
To check whether Cloudflare is down, open cloudflarestatus.com and compare it with what your own store returns. A 5xx error page that names Cloudflare as the failing component, while your origin server answers normally and the status page shows an incident, points to Cloudflare. A 521 or 522 error usually means Cloudflare is working and cannot reach your server. In that case the fault is in your hosting or firewall, and switching Cloudflare off will not fix it.
If Cloudflare itself is down, most stores should wait and communicate. Bypassing the proxy is a real option, but only if you prepared for it: your origin has a publicly trusted certificate, your firewall accepts direct traffic, and someone can still change DNS. With Black Friday on 27 November 2026 and Cyber Monday on 30 November, there are two weeks left to put that plan in writing and test it once.
Cloudflare publishes detailed postmortems, and three of them from the last twelve months show what store operators should expect.
On 18 November 2025, a database permissions change doubled the size of a Bot Management configuration file. The proxy software hit a feature limit and returned HTTP 5xx errors across Cloudflare's core CDN and security services. Impact began at 11:28 UTC, the main impact was resolved at 14:30, and Cloudflare reported all systems normal at 17:06. Cloudflare wrote that the dashboard was "mostly operational," but "most users were unable to log in due to Turnstile being unavailable on the login page." That matters for any fallback plan that depends on clicking through the dashboard.
On 5 December 2025, a WAF configuration change caused HTTP 500 errors for customers on Cloudflare's older FL1 proxy who had the Managed Ruleset deployed. Cloudflare put the affected share at about 28% of its HTTP traffic, with impact from 08:47 to 09:12 UTC.
On 20 February 2026, a bug in an automated cleanup task withdrew about 1,100 Bring Your Own IP prefixes. Affected visitors saw connection timeouts. Cloudflare's timeline puts impact between 17:56 and 23:03 UTC. Some customers could restore service themselves in the dashboard. Others could not, and Cloudflare engineers restored them by hand.
None of these was an attack, and each was fixed. They show that a Cloudflare incident during a peak sales hour is a realistic scenario to plan for, the same way you would plan for a payment provider outage. If you have not mapped which of your services depend on Cloudflare, the third-party dependency map workshop is a good place to start.
Work through these steps in order. Most of them take under a minute.
curl -sv --resolve shop.example.com:443:203.0.113.10 https://shop.example.com/ -o /dev/null, replacing the IP with your origin's address. If this returns 200 while the public URL fails, the origin is healthy.| Situation | Likely owner | What to do | Bypass proxy? |
|---|---|---|---|
| Error 521 or 522, Cloudflare status green | Hosting / your developer | Restart the web server, check firewall allows Cloudflare IP ranges, check server load | No |
| Error 520 or Cloudflare-branded 502/504 | Your developer | Read application and server logs, check last deploy | No |
| Cloudflare incident confirmed, short and improving | Store owner | Pause campaign sends, post a status note, wait | Usually no |
| Cloudflare incident confirmed, hours long, during a sale | Owner plus developer | Run the prepared bypass if every precondition is met | Only if prepared |
| Dashboard login failing | Developer | Use the API token stored outside Cloudflare | Via API only |
| Storefront up, orders stopped | Developer | Check checkout, payment webhooks, Turnstile or bot rules on checkout pages | No |
The last row is easy to miss. Cloudflare's November 2025 report lists Turnstile among the affected products, so a store that uses Turnstile or bot rules on login or checkout forms can lose orders while product pages still load. That is the "green uptime, zero orders" pattern described in WooCommerce orders stopped.
Cloudflare's proxy status documentation is direct about the cost. When a record is DNS-only, Cloudflare answers with your server's real IP, which "exposes your origin IP address to anyone who queries the record." Cloudflare also "cannot optimize, cache, and protect those requests." Your origin then takes the full Black Friday load without the cache in front of it, and with no WAF rules.
Three preconditions decide whether a bypass works at all:
If any precondition fails and you cannot fix it calmly, do not bypass. A half-done bypass replaces one outage with certificate warnings and an exposed server.
Fill this in now and keep a copy somewhere that does not depend on Cloudflare (a shared document, a printed page, a password manager note).
