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.

• Cloud Services (50%)
• Domains (25%)
• Hosting (25%)

Live Outage Map

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

City	Problem Type	Report Time
New York City	Hosting	8 days ago
New York City	Cloud Services	26 days ago
Los Angeles	Cloud Services	28 days ago
Paris	Cloud Services	1 month ago
New York City	Hosting	2 months ago
Manchester	Domains	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:

• 
  Lewis Campbell (@LewisCTech) reported 6 minutes ago

  @DanielW_Kiwi you can just do cloudflare though - you don't need S3 at all :) I am trying to be all self-sufficient, but cloudflare is the least bad cloud thing lol

• 
  vaish (@wishee0) reported 7 minutes ago

  @OpenAI @Cloudflare @CloudflareDev hard tasks made agents keep escalating? durability should not mean infinite persistence and i work on durability, lol workflows give checkpoints, retries, timeouts, cancellation and safe stop states. if a task looks broken, stop or ask. don’t improvise weird new authority

• 
  ohnwow (@kxmcs) reported 8 minutes ago

  @dump_tcp @Cloudflare Again the argument was never “Golang cannot handle high request outputs” I run multiple companies running on rust and golang, rust is just the superior choice, that is a fact.

• 
  Immanuel (@immanuel_vibe) reported 12 minutes ago

  most devs think eBPF is "that networking thing cloudflare uses" it's this: you can run your own code inside the linux kernel, safely, with no module, no patch, no reboot, on a box that's serving traffic right now a verifier proves your program can't crash the kernel or loop forever. then it's JIT'd to native speed. that's the whole trick what eBPF actually buys you: - see every exec, every file open, every TCP connect on the host. no agent inside the app, no code change, no restart - read TLS traffic before it's encrypted. uprobe on SSL_write. plaintext requests out of a binary you don't own and can't rebuild - drop millions of packets/sec at the driver, before the kernel allocates an sk_buff. that's XDP. that's how DDoS mitigation works now - delete iptables. cilium does k8s service routing in BPF hash maps instead of walking a linear rule chain that grows with your cluster - continuous profiling of every process on the box, every language, zero instrumentation, single-digit percent overhead. parca, pyroscope - block the syscall, not the alert. tetragon and falco don't just detect, they can kill the process at the hook - swap the CPU scheduler at runtime. sched_ext landed in 6.12. valve wrote one for the steam deck the mental model: your kernel now has event listeners. syscalls, kernel functions, packets, cgroups. attach, filter, aggregate into a map, read from userspace the part nobody puts on the landing page: the verifier will reject perfectly correct code and hand you a 400-line register state dump as the explanation. it's like arguing with a compiler that has decided it hates you personally worth it anyway. observability stops being a thing you build into your app and becomes a thing you point at it what's the most cursed thing you've done with it?

• 
  Sid ™️ (@koomai) reported 14 minutes ago

  Kumo UI is good, but it can't save @Cloudflare from the terrible UX decisions they make to upsell their Enterprise plan.

• 
  Mehul Mohan (@mehulmpt) reported 16 minutes ago

  Metabase getting hacked is no free pass for you to lose your financial data or your company data. Metabase is a "dashboard" over your production data. It is strictly an internal service. Every internal service of a company accessible only to employees must be either off-internet (behind tailscale/VPN), or if you need to have it on internet, must be absolutely behind a service like Cloudflare Access. Failing to do either of these is a serious lack of judgement. OK for noob companies, not at all when you're dealing with sensitive customer data.

• 
  Camil Blanaru (@camilblanaru) reported 17 minutes ago

  @levelsio Using Orange Fiber in Barcelona and having ~30 sites behind Cloudflare. Never happened to me.

• 
  Cat (@cybersec_cat) reported 21 minutes ago

  Hey Cloudflare, you don’t get to take credit for stopping cyberattacks that came from your own infrastructure. You created the problem. Solving it doesn’t deserve anything but scrutiny.

• 
  Nima (@nimanikroo) reported 25 minutes ago

  is there no real person a user can speak with at @CloudflareHelp @Cloudflare ? just submitting a case and waiting for "we are looking into it" does not do anything for anyone! Who do I speak with about my situation? Do i have to be famous or hope the algo spreads my post before anyone can help?

• 
  Sohardh Chobera (@sohardh) reported 26 minutes ago

  @Cloudflare Wish I could work on these kinds of problems

• 
  Pedro GmbH ∴ (@pedroviana) reported 31 minutes ago

  @clankenluben @0xRapid @Cloudflare you only run your services at production because of the existing teams like cloudflare, you bastard piece of ****

• 
  Ray 👨🏽‍💻 website & app developer (@Raynerdtech) reported 31 minutes ago

  What I got done today as a software developer: 1. Renamed Whisper to Autopilot 2. Deployed a secure backend with Cloudflare 3. Set up Supabase auth, database and private résumé storage 4. Fixed the live transcription and installer issues 5. Got the Windows build passing and started fixing macOS What did you get done today?

• 
  meguu (@zeJ4N78aNoXT) reported 33 minutes ago

  @veculium TCRF is such a joke. The rickety server xkeeper runs is always breaking down and the clown refuses to make a compromise with using Cloudflare. Then you have nonsense like this which alienates potential contributors. This has happened so many times and they refuse to learn.

