WhatsApp Outage | Hacker News Reader

WhatsApp Outage

downdetector.in

90 points__debugger__61 comments

Curious that since it's the .in domain, the map just shows India. But if you replace it with .com, or .de, it will show the map of those countries (with .de it redirects to a German version of the domain).

Someone said the other week that downdetector just monitors how many people check their site for outages, and if in a five minute time period 20000 people are looking at them to see if WhatsApp is down... then, that's what they report.

I wonder if that behaviour could be engineered to fake an outage, though I'm unsure as to what end.

It wouldn't be much of a sabotage since the users would actually be able to access the 'service impacted'. Perhaps it could affect a stock price, but someone who sells a stock (i.e. Meta in this case) because WhatsApp had a 1-5-10-30mins downtime, is doing themselves a disservice and us a favor :)

>and us a favor :)

So there is a plan.

As I always say, downdetector has successfully predicted 20 of the last 3 outages. It really is the perfect website for the ad-supported revenue model.

I mean.. let's be fair, you open that site only when something isn't working, and are curious, if it's your phone/pc to blame, or if others have the same problem.

Proof that services written in Erlang CAN go down!

(Though it's probably DNS...)

Yes: They are software running on hardware, after all.

Is it still written in Erlang?

Erlang and Rust!

Count the number of unsafes in a Rust code base to see how secure the code is.

Serious question: is making a joke frowned upon on HN?

Generally, yes, although if the joke is nerdy enough it is sometimes liked. It's a fine line to walk, and you're probably better off not doing it.

But the thrill of having walked it and lived (in terms of karma)... Nothing like it!

Not against the guidelines(1) but generally speaking yes. Jokes without an accompanying insightful or useful comment are downvoted.

1: https://news.ycombinator.com/newsguidelines.html

I like to think of HN conversations as taking place in a third or fourth year university elective course. You are here because you want to learn something. Hence, all comments should be substantive and contribute to the learning of the class. In this scenario, jokes should be occasional, and should be embedded in a larger constructive comment, much as they would be in actual university class.

I was about to say something nearly identical to this, but couldn't have put it better than you did! In this specific scenario, I think the joke was in good taste and might even spawn some serious discussion and questioning (as seems to be the case already).

In fact, I came to this thread wondering if such a comment (about Erlang) would be made, and I was not disappointed. I am an Elixir (/Erlang/Gleam/BEAM/OTP) fanboy and WhatsApp is basically the canonical example of OTP's strengths, so perhaps I should be outraged! But I am not, and that says a lot about the quality of the joke (it's not a "Reddit-style" joke - it actually makes one think). It's fun and enlightening to consider the ways in which Erlang's reliability guarantees can be hamstrung by (presumably) external factors such as DNS/hardware/BGP, and a good reminder that things are messier than just which language or platform one chooses.

Anyway, yes. I would distinguish between enlightening jokes that arouse curiosity, and Reddit-style low-quality "humor" and trolling.

Yes, which only makes it more fun to make jokes that get upvoted despite the frowning.

FWIW, WhatsApp shouldn't go down because of DNS. At least, it didn't while I was working there. There are (were?) IPs baked into the app packages so that when you're on a network with bad DNS, the app still works. This also protects against authoritative DNS outages.

That said, the haiku readily transforms to BGP:

    It’s not BGP
    There’s no way it’s BGP
    It was BGP

If the sibling comments are accurate and Instagram and FB are also having trouble, then it's most likely shared infra that's problematic. Load balancers, BGP/routing in general, abuse detection, container management, some of the databases (although when I was there, the fb infra databases that WA used seemed pretty individualized and robust --- I don't remember any failures, certainly not coordinated failures)

Meta is experiencing downtime / routing issues in general. Reels on IG is broken atm.

Instagram down as well.

Everything else works fine for me, but the stories I shared 18 hours ago have vanished. On the webpage, Instagram's API returns a 502 error when I try to view them.

The horror!

Thank you for the helpful report of your experience. I liked how you dug into the API and got the error code.

Sorry guys, I rm -rf ‘d the real database instead of backups!

GitLab guy: that will be fine.

The hero we don't deserve.

Whenever these massive services get any substantial downtime I fantasize about it extending over a large amount of time. Just imagine getting rid of Instagram and TikTok for a while, considering the insurmountable addiction most people have to them. I bet the massive change in behaviour would be palpable very quickly, especially in kids who've known nothing else.

Does Meta provide a status page for its consumer products?

https://metastatus.com/

Not sure if it officially belongs to Meta though

It does, they link to it at https://developers.facebook.com/support/.

