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 (33%)
• Cloud Services (30%)
• Web Tools (15%)
• Hosting (15%)
• E-mail (7%)

Live Outage Map

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

City	Problem Type	Report Time
Paris	Cloud Services	6 days ago
New York City	Hosting	9 days ago
Manchester	Domains	29 days ago
Angers	Cloud Services	1 month ago
London	Domains	1 month ago
Noida	Hosting	2 months 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:

• 
  Kirk Patrick (@kirkfps) reported 8 minutes ago

  @cerebras We landed on almost exactly this architecture a bit over a month ago: unified Postgres embeddings, hybrid retrieval, reranker, and MCP serving LLM-free retrieval primitives with the client orchestrating. Where we went further: - Append-only canonical store before indexing, so the whole index rebuilds without depending on source systems - Formal provenance on every derived artifact (derived-from, source version, model, prompt, date), with source vs derived queryable and derivative indexing as explicit policy - Zero LLM calls in the query path. Retrieval-only worker; synthesis delegated via MCP to the assistants users already pay for - A verified output layer that makes hallucination inexpressible, not just unlikely. The retrieval contract is empty-never-nearest: outside the validity window the system returns nothing instead of the closest plausible match. Absence is a well-defined, provable result over a bitemporal canonical store, and abstention is typed (Found | Gap), gated by byte-exact verification against the source, outside the model. Most systems make hallucination improbable; this makes unsourced claims impossible to sign - Governance via the ARSIA Protocol, our open standard aligned with the EU AI Act and GDPR: query-time ACLs against the IdP, PII protection, EU data residency, auditable corpus - Edge on Cloudflare, data layer on dedicated OVHcloud servers. Fully European stack, replicable by any company without an inference fleet The result is a complete stack, closed at every layer: ingestion can't contaminate (canonical + provenance), storage can prove what existed and when (bitemporal invariants), and output can't assert what it can't verify. Correctness lives in the data layer, not in model behavior. And the entry cost is almost embarrassing: a small company can run the whole thing end-to-end on Cloudflare alone (Workers, R2, D1, Vectorize, Workers AI). The two-person deployment runs on roughly $5-10/month. No seats, no per-user pricing, no GPU fleet. Convergent evolution is the best validation there is.

• 
  q32 (@q32com) reported 8 minutes ago

  @TimJayas amazon, google and cloudflare have more than enough compute to host those models but they won't. and shouldn't. there is simply no way to red team an open weight model of that size. it could be full of embedded backdoors, and you would never know.

• 
  Raj Sitaula (@raj_sitaula) reported 9 minutes ago

  We've contacted Cloudflare support multiple times, but have received no meaningful response. Meanwhile, our reputation and readers are being harmed by what appears to be an abuse of automated reporting systems.

• 
  Chuck Reynolds (@ChuckReynolds) reported 9 minutes ago

  @TheFrosty @Cloudflare yup. falls back to the "hire slow, fire fast" scenario... i saw the writing on the wall early... /me out.

• 
  Lamb O'rghini (@ld35253) reported 11 minutes ago

  @unusual_whales Old news and cloudflare sucks

• 
  Shakib Ahmed 🇮🇪 🇪🇺 (@Magurmach) reported 11 minutes ago

  Seems like @Cloudflare verification is breaking @mygovid website login. Even Cloudflare's feedback submission is broken.

• 
  Rosario Tech (@RosarioTechSolu) reported 14 minutes ago

  Lesson from this week: partial-synced a static site with scp instead of the whole dir, missed some icon assets. Cloudflare edge-cached the resulting 404s for 4 hours — my token has no cache-purge scope. Fix: sync the whole dir always, bump a ?v= cache-buster on asset URLs.

• 
  joexiii 🏴‍☠️ (@j03xiii) reported 17 minutes ago

  @HamEggsnSam I do a lot different **** on my network though actually. Like I have a whole subnet just for my wife’s work computer (I don’t trust her IT guy, long story) that is entirely segregated from my main net and uses Cloudflare for DNS. My gaming network is also segregated and uses Cloudflare for reduced latency. Basically my net traffic looks normal and crazy at the same time which probably drives my ISP nuts. At least I like to think it does I also run IDS/IPS and have caught and stopped a least a handful of probes trying to get in. I’ll probably spin up a honey trap in the near future too just for fun

• 
  hokey type (@cygniius) reported 23 minutes ago

  hi @lastfm team, I hope you guys will consider adding a AI DDOS/scraping blocking feature (such as cloudflare or Anubis) to the site as I’m assuming the site is constantly being hit by AI scraper bots. Many other sites have implemented this already so not sure if backend issue?

