In more than two decades working in infrastructure, big and small, I’ve never had important backups live in one place. even if that was just rsyncing the nightlies to a system in a different DC or whatever.
I know nothing about Parler’s infra team, including whether or not they had any control over software security or not. But like, them’s got some big data there, and I don’t care who you are, abstracted or not, that’s not an easy thing to just move to another provider with a different setup. Just moving that much data around is going to take awhile.
If I had 36 hours to move everything, this would (roughly) be my steps. in order of what I’d imagine was most important for a site like parler:
1 - figure out how I dump my entire object store and put it somewhere, because I’m betting that’s a big ol’ pile of data and it’s going to take many hours just to get out
2 - Immediately setup an off-AWS DB replica and start writing to it
3 - Start an immediate registry transfer process for domains (unlock them, get the transfer codes
4 - copy my artifact store somewhere (they mentioned “bare metal” though, so who knows if they were even containerized)
This assumes that 1) all the code and deployment is around something like git so they can ignore that, and that systems were created using terraform or at least with config management and not CloudFormation, so your setup is somewhat portable.
maybe with all of that you can deploy some servers somewhere, load your data and start reconfiguring your apps/containers/services to talk to one another again and get up pretty quickly. But given the number of SaaS components they were apparently using, their statement of being “platform-agnostic” sounds like bunk and they were probably pretty tied to some of those APIs, and untangling that has to happen before you can even pretend to come back online.
And remember, they’ve got a firehose of clients waiting on the other end, it’s not like they can trickle back, there’s however many API calls/sec being made to these endpoints from existing clients. They have to get the crushing load part right from the beginning.
As a purely intellectual exercise, this is a hell of a challenge. As an example of karma, it’s delicious.
