Yet another Hacker News

Sudden drop in Cloudflare bot traffic after strike on Yandex datacenter

260 points walrus01 2 hours ago 94 comments

toomuchtodo 2 hours ago | parent

Cloudflare should consider donating to the Ukrainian military for the public service they are performing.

exe34 1 hour ago | parent

I was thinking, we should all contribute to a Yandex fund for the Ukrainian mil to help keep the level of bot activity on the rest of the internet under control!

toomuchtodo 58 minutes ago | parent

https://war.ukraine.ua/donate/

(I’d welcome a data center and oil infrastructure targeting specific fund)

rvz 2 hours ago | parent

“AI agents” is just a fancy rebranding of “bots” with some “intelligence”.

WiSaGaN 2 hours ago | parent

isn't this only counting russian origin already?

walrus01 2 hours ago | parent

The URL is CF's public metrics for bot traffic from a single specific Russian ASN, Yandex.

https://www.peeringdb.com/net/1751

cowlevel 1 hour ago | parent

oh well of course if you blow up Yandex then Yandex stops sending you traffic, because they got blown up.

walrus01 40 minutes ago | parent

And nothing of value was lost.

cowlevel 34 minutes ago | parent

Yandex is the best search engine that doesn't follow American censorship norms. For this reason Kagi uses them as one source of several.

r721 2 hours ago | parent

*Sudden drop in Cloudflare bot traffic from AS13238 YANDEX — Yandex LLC

UPD Definition of bot traffic from Cloudflare's glossary:

>Bot vs. human traffic

>The percentage of HTTP requests classified as bot versus human, based on bot scores. Requests with a bot score between 1 and 29 are classified as likely automated (bot), while requests with a score of 30 or above are classified as likely human. For more information, refer to Bot classes.

>By default, Radar shows bot vs. human traffic for requests to HTML content. This filter is meant to represent traditional web traffic by excluding API calls, asset requests (images, scripts, fonts), and other machine-to-machine requests.

https://developers.cloudflare.com/radar/glossary/#bot-vs-hum...

https://developers.cloudflare.com/bots/concepts/bot-score/

walrus01 2 hours ago | parent

Correct, I probably didn't didn't write that in the clearest possible way, I meant "detected bot traffic hitting cloudflare originating from Yandex".

londons_explore 1 hour ago | parent

Yandex is Russian Google.

Of course they have bots... That's how you index the web.

If we took Google offline I bet there would also be a reduction in bot traffic from them too!

Indexing is also probably a low priority task, so if half their datacenter capacity is taken offline, their bot traffic probably reduces by more than 50%.

cromka 1 hour ago | parent

I don't think these are the boys they refer to?

EDIT: I was wrong, they're just generic bot vs human traffic numbers and you're right.

AlienRobot 1 hour ago | parent

I wonder how this will affect the Yandex search engine. I think I've read they split the Russian one and .com, but I'm not sure.

dgellow 57 minutes ago | parent

Yandex has cloud offering, their servers are used for actual bot traffic, not just crawlers. Yandex hasn’t been only a search company for a very long time.

nik282000 10 minutes ago | parent

Google hosts a significant number of bots. About half of my fail2ban list is google cloud IPs.

asdfsa32 52 minutes ago | parent

Don't let logic get in the way of a good propaganda.

lukan 36 minutes ago | parent

Logic says google bots do not come disguised, but as google agents.

Or is there other information?

Now there is probably also AI agent spam from google, but not at this level I believe. Also I am not fully clear what they count as spam, I doubt they would include google in this list, so do they include yandex bots here, if they come officially?

konart 32 minutes ago | parent

Neither do Yandex bots. What's your point?

lukan 24 minutes ago | parent

If official yandex bots ain't included in this data here - then it means spam/scam bots come indeed from yandex servers. As they stopped working, when the yandex data center was hit.

asdfsa32 20 minutes ago | parent

But what do you mean by "if"? It is a well defined metric and there is no "if".

konart 18 minutes ago | parent

I'll just link this comment if I may: https://news.ycombinator.com/item?id=50042416

asdfsa32 22 minutes ago | parent

I am unsure what you're trying to say. But it seems like you don't understand what Bot means in general and how is it specifically defined for the metric provided in that page.