CLOUDFLARE OUTAGE RUNBOOK: <store name>
Last tested: <date>
CONTACTS
- Decision owner (can approve bypass): <name, phone>
- Technical owner (can execute): <name, phone>
- Hosting support: <URL, phone, account ID>
FACTS
- Origin IP(s): <stored here, not in public docs>
- Hostnames behind Cloudflare: shop.example.com, www.example.com
- Origin certificate type: [ ] public CA [ ] Cloudflare Origin CA
- Origin firewall: [ ] open to all [ ] Cloudflare IPs only
- Turnstile / bot rules on checkout: [ ] yes [ ] no
- API token with DNS edit scope: stored in <vault entry name>
STEP 1 DIAGNOSE (first 10 minutes)
[ ] cloudflarestatus.com checked, incident ID: ____
[ ] Error code on storefront: ____ Ray ID: ____ Time (UTC): ____
[ ] Direct origin test (curl --resolve) result: ____
[ ] Server load / logs / recent deploy checked
STEP 2 COMMUNICATE
[ ] Pause scheduled email and SMS campaigns linking to the store
[ ] Pause or lower paid ads budgets if outage exceeds <X> minutes
[ ] Post note on social / status page
STEP 3 BYPASS (only if all preconditions are YES)
[ ] Public certificate installed on origin YES / NO
[ ] Firewall rule for direct traffic prepared YES / NO
[ ] Origin can handle uncached peak load YES / NO
[ ] Decision owner approved at <time>
[ ] Set records to DNS-only via API or dashboard
[ ] Test checkout end to end with a real payment method
STEP 4 RESTORE
[ ] Cloudflare incident resolved on status page
[ ] Records back to proxied
[ ] Firewall restored to Cloudflare-only
[ ] Test order placed, order email received
[ ] Note written: start, end, orders affected, what to change
Subscribe to Cloudflare status notifications too. The status page supports email, webhook, Slack, Discord, Google Chat and, where available, SMS, filtered by service or location.
The numbers in this example are made up to show the decision, not taken from any real store.
A WooCommerce store selling homeware sits behind Cloudflare with proxied DNS, a Cloudflare Origin CA certificate and an origin firewall that accepts only Cloudflare IP ranges. At 10:05 local time on Black Friday, the storefront starts returning a 500 error page that names Cloudflare. The agency's developer checks cloudflarestatus.com and finds an incident under investigation. A curl --resolve test against the origin returns 200 in under a second, so the server is fine.
The owner asks about bypassing. The developer goes through the runbook: the origin certificate is Origin CA, and the firewall blocks direct traffic. Fixing both under pressure would take longer than the typical incidents in Cloudflare's own reports, and the result would be an unprotected origin with no cache during the biggest traffic hour of the year. They decide to wait. The owner pauses the 10:30 email send and posts a short note on the store's social accounts. Cloudflare resolves the incident at 10:50. At 10:55 the developer places a test order and confirms the order email arrives. Then they resend the campaign.
The week after, the agency installs a public certificate on the origin and writes a firewall rule that can be enabled in one command. Next time, bypass will be a real option.
Limitations: you cannot fix a Cloudflare incident from your side. A bypass helps only when the outage is long and your origin can carry the load alone. If you run on a hosted platform, the infrastructure decisions belong to the platform, and the Shopify outage checklist applies more closely than this runbook.
Check cloudflarestatus.com, then read the error code. Errors 521 and 522 usually mean Cloudflare is up and your origin is unreachable. A direct curl --resolve request to your origin IP tells you whether your server responds without Cloudflare in the path.
Only if you prepared: a publicly trusted certificate on the origin, a firewall that will accept direct traffic, an origin that can handle uncached load, and a working way to change DNS. Without those, waiting is usually safer.
Sometimes not. During the 18 November 2025 incident, most users could not log in for part of the outage. An API token with DNS edit permission, created and stored in advance, gives you a second route, but the API can be affected by an incident as well.
Last reviewed: 2026-11-14
Pingvera runs one-minute HTTP checks from multiple regions with quorum before alerting, so a single-location blip does not page anyone at 3 a.m. On WooCommerce stores, its order anomaly detector alerts when the site is up but orders stop, which covers the checkout-only failures described above. A white-label status page on your own domain gives customers somewhere to look while you work through the runbook.
Pingvera watches whether an online business actually works — uptime, checkout, orders, domain, SSL and server — and alerts you in Telegram, email or a webhook before a customer has to tell you.
Start freeRead next: Is Shopify Down? How to Check and What to Do During a Shopify Outage · Third-Party Dependency Map for Ecommerce: Template and Workshop.
Read next: Payment Gateway Down? A Store Backup Plan for Peak · Gift Card Fraud Prevention for Online Stores (2026) · Holiday Return Policy for Online Stores: 2026 Setup Guide · How to check if a website is really down · Check availability from several regions.