3 ms·

WhatsApp going down is a total shock. I thought that ran on completely separate infrastructure.

• _Understated_ 5y ago
  If the issues is DNS-related or routing-related then since it's all owned by FB, they own the records too so it's likely that literally everything that has a DNS record under the FB cost code will be hit. The WhatsApp servers will be sitting idle most likely!
  • latortuga 5y ago
    Leaving aside whether or not this is true, why would you apply DNS changes to all of your properties at once? Especially if it's a sensitive change? Seems much more likely to me that WA/IG/Oculus somehow rely on facebook.com behind the scenes than that all FB domains were affected by a config change.
    • notwedtm 5y ago
      A misconfiguration of BGP would absolutely do it.
• toast0 5y ago
  I worked at WhatsApp until Aug 2019. WhatsApp hasn't been on completely separate infrastructure for quite some time. It's in FB datacenters, so if FB BGP is messed up, so is WA. There is a separate WA ASN that's used for DNS, but it's still FB infra, and announced through the FB ASN, so that doesn't help either. Instagram still has their DNS with AWS Route 53, so their DNS is still up, but their site isn't because it runs on FB Infra too. Within the FB datacenters, WhatsApp is somewhat isolated (at least when I left, chat was run on dedicated hosts allocated to WhatsApp only, but using FB container orchestration tools, and FB specified hardware, etc). Edit to add: WhatsApp has some mitigations against DNS not working, but in this case, it looks like DNS being dead was a symptom of something (probably BGP config error, if the Reddit post earlier was accurate) and that something also broke the underlying servers.