Night of a cascading failure

It’s a bit like DNS: you ask it for someone who’s providing X, and it gives you a list of candidates: hosts and port numbers, that sort of thing. The actual PBB node is the parent of the whole special set, and all of the others have names which continue on from it. Now, confident they had a good number thanks to their check, they proceeded to use it in a string comparison operation. There are a lot of things which could have gone better with this, and hopefully other people can look for similarities in their own software stacks and avoid the same kind of outage. Well, having the whole world die because that lookup service went down is certainly a bad thing.

Source: rachelbythebay.com