From that pages point of view, a Google bot, Yandex bot, or your uncle Tom's AI agent spam is the same, a bot.

lukan 3 minutes ago | parent

I know what it means in general, but like I said indeed not "how is it specifically defined for the metric provided in that page."

caaqil 35 minutes ago | parent

It's a nice reminder that HN is just filled with the same primitive tribal normies and is always getting oneshotted by the same transparent xenophobic hysteria that plagues the rest of the internet.

Citizen_Lame 21 minutes ago | parent

Will someone think of the Russkies.

loeg 19 minutes ago | parent

> HN is just filled with the same primitive tribal normies

Yes, but it's mostly the opposite tribe of the one you're complaining about.

GlacierFox 13 minutes ago | parent

Yeah for real, enlightened one. Russia is great by every metric /s.

wewxjfq 21 minutes ago | parent

From submission title down this comment chain, I'd say the logic stops and propaganda starts with your comment.

asdfsa32 18 minutes ago | parent

Thanks for making it clear that you didn't look past the submission title.

walrus01 29 minutes ago | parent

The other Yandex AS which is their hosting/hypervisor platform, broadly similar to AWS, Azure, whatever also shows a major drop in bot traffic:

https://radar.cloudflare.com/bots/as200350?dateRange=7d

https://www.peeringdb.com/net/20950

schleck8 1 hour ago | parent

The HTTP traffic from Yandex is down to ~ 0 % after the third strike, after the first two strikes already degraded it. I wonder how this looks like at the network level because their search page still serves, but presumably their crawler is inactive now

Edit: Here is some more info, they do have availability issues as is to be expected

> Because three of its primary facilities were knocked offline within four days, Yandex’s failover mechanisms broke down[1][8].

> Outages have cascaded across Yandex Go (taxis), Yandex Pay, food delivery, smart home platforms, internal logistics, and third-party corporate services hosted on Yandex Cloud[2][7]. Problems have spilled into neighboring countries using these systems (Belarus, Kazakhstan, Armenia, Serbia)[2][7].

Edit 2: Here is the Yandex Cloud status page, their GCP/AWS equivalent. The entire compute core is offline, and even worse, Cloud Backup

https://status.yandex.cloud/ru/dashboard

> ru Compute Cloud, ru Kubernetes, ru PostgreSQL, ru ClickHouse, ru MySQL, ru YDB, ru Kafka, ru Application Load Balancer

Interestingly, Yandex tells its users on the status page to switch to competitors Selectel, K2Cloud, and VK Cloud. Which will probably be hit next

Bluestein 1 hour ago | parent

This will be a case study in catastrophic events for HA and Ops folks.-

AlienRobot 52 minutes ago | parent

I don't think there is much you can do about being targeted by a military, simply because of how unique the attack vector is.

Hacking is smart but happens through software, so if you have hardware (i.e. physical) barriers, that stops the hacking.

Natural disasters happen physically but they are dumb, so if you have software redundancy (i.e. by copying the data to multiple locations), that stops them from destroying everything.

A military is both physical AND smart. They can target multiple locations at once and destroy whole buildings.

simondotau 36 minutes ago | parent

You can learn to quickly diversify availability (or at least data safety) beyond the region of conflict when a kinetic war breaks out. The moment “kinetic sanctions” became the strategy, the risk to data centres should have been self-evident. And if it wasn’t immediately obvious, it should have been blindingly so when targeting expanded from oil refineries to Wildberries warehouses.

jacquesm 31 minutes ago | parent

> I don't think there is much you can do about being targeted by a military, simply because of how unique the attack vector is.

You could choose not to invade other countries.