• 
  可达鸭 (@hzl895f) reported 24 minutes ago

  ChatGPT Work has become the second brain for my small web studio. I use it to review restaurant sites, turn messy notes into clear Astro/Sanity build plans, troubleshoot Cloudflare issues, and prepare better Codex prompts. It has cut a lot of trial and error out of my week.

• 
  Andrew Mitchell (@4ndrew_mitchell) reported 26 minutes ago

  I put my site behind cloudflare, should help with anyone who ever had trouble accessing it before.

• 
  Thomas Schulz (@thomasschulzz) reported 26 minutes ago

  @danielbelfort @jacobmparis honestly the primary reason i pay for vercel. i know they got me covered. all these people talk about cheaper prices on cloudflare and whatever, but u dont get this kind of support. i'm in an early testing group for their product eve and they send PRs to fix issues immediately

• 
  Krishna Singh (@krishnasinghdev) reported 27 minutes ago

  @cheaf25master I will go with cloudflare rest of them charge less in early years but they go heavy after that also few does not support important configs, only concern with cloudflare is that you wouldn't be able to change nameservers. P.S. earlier I love to buy from Google domains now they sold the business to squarespace

• 
  Victor Barsoum (@VBarsoum) reported 30 minutes ago

  Hey @CloudflareDev @eastdakota now that CloudFlare is more cloud compute than CDN, any chance of limiting compute Geo for us miserable plebs in regulated industries (I know it is literally the opposite of what you stood for.. but that's your problem for building good product)

• 
  SerPepe (@SerPepeXBT) reported 31 minutes ago

  token usage down cuz we’ve been heavily throttled due to us being a “competitor” to cloudflare, we used them for most of our models. we just switched over to another provider and slowly rolling out update.

• 
  Akintola Steve (@Akintola_steve) reported 32 minutes ago

  @ub360_ai Ops… It was a Cloudflare issue I had to fix it on the spot immediately I noticed it. Thanks

• 
  Humoud Almunawer حمود المناور (@homoudalmonawer) reported 35 minutes ago

  Fix it in 3 steps: 1) Update WordPress to 7.0.2 (backports: 6.9.5, 6.8.6). Under 6.8 isn't affected. 2) Don't assume auto-update ran. Open the dashboard and confirm the version. 3) Behind Cloudflare? The WAF rule shields you since Jul 17, but a shield isn't a patch.

• 
  Naamã (@Naam36831652) reported 35 minutes ago

  @divinetrickr Just look at what Kiwi Farms is suffering now. Poor sods can't even get Cloudflare.

• 
  mRr3b00t (@UK_Daniel_Card) reported 37 minutes ago

  @OliverSild I guess the last 20 odd years have taught me: expect the unexpected... From msblaster to wannacry was over 10 years gap.... The reason why i try and be measured is because there is a lot worse in the world (entire) than WordPress sites being pwn3d. (not that i want them to be! I'm very happy with how this seems to have gone (so far) and I'm very glad your org, cloudflare and others also have helped make this the way it is (team sport) 20% of 5 million sites is still a big chunk but from my sampling it wasn't CNI - it was small things... my terrible weekend got to: ah this ones ok quite fast :)

• 
  Adam Bossy (@abossy) reported 38 minutes ago

  I wrote my own agent harness because existing ones to be far too limiting: - Claude agents SDK: closed-source - OpenAI agents SDK: immature until recently (e.g., lacked subagent support) - Pi/Flue/Vercel/Cloudflare: Typescript-oriented (prefer Python) - LangGraph Deep Agents: weirdly locks every prompt into using Task and Todo list - OpenCode: responses would stop mid-way inexplicably - Pydantic: not an AI company but looks worth trying ... I got sick of each one missing one vital feature. I wanted this for a Python-based agent I've been writing for months without switching to Typescript, while having powerful features like Pi's session tree capability and Vercel's UI widget support. So I wrote it by defining an extremely clean and modular agent loop with general interfaces that are likely to last many major model iterations. Then I had my coding agent research all the modules (memory, compaction, etc) that need to be plugged into the loop, I checked the work and was happy with the design choices the agent made. Coding agents do extremely well when they can interpolate designs across 5+ samples. Since agent harness patterns are stabilizing and they're mostly open source, I didn't have to do the painstaking work of designing each module individually. These aspects of agent harnesses will evolve dramaticaly, so it's also not worth the painstaking work based on my goal of quickly getting something up that's feature complete. I thought this would produce junk, but it worked surprisingly well on my first run. It's super clean, model-agnostic and Python-based, which is wonderful. I'll need to refine it, no doubt, but it's an incredible example of "custom software", even for something as critical as an agent harness.