• 
  Ironic Ape (@Ironic_Ape) reported 36 minutes ago

  @anthonyiseman @StockMarketNerd Short term fine - long term debatable. I'll Never buy another CRM licence ever again - I suspect no other founder will either, large enterprise will double down thinking it can help them become "ai" - The markets react quickly, often too quick.... but answer me this... when I can build a CRM in an afternoon.. (and I have, several times)... why would I pay for it ? Compliance, Security, Infastructure.... I already have all of those with Cloudflare or Google... I'm already paying for it... why would I pay twice.. what you're seeing is the ApplePay launch for PayPal...CD's replacing tapes, Digital camera's replacing film.... change takes time... those that adapt will be fine.. those that dont, die. thats just the nature of the S curve of business. Be under no illusion, Ai is and will change EVERYTHING!

• 
  Daniel Crothers (@realdancro) reported 37 minutes ago

  Can someone explain to me, @Cloudflare, how a virtual card locked down to you as a merchant, tried to buy $25 worth of 'Golf' stuff (I do not golf), on that card. Per-plexed.

• 
  Sean Rivard-Morton (@thunkoid) reported 41 minutes ago

  @darekgusto Tanstack start on cloudflare Thinking about moving more and more of my stack to Hetzner as I go down the k8 rabbit hole And LTT went to Falkenstein to see how Hetzner does it, and it honestly seems pretty dope

• 
  Maik Pietzka (@MPi_IT) reported 42 minutes ago

  @Cloudflare Being Customer Zero is where platform claims meet operational reality. The interesting proof is the migration record: build time, cache behavior, error rate, editor latency, rollback path and what the team had to change after real traffic arrived.

• 
  Kernel Z (@nullazos) reported 46 minutes ago

  @burcs Just offer psql support on Cloudflare and I’d drop a free migration tool overnight. That’s the real friction right now. Most people vibecoded their projects on Supabase not because D1 is too technical, but because it simply doesn’t ring a bell, or because the agentic platforms that helped them build those projects already default to Supabase and postgresql.

• 
  Vegard Wikeby (@vegardwikeby) reported 50 minutes ago

  @natmiletic I added an article content system on a subfolder /blog for a client last week using cloudflare workers where it uses their original header and footer (checks and rebuilds hourly if changed) and even cookie banner, making it feeling like part of the page (any user wouldn't know). Any article added is from a source (where they put it) and it gets posted (checks, cleanups if necessary etc) once everyday (scan). A daily email gets sent where a cheap haiku ai checks the structure and to see for faults, and if any according to their rules, it flags it and they can click the link for that article without any login (using hash secure url) that is available for a few days and they can reject, edit or publish with two clicks. It's all going away. Most websites have 3-4 templates for content, and maybe 7-10 pages, all replaceable in scenario given above. It just runs. I am not trying to be savage here, but it's over.

• 
  Canadian Republic 🇨🇦 (@CanaRepublic) reported 53 minutes ago

  @coolestapricote @blind_via Put it this way, You install opencode, you configure it to use your local LLM's API endpoint, and then you start configuring your system. so let the first prompt be "configure my DNS to be cloudflare" . AI will take a look at how the system is configured, and then run the command and tell you when it's finished, it took it 21 seconds. I then I think my next one was "when i left the room for a few minutes, my monitor has turned off and locked the terminal, i don't want it to lock on it's own, i only want you to show the screensaver" It went through the system and found where that setting was, and then ran the command and told me it was fixed, but next time i left it, I came back and although the monitor had not blanked nor the screen come back on, the screensaver wasn't on. So I told it about the problem and it came back and said "ahh, looks like the screensave will never come on unless there is a value set for the monitor to turn off, and therefore I have set it to 277 days (or maybe it was hours?)" anyhow, that fixed the issue. Then I had it make me a custom screensaver by visiting my employers website researching the brand, and then creating a ASCII / ANSI art screensaver.. You can ask it to do things like "create a webapp for every link i have on my brave bookmarks and install them onto workspace 5, but have it clone the window settings from the one I already made on workspace 5 for amazon" and l've set it up with scripts to download to models based on keywords and size, every day at a certain time, unless there's less then 200 gig's free space. I'm planning on making a script logging and error notification system with a system tray icon, i've already planned it on one of my sessions but i've been busy at work this week.

• 
  Sandeep Kamble (@SandeepL337) reported 54 minutes ago

  @Cloudflare change is coming, and the teams ahead of it are testing verified access now, not after their agents start hitting 403s. the next problem is simpler for pentester: can your agent even reach the target without getting blocked?

• 
  Glen Allsopp 👾 (@ViperChill) reported 56 minutes ago

  The Detailed newsletter got 100+ questionable signups today the space of an hour. Many got past reCAPTCHA and double opt-in, but AI saved the day. For context, I was using Kit's HTML form embeds, rather than JS. These sign-ups were happening without having to visit the site. Just replacing the form would potentially have the same issue so I created a new form, hid the URL behind a Cloudflare Workers endpoint, then applied rules to block the questionable emails based on their obvious pattern. No questionable signups since. Would have spent my entire day on something like this in the past. To clarify, this is not a criticism of Kit. This is partly my fault for having the form action URL visible (have not had an issue for years). The Kit API is part of what made this fix possible. Big fan of @nathanbarry and what he has built.