AlienRobot 25 minutes ago | parent

I guess, but that doesn't sound like something in the control of a single individual company.

loeg 17 minutes ago | parent

This (as with defense) is kind of outside Google or Yandex's purview.

whatdouthink 15 minutes ago | parent

But apparently this does not apply to you.

Hamuko 37 minutes ago | parent

I feel like this isn't even a particularly new thing. I worked at a medium-sized SaaS company and we did disaster recovery drills for moving our stack from eu-west-1 to eu-central-1 in case something catastrophic happened in Ireland. If you want HA, you kinda have to assume that any single country in the world can have some kind of a crippling disaster. Unfortunately for Yandex's clients, Yandex only has regions in two countries, and not that many regions either.

walrus01 16 minutes ago | parent

I don't know how useful it will be as a case study since an HA/Ops person in (let's say, the US, Canada, AU, western european countries) has access to do things like buy hypervisor capacity or colocation in just about any location. I could go buy colo in Brussels or Panama City or San Diego or Edmonton or Melbourne tomorrow by contacting the right people.

The Russians are much more restricted from doing anything geographically outside of their own borders (for very well justified sanctions reasons), so their limitations are not the same as an HA person in almost any other country. Maybe some Iranian HA/Ops people will be comparing notes with them.

ivan_gammel 4 minutes ago | parent

Sometimes there exist data or processing residence requirements that bind you to a specific region (e.g. EU-only). Also, „all AZs in this region no longer exist“ is not on everyone‘s bingo card, this is a good reminder why multi-AZ deployment in the same region may be not enough.

For data center guys this is definitely a case study, because drone attacks are probably on the threat list for everyone now, for various reasons, including domestic terrorism and hybrid warfare. If you want to defend your investment, time to think what covers you from the air.

bamboozled 59 minutes ago | parent

I just searched "blyat" and it worked, so?

eptcyka 53 minutes ago | parent

Given that blyad is just punctuation at this point, I am not surprised.

throw-the-towel 42 minutes ago | parent

The username checks out!

("Ept" and "suka" are two other swearwords, that are also used much like punctuation.)

walrus01 39 minutes ago | parent

I predict a great many more videos coming out of Russia of Firepoint UAVs flying into refineries and such, punctuated by background audio commentary that can be summed up as "cyka blyat"

pluc 1 hour ago | parent

That's some intense mitigation from Cloudflare

Alifatisk 59 minutes ago | parent

You gave me a good laugh, cheers!

bamboozled 1 hour ago | parent

I'm going to be really interested how much this cleans up the internet, I mean the damage is done if you read comment sections on youtube etc, but let's see how long the BS persists.

cowlevel 44 minutes ago | parent

It's only counting Yandex search crawler traffic, not any actually bad bots.

speckx 58 minutes ago | parent

I have to wonder what will happen when the remaining 6 data centers get hit as well. It's bound to happen.

afdbcreid 48 minutes ago | parent

If there will be an world-wide war, sure. But at that point, the Internet as we know it will probably collapse anyway.

cowlevel 45 minutes ago | parent

There is one, but it's taking place mostly by information means, not physical.

walrus01 37 minutes ago | parent

https://postimg.cc/YjjDtrT2

Hamuko 36 minutes ago | parent

Are there even six remaining?

dintech 55 minutes ago | parent

Oh no, how is Trump going to win the midterms now?

surfingdino 54 minutes ago | parent

So, is sending hot hate towards a bot farm going to be a paid service on Cloudflare now?

876433790 42 minutes ago | parent

Would be ironic, since they provide services to criminals who commit DDoS-attacks, spread malware and conduct phishing campaigns:

https://cofense.com/blog/how-cloudflare-services-are-abused-...

https://www.heise.de/en/news/Report-Cybercriminals-use-Cloud...