"Status and outages of Meta business products"

This is for their business products, not consumer "products" (where _we_ are the product). However, they tend to go down and up together, as they're tied to the same data and servers.

I whois-ed it.

metastatus.com Registered to Domain Admin Email domain@fb.com Last Updated Oct 03, 2023 Expires On Feb 22, 2025 Registrar RegistrarSEC, LLC Name Servers ns-1064.awsdns-05.org ns-1803.awsdns-33.co.uk

The email clearly points at FB.

It's not a consumer products! It's a transformative business platform that helps enterprises of any size engage audiences, accelerate sales and drive better customer support outcomes on the platform with more than 2 billion users around the world!

It’s basically OnlyFBFans — expose yourself, but don’t get paid

The issue I noticed is I still could make calls from and to local numbers (+33 to +33) but not from/to foreign numbers (+33 to +else)

That's odd – I wouldn't expect WhatsApp to use the phone number for any internal routing-related decisions (or as anything other than an account identifier in general). If it's a routing issue, that could explain the problems being localized, but I wouldn't expect the actual account phone number to have any bearing on it.

I've had no issues using it all day across several countries for both calls and texts.

Can't send any messages on Messenger on my phone, but facebook.com appears to be working fine.

All meta properties look affected.

Yeah, got a whole bunch of unread

I'll be that commenter this time, all green here! https://metastatus.com/

Not all green for me there!

https://metastatus.com/whatsapp-business-api

Major disruptions: Cloud API

Apr 3 2024 2:10 PM EDT

We are currently experiencing an outage impacting service on Cloud API. This issue started at 04-03-2024 11:10 AM PST. Our engineering teams are investigating the issue. We will provide another update within 4 hours or sooner if additional information is available

This will die out when Apple enabled RCS.

lolno

Why the hell are you going to abandon the app that you have and where all of your friends are? What incentive do you have? RCS isn't even feature-complete compared to WhatsApp and has for example no encryption.

It does have encryption in Google Messages like WhatsApp https://arstechnica.com/gadgets/2021/06/google-enables-end-t...

That's not RCS though, that's Google Messages. It's a platform-specific encryption feature not part of RCS in any way and will not be a part of the implementation when Apple adds RCS.

RCS has enough features for most people, and it will be pre-installed. After people will realize that mobile operating systems can do modern texting without any crutches from their respective app stores, they will increasingly not bother with them, and that will make Whatsapp to gradually be less and less important in an accelerating pace.

Those who actually care about e2ee don't use Whatsapp, because it is closed source and it nags users into enabling plaintext cloud backups.

Of course, if Apple's RCS implementation won't work without carrier-provided RCS, it will be dead in arrival outside North America and China.

> if Apple's RCS implementation won't work without carrier-provided RCS

Could you explain? If RCS works without carrier support, why would Google rely on carrier support in the first place? (It seems for phone number verification, they rely on SMS. I'm confused.)

RCS in its standard form is to be offered by the carriers. However, because carriers in the most parts of the world refused to offer the service, Google took over that role.

https://www.theverge.com/2019/6/17/18681573/google-rcs-chat-...

Apple has so far been quiet about how, if at all, they are going to deal with carriers that have no native RCS support. Maybe they will run their own servers and federate with Google, maybe they will use Google's servers directly, or maybe it simply won't work at all without carrier support. We have no idea yet.

Edit: IMO the most worrying option would be that Apple and Google will make a some sort of exclusive bilateral deal that will make them, and only them, to be able to offer non-carrier-based RCS support, locking out any other operating system vendors from the platform. That would probably have antitrust implications, but enforcement is often slow.

Is RCS possible without phone number? Like, Google would just set up an RCS server that accepts GMail addresses, and only falls back to your contact's phone number if they can't route the message? I just find it odd to have carriers involved at all (through the phone number), but sure, it's probably compelling to use an established concept.

Not sure if I understood it correctly, but this specification seems to discuss about SIP URI based non-mobile identities: https://www.gsma.com/futurenetworks/wp-content/uploads/2019/...

Thanks!

RCS failed. It failed the last >10 years and it will fail the next >10 years. There are too many companies and too much bureaucracy. It isn't an open standard anymore - Google extended it and now wants Apple to use their implementation. The rest of the world moved on from SMS and so should the USA.

RCS is a joke of an instant messaging solution, so I highly doubt it.

Why on earth would I rely on any messaging infrastructure provided by cell operators or alternatively Google?

Why would I tie all my messaging to a single device and the requirement of having a particular SIM card in it on top of that?

Absolutely nobody other than Google still believes in RCS, and I think doing the opposite of whatever Google does in trying to "solve instant messaging" is a good heuristic, given their track record.

Reply on news.ycombinator.com