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 (41%)
• Cloud Services (25%)
• Hosting (16%)
• Web Tools (13%)
• E-mail (6%)

Live Outage Map

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

City	Problem Type	Report Time
Manchester	Domains	10 days ago
Angers	Cloud Services	21 days ago
London	Domains	23 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:

• 
  Zubair Ibn Zamir (@ZubairIbnZamir) reported 7 minutes ago

  @Cloudflare @CloudflareDev Workers Builds queue stuck: pushes to GitHub main keep creating build records but they stay in queued and never move to running. 3 builds stuck now, oldest 15+ min. Anyone else affected today?

• 
  Simon Høiberg (@SimonHoiberg) reported 10 minutes ago

  When you move to self-hosting your products, security automatically becomes the number 1 concern. I used to have CloudFront distributions sitting in front of every public endpoint. Another popular solution is using CloudFlare tunnels and Tailscale. The only issue is - it's still "cloudy". And personally, I wanted to get rid of that. Fortunately, I found a surprisingly simple setup that can do roughly the same as CloudFlare tunnels - but fully self-hosted. Let me show you 👇

• 
  🅑🅛🅐🅝🅒🅞🅢 🅚🅗🅘🅜 (@blancoskhim) reported 11 minutes ago

  SIEM platforms keep promising universal telemetry, but the data they can digest is only as good as the connectivity breadcrumbs users emit. Azure AD audit logs, an hour of Cloudflare rate-limit blocks, and a handful of IIS scans processed five minutes after a denial of service feels resilient on paper but not under pressure. Vendors patch over parsing limits with more throughput specs, since engineering integration timelines miss the ops reality of batch-delivered cloud raw streams. Budget watch starts fading when organizations pour yearly outlay into a solution to sequence fifteen states but lack hourly human validation. Those raw spans hosted by archival setups avoid notice until a DC cut strands response with a forty-minute vault wall. Persistent alerting skew (cross account aggregate vs per risk false drop) makes root cause tallying elastic throughout halves any productive maintenance window. #cybersecurity Architect logging like a proof by exhaustive case inspection; the dull part functions better unhyped versus fragile. Compression can justly follow defensible live baseline logic driving shift-left triage usefulness given people coverage churn bandwidth cost reality tradeoffs. Keep implementation practical.

• 
  DO-SAY-GO (@realdosaygo) reported 13 minutes ago

  you can't. you'll spend a month chasing down all the things. then get insta banned by CloudFlare or Datadome. I'm deep in this tech and it still took me 300 hours. Even if you're a 10x engineer vs me, you're still droppin 30 hours on this. At 200-500 an hour that's 6-15K. Or you could just pay me the equivalent of 30 minutes, and you can have it right now. Up to you, bad ;)

• 
  Nitewalkar (@Nitewalkar) reported 13 minutes ago

  Day 4 of building with Grok Build and using Openclaw to manage what I build. We have made; - Fully Functioning POS App - tailacale - pull from open inv - create new sale - return/refund - Android ready. - Fully Functioning Ops App - work orders - forms - compliance docs - service agreements - calendar - team message board (avail on web) - cloudflare w/ms auth - upload purchases Clawbot Learning - email booking -> work order and draft invoices - email purchase receipt -> update PO, draft purchase/bill - email scan and monitoring - calendar management Websites - hosted docker on backend - rebuilt exclusovely with build - need fine tuning and revision then go live!!!!!!! Next - shared inbox/alias config for custom ms auth logins on one license ms365... ● Saving $700/year in GoDaddy. ● 1.19%/transaction on POS. ● 1-2h/day in admin time saved with OPS ● Clawbot monitors/manages space on always on pc. ● Build script monitors drive and pc it lives on with cron script. Reads logs and fixes issues. SEND MOAR CREDS BROS!!! THIS IS AMAZING!!!!!!! Entrepreneurs ********* @grok @xai @openclaw