Or maybe they would just want to get rid of the competition.

ericpauley 52 minutes ago | parent

Interestingly most traffic coming out of this AS was actually UDP: https://dashboard.terracenetworks.com/sources/asns/13238

Pretty benign AS as far as actual exploit or malicious crawling is concerned.

cowlevel 45 minutes ago | parent

Probably QUIC if true, but you should also ask how these people get their data, because this isn't something you can just see or look up.

bearjaws 51 minutes ago | parent

Someone needs to inventory all the accounts that go dark on social media so we can flag them going forward.

timr 49 minutes ago | parent

Isn't this showing the percentage of bot traffic from the Yandex IP range?

If so, it isn't particularly surprising that the line would go down.

EDIT: yes, it is. Looking at the same graph for all of Russia, and there's no discernible pattern:

https://radar.cloudflare.com/bots/ru?dateRange=7d

same for the 48h view:

https://radar.cloudflare.com/bots/ru?dateRange=2d

pKropotkin 43 minutes ago | parent

Could be just crawlers

throwaw12 40 minutes ago | parent

Someone needs to learn how to read the Bot Traffic page.

That is a drop in traffic specifically from Yandex bots (which includes their crawlers and other apps hosted in their DC)

Now compare it with Bot Traffic in:

* Russia: https://radar.cloudflare.com/bots/ru?dateRange=7d

* Europe: https://radar.cloudflare.com/bots/europe?dateRange=7d

* Worldwide: https://radar.cloudflare.com/bots?dateRange=7d

Dip is barely noticeable even in Russia.

Its like saying, "API requests are completely down" (and small disclaimer: only from your server, because your server is down)

LiamPowell 37 minutes ago | parent

Everyone is mentioning crawlers, but wouldn't most traffic from a datacenter's ASN usually be bots just in general given that there's not many humans there to generate human traffic?

f311a 37 minutes ago | parent

I can’t believe there are so many dumb comments in this thread. All it shows is that crawling from Yandex search engine stopped from one Yandex data center ASN.

No one runs opinion/malicious bots from data center IPs that people mention here.

Did people actually stop thinking?

walrus01 35 minutes ago | parent

On the other hand there are plenty of datacenter located command/control systems that puppeteer vast fleets of residential proxies (and Russia is by no means the only one doing so), so it will be very telling to see what social media accounts go silent over the coming days and weeks.

> from one Yandex data center ASN

This is only one of two extant Yandex ASN, not an ASN specific to one geographic region/datacenter.

https://www.peeringdb.com/net/1751

The other one is this: https://www.peeringdb.com/net/20950

And the 200350 also shows a near complete drop off in bot activity:

https://radar.cloudflare.com/bots/as200350?dateRange=7d

monday_ 7 minutes ago | parent

You are replying to a war propaganda post with a heavily editorialized title. Of course it's swimming in bad faith engagement, that's how these things work.

blingo2665 31 minutes ago | parent

An eye for an eye for the Russians. I love western imperial propaganda though, and how HN so eagerly amplifies it. There’s something self-fulfilling about willingly becoming a foot soldier of the Empire.

Here's to a world where the West is monolithic, and may Iran, Iraq, Yugoslavia, Syria, Libya, Lebanon, Palestine be only be the beginning. And may every new crusade for freedom and democracy leave behind another Gaza, in pursuit of upholding human rights and civilizing the world.

beaker52 27 minutes ago | parent

It’s not really surprising that bot traffic originating from a network went away when _all_ traffic originating from that network stopped.

bhouston 27 minutes ago | parent

The US is still responsible for almost 50% of traffic of which the top are Amazon, Google, Microsoft, Cloudlare.

https://radar.cloudflare.com/bots

accountrequired 24 minutes ago | parent

When Brazil and Afrinet? :P

andrepd 23 minutes ago | parent

Now do OpenAI's dc...

salvuccio 14 minutes ago | parent

Low-IQ Americans will think these bots are X accounts posting memes against Democrats, which is why the media is astroturfing this news to target American liberals. In reality, most of these bots are just scripts performing normal enterprise jobs, but most people know nothing about how the world works.