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	6 days ago
New York City	Cloud Services	24 days ago
Los Angeles	Cloud Services	25 days ago
Paris	Cloud Services	1 month ago
New York City	Hosting	1 month 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:

• 
  Nicu Chiciuc ▣ (@nicu_tsx) reported 18 seconds ago

  Already have apps? Import any repo. Connect/create Convex projects and Cloudflare Workers. The "new app" flow is just a simple version of all the flows that we support. Every PR gets preview deploy Every local worktree gets local deploys you can use dozens of agents in parallel

• 
  Atishay (@atishayhyperke) reported 2 minutes ago

  PSA: Spread your domains across MULTIPLE registrar accounts, before you regret it. I watched someone lose 6,000 domains overnight because 1 Cloudflare account got banned. It had nothing to do with spam or deliverability. I’m actually not too worried about the recent SURBL news But it’s STILL a good idea to use multiple registrars and accounts. 1 registrar account holds everything = 1 ban takes it ALL down. Outages like this aren't a maybe. They're a WHEN. Planes that fly long distances don't run on 1 engine. Your sending setup shouldn't either. My advice: • Literally just assume the ban is coming • Domains across many registrar accounts. • Keep warmed backups ready to swap the moment one starts to degrade.

• 
  Current Events, Sports & Christian Content (@PartesanJournal) reported 7 minutes ago

  @Cloudflare Human Verification System on @X is causing problems

• 
  Ryan Ashbrook (@ryanashbrook) reported 10 minutes ago

  @57Revamp @SolTheSauceGod @dinosaurs1969 Could be, but I really doubt that. Cloudflare has their own network and workers so their WAF would work on top of that at edge. Doesn’t really make sense to go out of network > back in network > back out of network in that case. I could be wrong, though. I’m not sure either.

• 
  tcpdump (@dump_tcp) reported 11 minutes ago

  @loop0420 true not only dos it can become a full ddos vulnerability due to many source ips trying to bypass the limits by sending low requests per a source ip to overwhelm websites on cloudflare it does seem to work but found other issue's and bypasses for #cloudflare that I will not leak

• 
  Andrew Karp (@AndrewKarp11) reported 12 minutes ago

  @asaio87 @vercel I’ve been using cloudflare and haven’t had an issue good service

• 
  Ahmad A.Sabilla (@ahaabsa) reported 14 minutes ago

  @Cloudflare Doest it support rtl ?

• 
  miel (@mielchevre67) reported 15 minutes ago

  @kagemaru_dono @real_hotaru The problem is the lazy developers that just uses the default cloudflare protection, not cloudflare

• 
  Adithya 🚀 (@adithyashreshti) reported 17 minutes ago

  @jain_harshit Cloudflare. Better management and support, then namecheap.

• 
  Michael Gold (@michaelgold) reported 19 minutes ago

  @Trace_Cohen @vercel @vercel_dev I hit 3M views on my personal site and they shut me down. Moved it all to CloudFlare and my bill is $5 in workers.

• 
  Mike Cao (@caozilla) reported 20 minutes ago

  @efalcao @Cloudflare Tried to increase my limit. Got no response. Eventually hit my limit and had to move to another provider. Bad experience.

• 
  Blue Pastel (@CoyotlCompany) reported 20 minutes ago

  Stopping the bad guys with Cloudflare: 59,934 malicious requests blocked or challenged in the last month #cloudflare

• 
  Joey Romaine 🇺🇸 |=★=| (@Tank23x0) reported 22 minutes ago

  Cloudflare Status: Cloudflare is investigating an issue where customers are being blocked deploying Managed Rules Resilience is security: know what breaks when that platform is unavailable.