• 
  KANAPURO 🎭 TEAM COMEDY (@kanapurottv) reported 13 minutes ago

  @Cloudflare pls fix workers bro pls pls psl psls pls

• 
  AwesomeAI (@Awesome_AI_News) reported 14 minutes ago

  Cloudflare has released new service regulations requiring all AI vendors to separate search crawlers from training/agent-specific crawlers by September 15th. Mixed crawlers accessing pages with advertisements will be automatically blocked. This rule applies uniformly to new customers, existing users creating new sites, and all free users; website administrators must manually modify backend configurations to allow crawling, directly affecting the standardization of AI crawlers in the industry. Cloudflare 发布服务新规，要求所有 AI 厂商在 9 月 15 日前拆分搜索爬虫与训练／代理专用爬虫。未区分的混合爬虫访问带广告页面将被自动拦截。该规则对新入驻客户、老用户新建站点及全部免费用户统一生效；网站管理员若要放行，须手动修改后台配置，直接影响 AI 行业爬虫规范。

• 
  Kun Chen (@kunchenguid) reported 16 minutes ago

  i hope 2026 is the last year where we still have to manually click through any website to set things up in the last month, google cloud and app app review are the two repeated offenders that still need manual click-throughs - bad by contrast, github, cloudflare, hetner etc are pretty much entirely configurable by agents - good (why not computer use / browser automation? because i don't want to expose secrets in plain text and let the agent type them via keystrokes and capture them into screenshots)

• 
  Owaish Alam (@owaish3301) reported 17 minutes ago

  @singhal_swaraj @striver_79 I think the problem is from cloudflare

• 
  I..A..N..S (@IANSYT) reported 18 minutes ago

  @ATTHelp can you tell someone on network engineering to look into high RTT and loss on the ATT-Cloudflare interconnect in dallas AS7018 <-> AS13335 because this is unacceptable, < 1Mbit/s speeds sustained for over a week now

• 
  Thomas Gauvin (@thomasgauvin) reported 22 minutes ago

  @joelrunyon @nickgraynews @Cloudflare can confirm humans decided on the new ui for emails we moved email routing from a feature of a domain, to a feature of email service which is a top level product (which includes both routing and sending) what setup steps are unclear? what "locks" are you encountering?

• 
  world Evolution (ALT) (@EvolutionALT) reported 22 minutes ago

  4/8 Popular Repos (2026): yonggekkk/Cloudflare-vless-trojan (very active, has obfuscated versions) Surfboardv2ray/Trojan-worker & v2ray-refiner vfarid/v2ray-worker derivatives These support Trojan over WS+TLS on 443.

• 
  AlloyPress (@AlloyPress) reported 22 minutes ago

  Hey @Cloudflare, several sites using Cloudflare's login verification are stuck in a verification loop, repeatedly asking users to reverify without letting them through. Looking forward to a quick update.

• 
  Easyjose (@Onlyhumanme) reported 30 minutes ago

  @world_xyz @worldnetwork @Cloudflare Quite a poor branding and comms. Undermining other just to gain traction.

• 
  Turnpike (@turnpike402) reported 30 minutes ago

  The announcement from Cloudflare today is a huge step forward for the x402 ecosystem. Part of the problem at this early stage is convincing AI companies that they'll need to pay for what they scrape - CF is a loud voice telling them otherwise.

• 
  Aditya🌪️ (@aditya4f) reported 33 minutes ago

  - Claude = coding ($20/mo) - Supabase = backend (Free) - Vercel = deploying (Free) - Namecheap = domain ($12/yr) - Stripe = payments (2.9%/transaction) - GitHub = version control (Free) - Resend = emails (Free) - Clerk = auth (Free) - Cloudflare = DNS (Free) - PostHog = analytics (Free) - Sentry = error tracking (Free) - Upstash = Redis (Free) - Pinecone = vector DB (Free) Total monthly cost to run a startup: ~$20 There has never been a cheaper time to build. Who's stopping you?

• 
  UWillC (@uwillc) reported 34 minutes ago

  Half the internet blinked last week. The cause was a backhoe, not a model. June 22. A fiber cut on Zayo routes rippled into Cloudflare. X, Reddit, Zoom, Teams. Down. X alone passed 30,000 outage reports before it cleared in about 20 minutes. Every AIOps dashboard in those companies watched a problem none of them could fix. You cannot reroute around a cut you do not own. You cannot ask an agent to splice glass three states away. We keep automating the control plane. The physical plane stays one excavator from an outage. Your multi-cloud is a logical diagram. Underneath it is often a single carrier. An AI can monitor the fiber. It still cannot splice it. Your redundancy on paper: single-carrier underneath, yes or no?

