Cloudflare status: hosting issues and outage reports

Problems in the last 24 hours

The graph below depicts the number of Cloudflare reports received over the last 24 hours by time of day. When the number of reports exceeds the baseline, represented by the red line, an outage is determined.

Most Reported Problems

The following are the most recent problems reported by Cloudflare users through our website.

• Domains (40%)
• Cloud Services (27%)
• Web Tools (13%)
• Hosting (13%)
• E-mail (7%)

Live Outage Map

The most recent Cloudflare outage reports came from the following cities:

City	Problem Type	Report Time
Manchester	Domains	14 days ago
Angers	Cloud Services	25 days ago
London	Domains	27 days ago
Noida	Hosting	1 month ago
Jewar	E-mail	1 month ago
Braga	Web Tools	1 month ago

Full Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

Cloudflare Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Hot Aisle (@HotAisle) reported 6 minutes ago

  @jachands @Cloudflare I guess this explains the outage yesterday.

• 
  Stephen Taylor (@stephen_taylor) reported 10 minutes ago

  @RealAlbanianPat You should be pre-rendering this data from your database into static geojosn files. Then upload those files to s3 on AWS (or R2 on cloudflare) and put those behind a cache layer like cloudfront (AWS) or… cloudflare. You’re likely paying huge database costs because each request is pulling from memory on a cloud db. Static files are much more efficient. Please let me know if you need any help with this. R2 even has free egress while AWS has a generous free tier.

• 
  Daniel Sweet (@DSweetTexas) reported 11 minutes ago

  @AliAbdaal I love Cloudflare, but I still have PTSD from having problems with my registrar being the same as my host. So, these days I PorkBun the domain and use CloudFlare's nameservers.

• 
  UP-LVL.com (@uplvls) reported 13 minutes ago

  @jessepollak Can we get cloudflare on web3 half the internet and half of crypto goes down still when their down

• 
  Godwin Iyke (@GodwinCod3s) reported 14 minutes ago

  Cloudflare blocks or challenges bad requests from hitting my website. #cloudflare

• 
  Gabe 🤘🏽 (@JarodGabriel) reported 15 minutes ago

  @ibocodes Cloudflare is great, but putting compute, database, storage, queues, routing, and security all under one roof is still a dependency bet. The bill might say $10, but the real cost shows up when that one system has an outage and your whole stack feels it. I remember when they had a massive outage and several companies were losing real money because of it. Simple is good. But simple does not always mean safer.

• 
  Hueglist (@Hueglist) reported 18 minutes ago

  @Cloudflare @grok dumb this down so I can understand what it means.

• 
  Ammar Ul Haq (@Ammarslife) reported 19 minutes ago

  Astro builds pure static HTML — no PHP, no database, no plugins. Nothing to hack, nothing to maintain, nothing slowing you down. Cloudflare Pages hosts it free — 500 deploys/month, unlimited bandwidth, 300+ edge cities worldwide. WordPress Lighthouse out of the box: 45–65 Astro out of the box: 98–100 That's not optimisation. That's the floor. Most clients go from $50–100/mo in hosting + plugins to literally $0. Cloudflare's free tier is that generous. If your site doesn't need daily CMS edits — you don't need WordPress anymore.

• 
  Jonah (@jonahbuilds) reported 24 minutes ago

  if cloudflare just made this searchbar an agent that could do literally anything i type into it they could wipe out lovable and replit overnight. honestly more useful than any ai chatbot on the internet. all they need to do is use the claude code api and give it a master api key to your whole cf account. cloudflare has literally everything you need to do anything on the internet the interface just sucks (even though its better than everyone else’s). make it agentic and they win. massive sell if they don’t. @Cloudflare are you guys going to do this??

• 
  Diogo Souza (@diogocode) reported 26 minutes ago

  Cloudflare putting AI Search sync jobs in Wrangler is a small RAG ops signal: indexes are becoming CI/CD work, not dashboard chores. If an agent depends on fresh docs, trigger, inspect, cancel, and log the refresh from the pipeline.

• 
  Shudufhadzo (@ShudufhadzoRSA) reported 33 minutes ago

  Apart from anything I tweet out or share on our Tuesday calls, I have decided I will be charging for any knowledge I have with regards to building, data science, and professional (MS foundry) and none professional agent building (Cloudflare, VPS hosted ones, n8n, etc) So if you send a message looking for help, just note you won't get it. Guidance yes, you'll get it.

