found” | No | The domain name no longer exists in DNS | The registrar |
| “Connection timed out” | No | The domain points to an address, but nothing answers there | The host |
| “Connection refused” | No | The machine answers but closes the door, or is overloaded | The host |
| “Your connection is not private” | Yes, but blocked before display | Expired certificate, or your own computer’s clock | Check the device’s clock first, then the certificate |
See the technical detail: testing HTTP, DNS and the certificate
Test the HTTP response
Replace example.fr with the domain concerned:
curl -I --max-time 10 https://example.frNote the status code, the “server” line, any redirects and the time before the error. Repeat the test with http:// to see whether the redirect to HTTPS works. Do not use the option that ignores certificate errors to conclude that the site is reachable for visitors.
Check DNS
dig +short A example.fr
dig +short AAAA example.fr
dig +short NS example.frNo A or AAAA answer explains why no address is found. Then compare the answer from your usual resolver with a public resolver. If they differ just after a change, note the cache duration (TTL) before changing the zone again.
| What you change | Observed or maximum delay, depending on the case |
|---|---|
| Name servers of a .fr | 1 hour |
| Name servers of a .com | 48 hours |
| A name that has just been deleted (“does not exist” answer) | 10 to 15 minutes |
For a .fr, open:
https://rdap.nic.fr/domain/example.frThe useful fields are the registrar, the statuses, the name servers and the dated events. A recovery period means you should contact the original registrar immediately.
Read the certificate the server presents
openssl s_client -connect example.fr:443 -servername example.fr </dev/null 2>/dev/null \
| openssl x509 -noout -subject -issuer -datesCheck that the requested name appears in the certificate, that the end date has not passed and that the server presents the expected chain. A certificate that is valid as a file but missing from the server’s configuration does not clear the browser warning.
Check WordPress without making the outage worse
Before disabling a plugin, keep the files and the database. Check the administrator account’s email and the PHP logs. Then check, in this order: an interrupted update, the plugin or theme named in the error, the database connection and the rewrite rules. Test after each change: otherwise the cause becomes impossible to isolate.
Recognize the three common WordPress errors
WordPress shows some messages itself. They do not prove a general hosting outage: they point to the database, a plugin, a theme or an interrupted update.
| Page displayed | Trigger |
|---|---|
| “Error establishing a database connection” | Wrong credentials, a database that is down or overloaded, or a compromised site |
| “There has been a critical error on this website” | A fatal code error, most often an incompatible plugin or theme |
| “Briefly unavailable for scheduled maintenance” | The technical file that an update leaves in place while it runs |
After a critical error, check the mailbox tied to the administrator account. Since WordPress 5.2, a message may contain a link to recovery mode and name the plugin or theme behind the error.
The normal maintenance page stops being applied after about ten minutes. If the site is still unreachable after that, note the exact message and look for another cause instead of waiting longer.
What not to do
- Deleting plugins at random before keeping a backup and reading the exact message
- Ignoring the administrator account’s mailbox after a critical error
- Waiting indefinitely for a maintenance page to disappear on its own
Check the domain name renewal
For a .fr domain, the expiration date alone does not cause an immediate shutdown. In practice, the site becomes unreachable when the company that manages the domain asks for its deletion, often after an unpaid renewal invoice. That company is your registrar; AFNIC only operates the .fr registry.
After a .fr deletion request, the domain can still be recovered for 30 days from the same registrar. For a .com or .net, the timeline has a grace period, a recovery period and then deletion. In every case, contact the registrar named in the domain’s directory record without delay.
If you no longer know who manages the domain, use AFNIC’s public tool for a .fr or an RDAP service for other extensions. Note the registrar’s name, the domain’s status and its expiration date. A detailed reading of the technical response is in the panel above.
Check the clock, then the certificate
Start by checking the device’s date, time and time zone. A wrong clock can produce the same warning as an expired certificate. Then test from a second device. If the warning remains, the site’s certificate must be checked and renewed.
Since March 15, 2026, a public certificate can no longer be valid for more than 200 days. That limit drops to 100 days in March 2027, then to 47 days in March 2029. A manual yearly renewal is no longer enough: renewal must be automatic and monitored.
Understand why the old address is still visible
After a DNS change, some networks keep using the old answer stored in their cache for a while. The new value is already published at the source; you have to wait for each cache to reach the end of its validity period and ask again.
The delay therefore depends on the setting you changed and its cache duration. The “24 to 48 hours” often quoted does not apply to every change. Testing from several networks and with a DNS checking tool tells you whether a cache is still valid or the configuration is wrong.
Send your host a precise diagnosis
Note the exact address, the full code or message, the time you noticed it, the result of the test from another network, the latest known change, the domain’s status and the certificate’s date. Attach a screenshot without hiding the message.
Since [time and time zone], the address [URL] shows “[exact message]”. The same result appears from [other network/device]. The latest known change is [change, or “none”]. The domain is registered with [registrar] and expires on [date]. The certificate expires on [date]. Can you confirm whether the server receives the request and tell me which error your logs show at that time?Move from outage diagnosis to the incident procedure
Unknown files, new accounts, redirects or emails sent without any action on your part point to something other than an outage. Preserve the evidence before cleaning the site, cut off public access and follow the procedure for the first 72 hours after a hack.
What to ask your provider
What code or error do your logs show at the exact time of the outage?
A code, a timestamped excerpt and the service that produced it.
Are the server, the database and the domain name all active and renewed?
A status and a date for each of the three, not a global answer about “the site.”
Does the certificate renew automatically, and what alert is sent if the renewal fails?
The renewal mechanism, the next deadline and who receives the alert.
What monitoring detects downtime, who receives the alert and after how many minutes?
The tool’s name, the check frequency, the trigger threshold and a responsible person.
What maximum recovery time is written in our contract?
A number of hours or days stated in the contract, with any exclusions.
Sources
- IETF — RFC 9110, HTTP Semantics
Normative definitions of the 403, 404, 500, 502, 503 and 504 codes.
- IETF — RFC 9499, DNS Terminology
Confirms that “propagation” is absent from DNS’s normative vocabulary.
- WordPress.org — Common errors
The official causes of WordPress’s three error pages, including recovery mode since version 5.2.
- AFNIC — Charte de nommage, version du 5 mai 2025 (Naming policy, version of May 5, 2025)
Tacit renewal of a .fr and the deletion timeline.
- ICANN — Expired Registration Recovery Policy
The exact timeline for .com and .net after expiration.
- CA/Browser Forum — Baseline Requirements, version 2.2.9
The schedule for shortening TLS certificate validity.
- Let’s Encrypt — Official FAQ
Validity periods and caching periods for domain validations.