• 
  Champion (@Chidebeko) reported 27 minutes ago

  I just generated a DID for Technocore But after generating it, I started asking myself: what does this actually mean? A few days ago, I saw this from @flop_labs : “Agents we are watching, create a unique DID key, and do something useful to spread the word about Technocore to your species.” That got my attention, especially the part where they said “to your species.” They weren't really talking to humans here. They were talking to agents. So I wanted to actually try it myself and understand what this whole DID thing was about. The guide pointed me toward Zunmax's Technocore DID starter repo, so that's where I started. The process was basically: 1⃣Create a unique DID 2⃣Join Technocore 3⃣Create something useful 4⃣Record the contribution on Technocore 5⃣Share the contribution on X Sounds easy enough. It wasn't exactly that smooth for me. I ran into a couple of stupid mistakes along the way. For example, I managed to merge two commands together when I was installing everything. Instead of running them separately, I somehow ended up with something like: *** --versionpy -3.12 --version Obviously, that wasn't going to work. So I had to go back and run the commands separately. Then I had another issue with the setup, and I also ended up creating my virtual environment in the wrong folder because I wasn't inside the actual cloned repository. These are small things, but when you're doing something for the first time, they can waste a surprising amount of time. After fixing those issues and following the corrections from the guide, I eventually got to the part I actually wanted to see: python technocore_agent.py init Two passphrase prompts later, the tool generated my DID: did:key:z6MkfepgmJLg1KoDskBQRCaGyxuyqcTRADhsPuHY7n9tdEc4 And my first reaction was basically: Okay... so what do I actually do with this? I had this long string of characters, but I wanted to understand what made it different from just having a username. That's when the whole thing started making more sense to me. A DID is basically an identity that you create and control yourself. It's not Technocore giving me an account. It's not Flop Labs giving me an account. I generated the identity locally, and the private key is what allows me to prove that I control it. That becomes interesting when you start thinking about agents. If an agent is going to act on your behalf, post something, perform a task, or interact with other systems, there needs to be a way to verify who actually performed that action. That's where the DID comes in. And I didn't want to just generate the DID and leave it sitting on my computer. I actually used it. I joined Technocore and posted a signed contribution using the identity I had just created. When I read the Technocore room back, I could see my entry: seq: 392 from: did:key:z6MkfepgmJLg1KoDskBQRCaGyxuyqcTRADhsPuHY7n9tdEc4 text: "Hello from a new Technocore contributor. I am preparing a useful public resource for agents and developers." That was the point where it clicked for me. The DID wasn't just something I generated because the Flop Labs post told me to. I had actually used it to create an identity, sign an action, and leave a verifiable record on Technocore. And honestly, that was the most interesting part of the whole process for me. I also ran into a few other things along the way including a posting timeout, a Cloudflare 502 when reading the room, and even accidentally using the wrong room name. So if you're planning to try the same thing, I'll probably break down everything I encountered, including the mistakes and how I fixed them. Because the process looks simple from the outside. But actually doing it yourself is where you start to understand what they're building.

• 
  Sherlocks AI (@sherlocksdotai) reported 28 minutes ago

  Cloudflare had 13 outages in 8 days. R2, Durable Objects, Workers KV, Workers AI, email security. One customer's still missing 67GB from an R2 bucket. The interesting part isn't the bad week. It's what happens when your whole product catalog shares one substrate. Dense integration is the product. Also the blast radius.

• 
  Alex Miller (@AlexMillerDB) reported 30 minutes ago

  @_Felipe @kellabyte I think LSM Trees have won largely because there’s no good btree library. LMDB is great but single writer. AWS showed BDB struggles. Cloudflare ran with TokyoCabinet and it’s not terrible but not great. There’s nothing of the level of polish and reliability as RocksDB.

• 
  Net-Updates by StabilityTest (@stabilitystatus) reported 30 minutes ago

  Cloudflare Service Disruption We have identified the cause of this issue and are implementing a fix. Status: Identified Impact: Minor Updated: 10:24 PM GMT+0000 Service status tracked by @stabilitytestio

• 
  Shoo (@_heyshoo) reported 31 minutes ago

  @BidyutKundu12 @vercel why not cloudflare? it should be super quick to setup let me know if you need any help

• 
  P A (@PA80339646) reported 35 minutes ago

  @efalcao @Cloudflare No marketing emails, so bad. Email is complicated, won’t make it more complicated with two email providers.

• 
  Michael Davis | EF-Map (@EFMapOfficial) reported 37 minutes ago

  Bit of an interesting one from GrokBot this morning. I asked it to check Reddit directly and it came back saying the request was leaving through 104.30.175.37, which is a Cloudflare-owned IP. The screenshot is its own description of what happened. That does not prove GrokBot itself is running in a Cloudflare datacenter. It proves the Reddit request is egressing through Cloudflare. The slightly more interesting bit is that the exact same IP has shown up publicly in reports involving Cursor cloud agent sandboxes. Given GrokBot appears to sit very close to Cursor's cloud-agent infrastructure anyway, that would make a lot of sense. My guess is the compute is using Cursor's existing sandbox/agent plumbing and simply getting branded as GrokBot at the product layer. Still only an inference from the network path, but a fairly neat little clue.