• 
  UWillC (@uwillc) reported 35 minutes ago

  Half the internet blinked last week. The cause was a backhoe, not a model. June 22. A fiber cut on Zayo routes rippled into Cloudflare. X, Reddit, Zoom, Teams. Down. X alone passed 30,000 outage reports before most services recovered in about 20 minutes. Every AIOps dashboard in those companies watched a problem none of them could fix. You cannot reroute around a cut you do not own. You cannot ask an agent to splice glass three states away. We keep automating the control plane. The physical plane stays one excavator from an outage. Your multi-cloud is a logical diagram. Underneath it is often a single carrier. An AI can monitor the fiber. It still cannot splice it. Your redundancy on paper: single-carrier underneath, yes or no?

• 
  Easy简单点 (@Easyidea_) reported 36 minutes ago

  @Cloudflare I still think that a small group can not build a protocol- level product like x402 payment. Unless it can solve specific problems in B2B system.

• 
  tamimbuilds (@tamimbuilds) reported 39 minutes ago

  - Claude = coding. ($20/mo) - Supabase = backend. (Free) - Vercel = deploying. (Free) - Namecheap = domain. ($12/yr) - Stripe = payments. (2.9%/transaction) - GitHub = version control. (Free) - Resend = emails. (Free) - Clerk = auth. (Free) - Cloudflare = DNS. (Free) - PostHog = analytics. (Free) - Sentry = error tracking. (Free) - Upstash = Redis. (Free) - Pinecone = vector DB. (Free) Total monthly cost to run a startup: ~$20 There has never been a cheaper time to build.

• 
  David Frosdick (@DavidFrosdick) reported 40 minutes ago

  Been putting Cloudflare pages to use today. @NotionHQ database on the backend. Customer shops built for brands on the front end. Hold about 120 products. Protected login, stripe checkout or checkout on account. Customer account approvals. Order confirmation emails and invoices. All built so staff can manage products prices from inside Notion setting markup on cost price, customer account management and more. I might start switching my smaller Shopify sites over to this as it’s easier to manage for small ecom stores with 100 products.

• 
  Richard Radermacher (@_radermacher) reported 41 minutes ago

  @djgeisi For the most affected page, Cloudflare is not in use. I encountered errors with the HTTP method and DNS. It became even more problematic when I needed a single certificate that included both the www and non-www domains. For example in one case netcup is used.

• 
  Kshitij Shah (@stuli1989) reported 41 minutes ago

  Just got a massive Cloudflare bill because of a vibe coded CF bug - damn, this is how vibe coding can bite you in the ***. @CloudflareHelp - could you be generous and give me a one time waiver please?

• 
  Hira Siddiqui (@identityonchain) reported 43 minutes ago

  <Rant Ahead> Every major AI company is building memory right now. OpenAI just shipped Dreaming V3, which updates your ChatGPT profile automatically after each conversation. Cloudflare launched Agent Memory so AI agents can store context between sessions. X released an official MCP server so agents can read your posts and activity in real time. All of this is genuinely useful. But none of it works together. ChatGPT's memory stays in ChatGPT. Cloudflare's agent memory stays with whatever agent you built on Cloudflare. X knows what you post but that doesn't help Claude understand who you are. Every product is solving memory for itself, inside itself. Which means you're still re-explaining yourself constantly. Your job, your preferences, your current project, your writing style. Every new AI tool you try starts from scratch. If you use five AI tools, you have five separate versions of "you" floating around, none of them in sync. The reason this won't get fixed by the big players is pretty simple: memory is how they keep you around. The better ChatGPT knows you, the less likely you are to switch to something else. That's not a conspiracy, it's just product logic. Sharing memory across tools would hurt retention, so nobody does it. What actually needs to exist is a memory layer that sits outside any individual product. Something you own, that you control, that any AI tool can read from if you give it permission. Not because the companies agreed to share your data, but because the memory never belonged to them in the first place. MCP is already starting to act as the connection layer between AI tools. The infrastructure for retrieval exists. The auth patterns exist. The missing piece is a persistent store that any agent can plug into, that travels with the user rather than living inside any one product. Before DNS, every network handled naming differently and nothing connected cleanly. Then one standard emerged and suddenly the whole thing scaled. AI memory feels like it's at a similar point. The question isn't really whether something like this gets built. It's who builds it and whether it's actually user-owned when they do. </Rant Over>

