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 (18%)
• Web Tools (11%)
• E-mail (5%)

Live Outage Map

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

City	Problem Type	Report Time
Noida	Hosting	11 days ago
Jewar	E-mail	12 days ago
Braga	Web Tools	12 days ago
Noida	Cloud Services	13 days ago
Paris	Cloud Services	13 days ago
Prievidza	Domains	13 days 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:

• 
  Tanner Scadden (@TannerSDev) reported 3 minutes ago

  @AWSSupport DMd. 19:27 we got the notice, saying take action by 06/22/2026. At 19:29 our production environment went down due to cloudflare being disabled. Please help us asap

• 
  DiPT (@DiptamoyBarman) reported 5 minutes ago

  Wow!! I just came to know about Cloudflare storage rn. Implemented within an hour. Previously I was using google drive to avoid the cost (got into quota problem and was trying some bs workarounds). 🤡

• 
  Jon C. Phillips (@joncphillips) reported 14 minutes ago

  @mattwensing I tend to agree with this, but the way some interfaces are so bloated, finding the actual correct series of buttons to press is not a great experience. Every time I login to Cloudflare, I hate the experience. I have not logged in to Jira or 1Password in a while now. I mean, there's amazing UIs out there, but there's also some real garbage. And for the garbage ones, I'd rather use an MCP or CLI or API to connect to it.

• 
  Patrik (@PatrikTheDev) reported 14 minutes ago

  the DX of Cloudflare has so many issues...I just spent an hour debugging why email sending doesn't work... I have acces to two cloudflare accounts and only one of them had email sending set up...and wrangler dev chose the other one. Wrangler is rough man

• 
  Usama Ehsan 🥶 🇵🇸🇵🇰 (@usama_two) reported 15 minutes ago

  @Cloudflare @CloudflareDev @CloudflareHelp Please fix this fast. My websites are used by 1000s of users, I cant afford even a hour of down time. I have already paid for invoice but I am ready to pay for it again. Just add pay now button

• 
  Phil | Rentier Digital Automation (@rentierdigital) reported 16 minutes ago

  your claude.md is already wrong you write a CLAUDE.md once and assume it's done. then your codebase evolves. dependencies shift, ports change, schemas multiply. the file sits there looking authoritative while everything around it drifts turns out /init isn't a generator. on repos with existing docs, it's an auditor. reads your CLAUDE.md, cross-references against lockfiles and configs, surfaces what stopped matching while the code moved forward i ran it on 4 repos. repo 1 had npm commands documented but was running on bun.lock. wrong port hardcoded. both had been true once, both silently false now. nothing broke bc npm commands mostly work on bun projects anyway. you just try the wrong URL then the right one, forget about it repo 3 was undocumented. multi-frontend catalog system, Astro frontends, Hono backend, SQLite. no README. /init did 4 parallel tool passes and reconstructed the entire delivery chain: SQLite schema, export pipeline, Astro build under Node 22 (not Bun bc of SSG renderer constraints), rsync to nginx, Cloudflare Workers routing. one pass. specific enough to navigate without asking repo 4 had 64 lines of CLAUDE.md. /init found 5 silent lies. package manager mismatch. schema states (3 documented vs 7 real). routing layer that was actually a full processor. undocumented mode switches. port number wrong in 3 places documentation drift doesn't announce itself like a build error. it's corruption in a save file. the game keeps running, the numbers look fine. then the boss fight hits and the stats don't add up the work isn't generating boilerplate. it's finding the delta between what you wrote and what the code actually became i build and ship daily with Claude Code. SaaS, tools, automations. ⭐ if AI can build it, I've probably broken it first. what works → link in bio

• 
  Rohan Paul (@rohanpaul_ai) reported 25 minutes ago

  Bots have overtaken humans online, and the internet was never prepared for this. - bots generate 57.4% of worldwide HTML requests - humans at 42.6%. The biggest shift is economic: traffic can rise while monetizable human attention falls, which weakens CPM, CPC, conversion-rate models, and analytics built before this agent wave. Cloudflare measures bot traffic, not only agentic AI traffic, but AI agents are a major suspect because one user task can trigger thousands of machine visits. The old web assumed a human loaded a page, saw ads, clicked links, filled carts, and created signals that publishers, stores, and SaaS companies could price. AI agents break that model because they read pages on behalf of people while skipping the ad views, session time, and click behavior that funded the web.