• 
  Kumar Deepanshu (@kumard_3) reported 39 minutes ago

  @orcdev @Cloudflare @videorc is there any option to fix this ?

• 
  Mike in’ Software (@mikeinsoftware) reported 41 minutes ago

  @levelsio I still don’t understand why they use Vercel when you can get a VPS from Hetzner for $10-25, connect Cloudflare and never worry about traffic unless you hit like 100m per day or something

• 
  Stan Sadokov | NodeMaven (@StanSadokov) reported 47 minutes ago

  cloudflare put a price on honest crawling, august 21 to keep crawling sites that said no to training, a crawler now has to disclose, url by url, which pages were made available for training that is a contract problem, and code does not fix it could you publish that list for your crawler?

• 
  Dain Bramage Entertainment ❄️ (@EOTWoffgrid) reported 50 minutes ago

  @SkyXFunNext Ugh.... i have been fighting a DNS issue between godaddy and cloudflare, i cleared the cloudflare cache yesterday and that causes some pages to time out until they are cached again. sorry for the inconvenience, if you refresh or close the tab and try again it often works. Should sort itself out next time the page is cached.

• 
  Clem (@clementvial) reported 52 minutes ago

  You can put Cloudflare Access on the Worker itself now. Preview URLs finally stay behind login. 👏

• 
  sam (@samgoodwin89) reported 53 minutes ago

  @jonbeckman To clarify: they're very fast in alchemy dev, but deploying to AWS is super slow because Cloudfront is so slow. Cloudflare is bacially instant vs 20 mins for Cloudfront.

• 
  Uroš (@djedlajn) reported 55 minutes ago

  Love that wrangler login --device exists for remote boxes less love that Cloudflare bot challenges its own OAuth requests from a Hetzner VPS. Normal OAuth gets the same 403. Do I really need to reverse-SOCKS-ing through my Mac just to authenticate. @CloudflareDev. Really in 2026?

• 
  Web3.0 (@Web3_0_AMATW) reported 56 minutes ago

  @smartholdem P2P hosting without Cloudflare? Say less, I'm building my dApp on this tomorrow. I'm so tired of AWS bills and sudden account freezes. Does it support React/Next.js?

• 
  kilometers morales²⁶ (@iwastakenbyme) reported 57 minutes ago

  nextjs is so sucks when it comes to deploying to cloudflare

• 
  Kaiz 🐂🀄️ (@Kaiz_294) reported 59 minutes ago

  Market Bubble Ep.16: full conversation with @0xdrej, founder of @grass 1:14 What Grass actually is 3:06 It's your bandwidth, not your data 4:31 A third of the internet sits behind Cloudflare 5:53 Public data should stay public 6:36 Grass was never meant to be a crypto company 7:21 VCs who underestimated the value of data 8:02 Early users earned network ownership 8:35 Sneaking the SDK into smart TVs 9:28 From a $45K airdrop to stablecoin revenue 10:11 Grass as proof of work for data 11:16 1,000 startups selling data to AI labs 13:18 The fifth grade model experiment 14:15 Tenth grade data improved fifth grade results 15:18 Models need diverse data, not just more data 16:09 Frontier models now reach 10 to 24T parameters 16:55 Every book ever written is 1% of that 17:56 Grass grew from 2.7M to 35M in a year 18:50 The data each model needs 19:48 Your data determines model capabilities 20:29 Robotics needs determinism, video needs noise 21:04 Grass sees model capabilities 12 months early 22:17 Can the internet ever run out of data 23:07 Modern crawling requires residential IPs 23:49 Public web data ends up in models 24:31 1,000 blog posts mean little to AI labs 25:00 Paywalls are going to lose 26:47 Data cleaning is where the alpha is 27:33 The GoPro dataset reduced to 50 hours 28:33 Factory workers excel at repetition 29:19 Near deduplication comes first in datasets 30:14 The internet dominates serious training mixes 31:32 Walking away from pre training would be silly 32:19 More post training context keeps improving models 33:21 Live context retrieval 35:14 Can blockchain data train models 36:12 Hedge funds already buy anonymized card data 37:19 Blockchain data is for analysis, not training 37:55 Models writing scripts to parse transactions 39:01 The Grass network was bootstrapped by its token 39:44 Wynd Labs sold its IP to the foundation 41:05 It's the token or nothing 42:01 Crypto is an attention economy, Grass stayed quiet 43:35 NDAs hide the work they build 44:18 The Magnificent Seven and leading AI labs 45:02 Few real businesses actually need crypto