Cloudflare just jumped off a bridge, down globally.
Downdetector down as they use Cloudflare, journalists are gonna be fukken mad as how they do they say 1000 people have reported an outage
Here's what it looks like from Mastodon end for anybody interested, green is successful queues, red is failed. Unfortunately a lot of the fediverse is behind antitrans supporters, Cloudflare.
Cloudflares status page is down now too

The Internet: designed to withstand an nuclear blast, reroute automatically.

Four tech companies: what if we can bring it down at any moment by typo'ing a regex string, lololol

lol, Cloudflare's status page got overloaded
Tech press trying to cover Cloudflare outage
Cloudflare say they are seeing signs of service recovery, but the failed queues are getting worse on my Mastodon server.
frame this modern art

Cloudflare continues to investigate an hour in

Mastodon failed queue continues to grow

Cloudflare 'Change Management' in action
Hour two of the Cloudflare outage, Mastodon retry queue at record length and failed attempts at highest amount on record too.
Wild, entering hour three and still seeing sites down and about 40% failure rate on Mastodon Sidekiq queue requests.
Fun fact, if you google booter services (DDoS services), almost every major one is a Cloudflare customer. So currently, by anti-DDoS provider Cloudflare being offline, the world is a safer place.
Cloudflare says it is resolved, and my Mastodon server queue is clearing now... so prepare for DDoS to resume, since their DDoS botnet customers are back online.

Here's the 6 month Mastodon Sidekiq track, you'll see the highest amount of failures on the far right today in red - that was the Cloudflare outage.

Hopefully it encourages more #mastoadmin to consider if Cloudflare is a good choice.

17 hours from now, somebody will reply to this thread saying 'It's not down, works for me'. Always happens. I love the internet.

Cloudflare RCA is out.

Basically a few things went wrong which then hit a coding error. https://blog.cloudflare.com/18-november-2025-outage/

@GossiTheDog but it was memory-safe! 😎

@bagder @GossiTheDog

The 500 in FL2 in Rust was memory-safe, but what about the error in FL that set the bot score to 0 ?

@GossiTheDog it's hard to know what to do with those.
I used to work with real time software for ship radars, so safety critical in a sense.
We used Design By Contract extensively, and when the code was under development we enforced a software halt, and logged a message.
There was great debate about this for production systems, as halting when someone is shooting at you is bad. We eventually settled for logging the error, but continuing in the hope it would be alright.
@GossiTheDog CloudFlare could have chosen to drop extra features and continue, but it's hard to know what the ramifications might be.
@tjbutt58 iirc linux "security best practices" say if the audit log disk is full, halt the server. Sysadmins were not amused to find out this config behavior in production
@GossiTheDog
Today is a good day to go look in your #rustlang code for all the unwrap() and expect() that you left for later... 😉
#cloudflare
@GossiTheDog The bottom line seems to be 'we weren't sure something was wrong internally until it was completely wrong'.
@GossiTheDog So they've set a limit for something but forgot to add input validation to prevent this very critical piece of the infra completely shit itself in case of a misconfiguration. Nice.

@GossiTheDog

The biggest question I got is why they didn't find out during the standard OTAP procedure they "allegedly" are following like a professional company would.

@GossiTheDog Rust? They should rewrite it in C.

@GossiTheDog

'It's not down, works for me'. My washing line is still working. Erm what day is it?

--
Mrs Trellis, North Wales

@X31Andy @GossiTheDog
Good morning Mrs. Trellis.
We've been trying to reach you about your extended warranty . . . 😁

@AG100pct @GossiTheDog

Dear John Mastodon,

My latest crop of poppies were the normal length this year so I dont have to worry about them being extended.

I put some into a salad for my husband who later told me the world was wonderful and he enjoyed watching the yellow and pink squirrels in the garden. I need to get his eyes checked because could only see the neighbours white tomcat.

Yours. Happily

Mrs Trellis, North Wales

@GossiTheDog
1.1.1.1 doesn't seem to be responding, while other Cloudflare DNS are.
@GossiTheDog I'll have you know that my Fedi server hosted behind cloudflare is working fine. I would've replied to you earlier but uh, my whole instance was down for some unknown reason.
@GossiTheDog it's not down. Works for me.
(Sorry. I couldn't resist)
@GossiTheDog With what software are you monitoring your instance?
@GossiTheDog Cloudflare has been back up for less than 30 minutes and one of the IRC networks I'm on just had a netsplit. But maybe they didn't get DDoSed, it could be a coincidence. 
@GossiTheDog
And don't forget their spam bots!
@GossiTheDog @tiotasram
“First we’ll poison you, then we’ll sell you the antidote”
@GossiTheDog that sounds like a protection racket
@GossiTheDog Impacted twitter upload as well.
@GossiTheDog not even their support work, or log in, one is really lock down :/
@GossiTheDog Gotta love the age of single points of failure in a supposedly massively distributed network of independent nodes
@elias_sorensen @GossiTheDog Paul Baran spinning in his grave.
@GossiTheDog Decentralized Mastodon Instances behind centralized CDNs - find the misconception
@GossiTheDog incoming euro-pop song about Cloudflare?
@GossiTheDog have they tried turning it off and back on again?
@GossiTheDog ah, that explains things 😂 thanks, I’ll go back into the forest for now
@GossiTheDog I feel like updates from tech companies shouldn't read like something I would expect from a Mel Brooks script
Our Plans | Pricing

Discover which Cloudflare plan is correct for your requirements. Find out more about Cloudflare plan pricing and sign up for Cloudflare here!

@GossiTheDog is it BGP, DNS or breach - place your bets here!
@GossiTheDog Cloudflare's Canadian site makes no mention of this.
I guess they are pretty busy right now.
@GossiTheDog i mean twitter loads some times again
@GossiTheDog meanwhile, literally every mid-sized to big company: we'll just pretend that didn't happen and will not change anything :)
@sijmen @GossiTheDog The same after CrowdStrike 2024, AWS 2025, etc.