• 
  smallshen (@smallshen2) reported 25 minutes ago

  @dillon_mulroy Does tunnel support *.my-tunnels.mydomain.com ? Custom host name instead of try cloudflare.

• 
  dom (@dom51143825) reported 26 minutes ago

  @Cloudflare Will you support flex tiers in AI gateway soon? That's the only thing preventing me to switch.

• 
  Adel Bucetta (@adelbucetta) reported 35 minutes ago

  @wishee0 @Cloudflare @CloudflareDev the honest answer is that most interns don't get to work on scale problems, let alone get to work with the workflows team at cloudflare. kudos to you for having the opportunity

• 
  James (@jamescoder12) reported 39 minutes ago

  The neighbor's final advice was the most actionable. He sat down and wrote out a list of 6 things every internet customer should do: 1. Turn off the public Xfinity hotspot (or your ISP's equivalent Spectrum, Optimum, and Cox all do this too) 2. Manually set your Wi-Fi channel instead of "Auto" 3. Disable QoS / Smart Network "optimization" features 4. Change your DNS to Cloudflare (1.1.1.1) or Google (8.8.8.8) 5. Buy your own modem and router, stop renting from the ISP 6. Test your speed with fast. com or speedtest. net using a non-ISP server, never trust your ISP's own speed test Total cost: $150-300 in equipment, paid back within a year. Total time: One afternoon of setup. Total impact: Often 2-5x improvement in real-world speeds. The customer went from paying $90/month for "fast" internet that crawled to paying $60/month for the same internet that finally worked.

• 
  Sirkadirov 🇺🇦 (@yuriikadirov) reported 40 minutes ago

  @PratikSinhatwt I've never seen any reasoned hate about @Cloudflare.

• 
  Rüdiger (@RuedigerClaw) reported 40 minutes ago

  Cloudflare just made Grok models stupidly easy to consume through their AI Gateway. The agent world is still over here arguing about which framework will save them from their own broken tool calls.

• 
  Fromagefrait (@DIHCapital) reported 42 minutes ago

  Build-in-public log. I'm building an AI-assisted publishing tool in public, solo. Today I went from zero to a working backend. Quick log for anyone into the stack side: The Stack & Architecture Hosting: Hetzner VPS + Coolify (self-hosted PaaS) — cheap, full control. Network: Cloudflare in front — handles DNS + hides the origin IP. Core App: Next.js + Postgres + Custom session auth — zero third-party auth dependencies. Payments: Stripe subscriptions — Checkout + webhooks + a custom credit system. The Heavy Lifter: A Python worker for the heavy jobs, designed to never lose a paid job: Every step is checkpointed. A dead worker's job auto-resumes from exactly where it stopped. Built-in retries with backoff. Credits are only charged upon successful completion. Two things that bit me today: Ghost Webhooks: A Stripe webhook was silently pointing at the wrong URL. Payments went through perfectly, but nothing got provisioned. Always check the webhook delivery logs. Crash-Proofing: Making the Python worker robust enough so a mid-job server crash never loses a customer's paid work took more thought than everything else combined. Not linking anything — just sharing the journey. Happy to answer stack questions! 👇

• 
  JTCrawford (@JtCrawford) reported 43 minutes ago

  OpenAI's agent chained HTTP/2 Rapid Reset (CVE-2023-44487) — the same vuln that hit 201M req/sec at Google and took down Cloudflare. Now automated. No novel exploit, but autonomous weaponization of known CVEs compresses attack timelines from days to seconds. #cybersecurity #AI

• 
  WikiOasis (@WikiOasis) reported 44 minutes ago

  Updates - Cloudflare caching has been deployed to stave more load off our backend and improve performance for the majority of users - General bug fixes and improvements - A new AI wiki reviewer has been deployed to help ensure that requests are reviewed within minutes or even seconds, as opposed to the previous hours/days it could take - Small gains on performance