• 
  Search Engine Land (@sengineland) reported 41 minutes ago

  Fix gaps in this order: 1. Eligibility 2. Trust 3. Relevance First check crawler access, robots.txt, Cloudflare settings, NAP consistency and structured data. New content has limited value if AI systems can't access or confidently interpret the site.

• 
  Chris (@burnerDevAcct) reported 42 minutes ago

  @jackfriks the obvious answer is lovable/bolt/replit and supabase (yes, i know about the supabase issues). cloudflare and 3rd party db signups are too much for someone just starting - and any of the desktop apps she will be in permission hell (ie clicking "always allow" 400 times).

• 
  Bulls & Barons (@BullsandBarons) reported 43 minutes ago

  Similar to @MEADGod and @ponsdotfamily, our opps (who know how bullish this is) submitted a completely unsubstantiated false report to @Cloudflare. It included no evidence or logs or anything. Unfortunately, Cloudflare acts on these reports immediately without reviewing them and then has a very long delay to appeal. The site is temporarily affected by this. An appeal is pending and if not answered soon, we will simply remove Cloudflare from the app entirely. To be certain, we reviewed the production code, player profiles, and available server logs and found no phishing content or credential collection whatsoever. The site never requests seed phrases or private keys and uses private wallets like any other app. Your wallets, tokens, and buildings are safe. An appeal has been submitted, and official updates will only come from this account. TLDR: Haters gonna hate.

• 
  max.berlin (@maxjendrall) reported 43 minutes ago

  @vibecodeguild You can! You can put out public pages haha on custom domains even. Goddmanit, they need to fix the presentation. They literally rolled out cloudflare workers across their customer basis that are soooo powerful

• 
  Elion's Pops ⌨️🚀💰 (@EseDevlin) reported 45 minutes ago

  @uche_ui vercel hosted abi? blocked on Glo network. You should route it through cloudflare to circumvent the problem

• 
  Jas (@keepontruckin78) reported 52 minutes ago

  If you ever wanted to roll your own cloud give Nextcloud a try. If you need help setting it up or want to try it first please drop me a message. I don't get paid any sponsorship for saying this. I genuinely use the self-hosted platform myself behind cloudflare on a daily basis and ditched cloud subscription paid services for good.

• 
  FatherXLdn (@FatherXLdn) reported 54 minutes ago

  Step 6: lock it down (do NOT skip this). Cloudflare Access in front of the tunnel: only your email gets a one-time login code. Nothing loads without it. Then verify Access's signed JWT server-side, don't trust the email header. Headers can be forged; signatures can't.

• 
  ⓙⓘⓑⓞⓢⓢ (@jibayarea) reported 55 minutes ago

  is there a cloudflare outage??

• 
  Hi I'm Mehdi • Infra Lead @ Ecomflow (@MehdiChioukh) reported 57 minutes ago

  @Cloudflare biggest problem is D1 being limited to 10gb, please offer a real scalable solution to a SQL database and it will solve soooooooo many problems

• 
  hrvoje (@0xHP10) reported 1 hour ago

  This is the 3rd post in my Hetzner server series. I wanted to understand where the setup actually starts bending: CPU? Postgres? API latency? memory? logs? something else? So I mapped the current architecture against different traffic levels, cache hit ratios, and failure points. The biggest takeaway: 10M users is not the scary number. The scary number is how many requests reach origin. If every page is dynamic, you're cooked. If Cloudflare, Redis, and precomputed read models absorb 95-98% of public reads, the whole thing stays way more efficient than people expect. Current setup is around 30 EUR/month. My best-case 10M user version is still Hetzner + Cloudflare: private network, load balancer, stateless app nodes, dedicated Postgres + replica, Redis, workers/queues, R2 for assets/backups. Estimated cost: around 800-1800 EUR/month if the product stays cache-heavy.