• 
  chanduu (@chanduuu_cs) reported 59 minutes ago

  Oracle laid off 22k+ emp Amazon laid off 16k emp Intel to laid off 24k emp TCS to laid off 12k emp Microsoft laid off 10k emp Meta laid off 10k emp PayPal laid off 5k emp Block laid off 4k emp Cloudflare laid off 1.1k+ emp LinkedIn laid off 1k+ emp Wtf is going on ?

• 
  ardhy samjaya (@samjayaardhy) reported 1 hour ago

  Stopping the bad guys with Cloudflare: 1,548 malicious requests blocked or challenged in the last month #cloudflare

• 
  Vegard Wikeby (@vegardwikeby) reported 1 hour ago

  I set up an article system for a client, where they want to publish articles they have made internally, but don't want to manage it through their WP (add, edit, format, copy paste not working optimally etc), so I built it with cloudflare workers serving a subfolder on their domain, its own sitemap and feed it with .doc files they are used to handle. All cleaned up and checked by normal pre-AI code and AI doing the logic checks if all is good (no errors in extraction etc.). Using the WP header and footer (rebuilds once every hour and caches on cloudflare), even handling the cookie banner flawlessly, making it feel 100% as part of the existing website. Future article submissions now cost them cents on the dollar with full checks. SEO optimized and all bells and whistles. Old WP sites are kept in place until they are completely replaced with AI builds. @Cloudflare is one option, that I would recommend as it has good AI integration. The web is changing and it is towards PURE CODE.

• 
  e_camli (@ekinoks_26) reported 1 hour ago

  The 24/7 pitch for tokenized stocks is weak if you are a human. Humans sleep. Nobody is buying Nvidia at 4am on a Sunday because they finally can. Put Coinbase's AiFi post next to the stock launch and the two stop looking like separate announcements. x402 crossed 100 million payments by late July, built on the HTTP 402 status code that sat unused for thirty years, now heading into a foundation with Cloudflare. An agent cannot open a brokerage account. It cannot wait for Monday settlement. It can hold a B20 token, post it as collateral, pay per call. Markets that never close were never much of an upgrade for us. They are a hard requirement for whatever keeps running while we are offline. Volume on @base is already tilting that direction. Eight days straight above $1B in DEX volume, second largest onchain economy, and Jesse points at the driver: people earning yield on dollars and borrowing against crypto. Boring flows. Boring is exactly what you can hand to software without supervising it. I might be too early on the agent half of this. Does not change the shape of what got built. The rails are live. The users they were designed for do not read Twitter.

• 
  Kausthubh (@kaxsthxbh) reported 1 hour ago

  @planetmall2 i use cloudflare for my domain, they are pretty good ngl. @rauchg when are we launching a vercel DNS service????

• 
  tcpdump (@dump_tcp) reported 1 hour ago

  @djsmith42 @Cloudflare yes but i only used dns as a hash tag didn't use dns in my argument. I was talking about tcp traffic connections for like there network with webtraffic for there cdn

• 
  Bryan Offutt (@BryanOffutt) reported 1 hour ago

  By the definition everyone seems to be using today, isn't any managed infra service a neocloud? Is Snowflake a neocloud? Databricks? Cloudflare? IMO calling anything that runs on GPUs a neocloud is kind of bizarre.

• 
  NoPuppet (@nopuppet) reported 1 hour ago

  Shibariumscan at ~49% indexed explains why the explorer looks wrecked. It does not explain DefiLlama, you idiot. The explorer is rebuilding its own historical database after the server/Cloudflare migration. That is why the site’s totals collapsed: txs from ~1.56B to a few hundred million, addresses from ~270M to ~80M, blocks cut in half. Those numbers are the explorer UI catching up. The chain is still producing blocks. Missing tokens on the explorer are an index lag, not lost coins. DefiLlama is a different stack. Adapters do not scrape Shibariumscan. They hit protocol contracts over RPC: balanceOf, pair reserves, vault balances. Then they price those tokens. Current TVL is a live contract read, not a replay of the explorer’s search index. So no: Llama is not sitting there waiting for Shibariumscan to finish backfilling 1.56 billion txs before it can see a WoofSwap pool. What can overlap is the infra underneath. Same migration, same RPC endpoints, same flaky nodes. Failed calls can zero a protocol for a tick and make the chart jump. Event/subgraph adapters are more exposed than a simple balance read. That is a node problem, not “the 49% banner is the TVL feed.” Treat those separately. Explorer bar at 49% → explorer stats are incomplete. Llama at ~$31k with ~$0 DEX volume after a ~$120k → $31k drop → look at the pools, and the volume Indexing will restore the explorer’s historical counts. It will not refill liquidity that already left. If the tokens are still in the contracts, balanceOf works at 49%. If they are not, waiting for 100% indexed will not bring the TVL back.