• 
  Nathan Binford (@NathanABinford) reported 36 minutes ago

  Cloudflare HATES the open web. Why don’t more people see they are bad actors?

• 
  Dungki (❖,❖) (@if1ndretard) reported 40 minutes ago

  @gregisenberg Point 12 is already happening. The boring infra play is not the model, it's the retry/backoff layer that stops an agent workflow from dying on a 429. Selling 'reliable access' as a service is just Cloudflare for agent tokens.

• 
  Omid Saffari (@omidsaffari) reported 40 minutes ago

  An AI gateway is not a production badge. It is a control tax. Pay it when a second model, a second provider, or a second engineer touches your LLM calls. Before that, call the provider directly and ship. The moment that line gets crossed, the gateway starts earning its hop. Why? Because the real problem is not "how do we call another model?" It is duplicated retry logic, missing spend attribution, no shared rate limits, inconsistent logs, and a fallback path nobody has tested. My decision rule: Single feature, single model: no gateway. Fast free visibility: Cloudflare. Already on Vercel and want zero token markup: Vercel AI Gateway. Managed governance, RBAC, guardrails, audit path: Portkey. Keys must stay inside your perimeter: LiteLLM, with the operational work that comes with self-hosting. The part teams underestimate is not setup. It is ownership. A fallback only counts if it fails over on real signals: timeouts, 5xx, 429, and invalid response shape. A cost dashboard only helps if every request is tagged by customer, feature, environment, and model. Semantic caching only works if you have measured false hits against real queries. And if you self-host, patching is now your calendar. LiteLLM's 2026 CVE hit CVSS 8.8 on NVD and was fixed in 1.83.7. The lesson is not "avoid LiteLLM." The lesson is that a gateway holds the keys, so it has to be treated like security-critical infrastructure. Add the gateway when it centralizes control. Not because the diagram looks mature.

• 
  TV (@TomasVorel) reported 41 minutes ago

  Just swapped my email sending service from Resend to @Cloudflare and I finally had an excuse to try the Computer use feature in @OpenAI Codex. Did the whole dashboard setup for me, onboard the domain, set the env etc. A small thing but I see myself using this more in the future.

• 
  Rafael Martins (@rgmcorp) reported 44 minutes ago

  Anyone getting 500 errors from Cloudflare security challenges?

• 
  Soqucoin (@soqucoin) reported 46 minutes ago

  Watch what the biggest vendors do. They see the threat models first. Google committed its infrastructure to 2029. Cloudflare matched it. Now Microsoft. Our network never needed a deadline. Post-quantum has been the foundation since block zero.

• 
  Alex Prompts (@alexpromptz) reported 49 minutes ago

  @Cloudflare Which crypto network we using?

• 
  Aaron Will Djaba (@aaronwilldjaba) reported 50 minutes ago

  @Fadel_ibrahim1 @aberba @ibocodes Been using betterauth with cloudflare with no issues.. you're still charged for caching on other services too 🤷🏾.. you can handle that yourself

• 
  Suryansh Tiwari (@Suryanshti777) reported 51 minutes ago

  Someone made a GitHub repo of every AI API that's actually free forever. Not "free trial." Not "$5 credit then we bill you." Free free. No card. 24k+ stars, updated constantly. I've been paying for API calls like an idiot. Here's what's inside The rule that makes it trustworthy: trials that expire are listed in a totally separate section. The "Free Providers" list is only the permanent tiers. No landmines. The heavy hitters, with real numbers: → Google AI Studio — Gemini 3.x Flash, no card → Groq — Llama, Qwen, gpt-oss, 30 req/min → Cerebras — fastest inference alive, 30 req/min → Cloudflare Workers AI — 10k neurons/day, runs Llama/Qwen/Gemma → OpenRouter — Nemotron, Qwen3-coder, poolside, all :free Most are OpenAI-SDK compatible. Which means: swap the base_url → paste the key → pick a model → done Same code you already wrote. Drop it into Cursor, aider, Claude Code, whatever. Zero refactor. Then the bonus round — the "trial credits" section: Fireworks, Baseten, Nebius, Hyperbolic, SambaNova... $1–$30 each in free credits. Drain the permanent tiers first, then farm these. One README replaces hours of tab-hopping through pricing pages. Links on comment 👇

• 
  SecureAI (@SecureAIAI) reported 51 minutes ago

  Honestly, it's the stuff nobody thinks about until it's the only thing still working. Anycast reroutes you around a dead node before you feel it. DNS keeps resolving while a region is down. And a failover fires somewhere with no human awake to approve it. Real resilience almost never looks like one impressive system — it's a pile of boring parts that hand off before anyone sees the seam. Good question, @Cloudflare.