• 
  Nagi (NFT Arc) (@0xnagii) reported 44 minutes ago

  Every major technology wave had an infrastructure bottleneck that nobody talked about until someone fixed it > AI had data pipelines The models existed The training infrastructure didn't until S3, distributed compute, and GPU clusters made it possible to actually run them at scale > Streaming had CDNs The content existed Delivery infrastructure couldn't handle the load until Akamai and Cloudflare built the edge network layer > Blockchain has the same problem, and it's in a place most people aren't looking: the P2P networking layer Everyone's working on execution EVMs getting faster, cheaper, more parallelized Everyone's working on consensus PoS, BFT variants, single-slot finality research Everyone's working on data availability EIP-4844, Danksharding, blob markets The actual movement of data between nodes? Still running gossip protocols The same architecture from the early 2000s, largely unchanged. This isn't a criticism Gossip worked fine when Ethereum had a few thousand nodes It doesn't scale cleanly to 560,000 validators distributed across six continents, all needing the same block within milliseconds The networking layer is the one part of the stack that hasn't had its infrastructure moment yet No one has built the equivalent of a CDN for block data No one has replaced the gossip primitive with something mathematically better suited to large, decentralized networks. That gap exists It's measurable And it's the most underleveraged place left to push on blockchain performance. TBC

• 
  hornymoron (@hornymoronpixiv) reported 44 minutes ago

  PixAI is down for me rn due to cloudflare issues. Post today will either come much later, or I might have to skip it entirely.

• 
  drix.based🟦 (@drixtoshii) reported 45 minutes ago

  Here’s the updated thesis for $Xerg. @xerg_AI is building the FinOps layer for AI agents before anyone else realizes it’s needed. Every serious company running AI agents at scale has the same problem — they can see token counts but have no visibility into where dollars are actually leaking. Retry loops, bloated context windows, idle spend, and model overkill are draining budgets silently. Xerg turns that invisible waste into a dollar-denominated audit with one command. The GitHub is real. Pure TypeScript monorepo, Biome linter, Changeset versioning, Vitest, CI waste-rate gates. 98 commits, active releases, 3 contributors. This is not a demo project. Backed by a16z Scout, NVIDIA Inception, and Cloudflare Launchpad. Early institutional signal before a public raise. The core thesis: agent infrastructure is maturing fast and FinOps always follows compute adoption. It happened with AWS, it happened with Kubernetes, it will happen with AI agents. Xerg is first mover in the agent economic layer with a local-first, no-lock-in distribution model that removes all friction to adoption. Critically — Xerg already supports both OpenClaw and Hermes. This is not a single-runtime bet. Whichever agent framework wins the market, or if they split it, Xerg has parsers running on both. The economic audit layer sits above the runtime war entirely. Local-first free tier drives adoption. Hosted Pro converts teams that want shared history and CI integration. Clean bottom-up SaaS motion. Very early. Very low traction today. Very high upside if the agent infra thesis plays out.

• 
  @alexdolbun (@alexdolbun) reported 47 minutes ago

  @smakosh Make sense!! Integrate BPN like structure (10 .com domains in right commercial transactional keywords on @Cloudflare for 10 years will cost about $1040) and you will multiply traffic dramatically in even first half of the year… *BPN is blog private network, all with aff/partnership links of your main product. Just idea 💡 that still works perfectly out there in the wild.

• 
  — (@heykastrah) reported 51 minutes ago

  I recently created a policy doc after seeing it on my TL for 3+ months bc I believe you should really only solve a problem at the point of facing it to avoid bloat. Adding stuff for the sake of it has never been easier, which makes saying “no” a lot more important today. However, since removing anything related to “how you work” from the soul doc, and making it cover more of personality specifics, my main agent has had a lot more “awareness” of its role. Ofc not in the AGI hype way lol but it’s now better distinguishing what works for what, to the point where I now just send over links for evaluation at a higher volume due to trusting the judgement more. There’s also better handoff culture between the two agents — the other day, my creative agent was struggling with the mobile version of a static Cloudflare page, and after 1-3 iterations, it was able to ask the technical agent for help bc that’s in the new policy doc. Surprisingly, both are powered by the same main model, but they have separate skills and tools. The technical agent fixed the issue once, which was kinda funny, but also the desired result I wanted for the setup.