• 
  mugiwara no mean face (@m3anf4ce) reported 44 minutes ago

  Now cloudflare blocking me from watching donghua. If they take away my last bit of peace in this ******* country, I can promise I will be a problem.

• 
  Rahul Karajgikar (@Road_Kill11) reported 48 minutes ago

  so i tried self hosting observability myself on AWS turns out codex + terraform + aws + cloudflare is insane was able to set up clickhouse running on an m6g.large instance, connect it to EBS volume (for local disk) and s3 (cold storage), set up all the ALB/VPC etc. network stuff, and wire it up to my domain on cloudflare with very little effort - about 3 high level steps and about 30 mins of active effort: 1. step 1: asked codex to tell me all the api key/manual configuration i need to do. here i had to set up aws iam, and do some manual work on cloudflare to create apps/api keys/separated for dev/****, and connect the domain name i wanted to the clickhouse endpoint i would use after step 1, my agent was able to access my aws account with ssm after i signed in. 2. step 2: prompted a /goal to setup the entire infrasturcture stack using terraform + scripts, then deploy clickhouse + OTEL collector on the ec2 instance. codex then worked for 5 hours, wrote the entire code to set up the infra in terraform, deployed it with terraform CLI + some custom scripts, verified/tested that clickhouse was deployed and working fine with the local ssm connectivity after that, i asked codex to write some scripts for me to monitor my ec2 instance health, test scripts for verifying that logs/metrics/traces/spans etc. work properly and a few other scripts for making it easy to monitor/maintain the instance then i wired it into @dexto_ai, which was pretty easy, i just had to swap out the collector config to use my new clickhouse endpoint. it's working pretty insanely, and now i no longer need to deal with arbitrary caps on traces anymore. i can just scale my ec2 instance up when i need it, and use my startup AWS credits ($67 a month). mostly i cared about traces/spans anyway, and less about fancy monitoring features, so this is an incredible unlock compared to paying over $100 + usage based pricing on sentry/datadog etc.

• 
  🔥Phoenix (@GPhoenixForever) reported 52 minutes ago

  @LilithDatura Kind of like encryption with lava lamps at Cloudflare, noise vs signal down to the quantum fluctuations.

• 
  Vivek Kotecha (@vbkotecha) reported 56 minutes ago

  Most websites now have llms.txt for AI discovery. Almost none have x402 payment manifests. Discovery without payment is free-riding. The next layer is a per-tool contract. Identity. Price. Cap. Receipt. A /.well-known/x402.json file declares your service, your prices, your spending caps, and your receipt format. A Cloudflare Worker handles the 402 challenge and HMAC verification. Discovery makes you findable. x402 makes you payable. Findable without payable is a business model that does not close.

• 
  marcelo mezquia (@IntentSim) reported 1 hour ago

  Stopping the bad guys with Cloudflare: 15,448 malicious requests blocked or challenged in the last month #cloudflare

• 
  One&OnlyAarav (@WaterAarav) reported 1 hour ago

  Claude = coding. ($20/mo) Shypmenta = fully automates platforms below($6/yr) Supabase = backend. (Free) Vercel = deploying. (Free) Namecheap = domain. ($12/yr) Stripe = payments. (2.9%/transaction) GitHub = version control. (Free) Resend = emails. (Free) Clerk = auth. (Free) Cloudflare = DNS. (Free) PostHog = analytics. (Free) Sentry = error tracking. (Free) Upstash = Redis. (Free) Pinecone = vector DB. (Free) Total monthly cost to run a startup: ~$20. Building has genuinely never been this affordable, and rarely this effortless either.