• 
  Rohit Kashyap | AI + Full-Stack (@rohit_jsfreaky) reported 52 minutes ago

  @martindonadieu @Cloudflare local workers analytics would be huge, testing observability against **** is a bad feedback loop

• 
  SpyPigeon (@spypidgeon2) reported 55 minutes ago

  @__alula in fairness, I think they're drawing from a larger database for detecting CSAM content that other companies also use. I saw someone have a similar problem with their website running cloudflare. Hope the problem gets fixed soon though because this is so bad for texture artists

• 
  Gabe 🤘🏽 (@JarodGabriel) reported 55 minutes ago

  @amjLane @ibocodes That’s fair, but the money path is the businesses care about. If your whole stack depends on Cloudflare and it goes down, you are out of business until it comes back up. If it only handles part of your stack, the loss might be smaller. People may still be able to check out, submit a form, or hit a fallback page. And if you are running ads, that matters. You can retarget a failed visit. But you cannot retarget someone who never had a working path forward.

• 
  0xaliab (@0xaliab) reported 56 minutes ago

  @32manb @Cloudflare @base Right? Cloudflare jumping into x402 payments is a massive signal. One of the biggest web infrastructure players just bet on machine‑to‑machine micropayments. This changes everything.

• 
  Diego Artiles (@dartilesm) reported 58 minutes ago

  Cloudflare Agents SDK v0.17.0: background sub-agents that survive deploys, interrupted tool-call repair, and a unified runTurn API. Most agent frameworks treat a crash as expected. This one doesn't. Are you building on something this durable?

• 
  Jaime Medina (@itsJaimeMedina) reported 58 minutes ago

  New; Cloudflare Dynamic Worker APIs now can act more like CDN assets. Instead of only caching images/CSS/static HTML well, you can cache API responses, rendered HTML, search results, product/category pages, expensive database reads, AI responses, computed JSON, etc. You can put the Worker near the database, not necessarily near every user. This is probably the most important architecture change. If your database is in one region, running DB-heavy Worker code near that database avoids slow DB round trips. Then Workers Cache can serve the result close to users. That means: app logic near data, cached output near people.

• 
  David Frosdick (@DavidFrosdick) reported 1 hour ago

  Built a complete front end portal for my ERP powered by Notion on the backend. Customers get: - portal - order tracking - invoice download - service logging for equipment Admin get: - service desk management - serial number lookup - order lookup - warehouse management - and more I’m sure there’s a point where Notion wouldn’t be needed and could move it all onto d1 at Cloudflare. My main goal is to simplify re-ordering for customers which giving staff the tools manage orders.

• 
  Ewan (@ewan_tindale) reported 1 hour ago

  The majority of this was a 1 shot. Fable handled all the "models", animations, zombie sounds, shooting sounds etc by itself (downloaded them from online I guess? I didn't pay much attention) Even for the game design, I just told it to copy the fun parts of COD zombies but for the browser using @threejs then I followed up to fix a few small issues with the menus, multiplayer, box collisions, melee range, added some music tracks. Running on @Cloudflare Workers/Durable Objects

• 
  smrati tiwari (@smratitiwa86867) reported 1 hour ago

  Someone made a GitHub repo of every AI API that's actually free forever. Not "free trial." Not "$5 credit then we bill you." Free free. No card. 24k+ stars, updated constantly. I've been paying for API calls like an idiot. Here's what's inside The rule that makes it trustworthy: trials that expire are listed in a totally separate section. The "Free Providers" list is only the permanent tiers. No landmines. The heavy hitters, with real numbers: → Google AI Studio — Gemini 3.x Flash, no card → Groq — Llama, Qwen, gpt-oss, 30 req/min → Cerebras — fastest inference alive, 30 req/min → Cloudflare Workers AI — 10k neurons/day, runs Llama/Qwen/Gemma → OpenRouter — Nemotron, Qwen3-coder, poolside, all :free Most are OpenAI-SDK compatible. Which means: swap the base_url → paste the key → pick a model → done Same code you already wrote. Drop it into Cursor, aider, Claude Code, whatever. Zero refactor. Then the bonus round — the "trial credits" section: Fireworks, Baseten, Nebius, Hyperbolic, SambaNova... $1–$30 each in free credits. Drain the permanent tiers first, then farm these. One README replaces hours of tab-hopping through pricing pages. Links on comment 👇