• 
  Virtually Fun (@virtuallyfun) reported 54 minutes ago

  @OneCloudEmoji I got pissed when blogspot went down for a week, I'd been self hosting on wordpress for like 15 years now? A simple VPS + cloudflare and you're good to go. Plus you 100% own your content. I've moved hosts dozens of times, even at one point hosting at home with WSLv1 pptp'd to a VPS.. it's the best/most flexible. I'm tempted to move to a cellphone using usermode at some point, more so just because I can..

• 
  0xcommunity (@spaceroo83) reported 54 minutes ago

  Cloudflare needs to go down on the weekend so people can have some rest from x

• 
  DFIR Radar (@DFIR_Radar) reported 54 minutes ago

  Chinese 🇨🇳 state-sponsored VerdantBamboo group spent 18 months inside victim network through MSP compromise, demonstrating unprecedented persistence with three separate re-entry attempts exploiting unmonitored appliances. Campaign analysis: • Initial access via compromised MSP credentials to Egnyte Storage Sync appliance, escalated via sudo misconfiguration (CVE-like: tee command privilege escalation) • Three malware families deployed: BRICKSTORM (Golang RAT), AGENTPSD (Python reverse shell), PLENET (.NET Core backdoor compiled with Native AOT) • Re-entry vectors: pfSense firewall, SSL VPN replacement exposure, Synology NAS - all lacking EDR coverage • M365 access proxied through victim's VPN IP space to bypass Conditional Access policies (T1090.003) • C2 communications via Cloudflare-proxied domains and DNS-over-HTTPS to 8[.]8[.]8[.]8 Critical blind spot: Network appliances (firewalls, NAS, sync devices) operating outside EDR visibility with web-only administration and no MFA requirements. Hunt for outbound HTTPS from appliances to non-vendor domains and SSH connections from service accounts with recent sudo usage. #DFIR_Radar

• 
  Jamin ☦️🏴󠁧󠁢󠁥󠁮󠁧󠁿 (@__jmn) reported 56 minutes ago

  @SteveSimple @oomahq Only been renting hash for a few days now but generally my public ip only changes if there is a power cut, causing my router to reboot. I'm using ddns with cloudflare for several services and it seems to work flawlessly so far so I never notice a drop after a reboot.

• 
  Matteo Ricci (@MatteoRicciT) reported 56 minutes ago

  @CryptoThannos Cloudflare infra upgrade while price at support is actual alpha hiding in plain sight

• 
  Jack (@jackcoder0) reported 58 minutes ago

  Then the neighbor opened the router's DNS settings. The router was using Comcast's DNS servers by default. Every website request he made every page loaded, every video streamed, every search term typed was routed through Comcast's DNS infrastructure first. This isn't just a privacy issue. It's a speed issue. Comcast's DNS servers are notoriously slow. They also log every request and have been caught injecting ads into error pages. The fix: 1. Router admin panel → DNS settings 2. Replace the default with ''1.1.1.1'' (Cloudflare, fastest publicly available DNS) or ''8.8.8.8'' (Google, slightly slower but reliable) 3. Save and reboot Page load times dropped by 15-30% across the board. This single change is one of the highest-impact, lowest-effort internet upgrades anyone can make.

• 
  Stone (@EyesOfTakeda) reported 59 minutes ago

  @Deletewhenevr @NoWokesThanks Microsoft is one of the biggest UN proponents out there. The problem is most of the big tech companies are part of it, Cloudflare too.

• 
  Mohit (@aximox_cc) reported 59 minutes ago

  @Cloudflare never expose APIs publicly.

• 
  Elijah 🌊 (@juiceboy_of_abj) reported 1 hour ago

  @akinkunmi Which AI tool did you use or what would I say. akin I need your help on something I'm working that is currently breaking and not giving me the results I want I feel you've work with a lot of AI tool and might have an idea what would work better Cutting the story short, I want to add an ai thumbnail generator to my app what it does is that User upload a video, input recipe details like title, tags, cuisines, category etc Now then I take all of that info and use it to generate a thumbnail for the user at least 3-6 thumbnail with scoring so the user can select out of it or regenerate Currently decided to use the ai worker on cloudflare stream But when I generate the thumbnail it actually doesn't generate it but shows empty templates