2 ms·

From the rest of the post, it sounds like replication time. Datacenters started dropping an hour beforehand one by one, and they had all fallen over by 19:08. G

From the rest of the post, it sounds like replication time. Datacenters started dropping an hour beforehand one by one, and they had all fallen over by 19:08. Given that you have to push the rollback to routers around the world, and that peer routers have to propagate the changes from there, 18 minutes for a change like this sounds about right.

jlgaddis 11y ago
... although once the first datacenter once again announced the prefixes into BGP, those networks would have been reachable again, from everywhere. I imagine this is what happened at 19:27 -- the first datacenter came back online. Of course, the traffic load might have overwhelmed that single datacenter but that would be alleviated as soon as additional datacenters came back online ("announced the prefixes"). A portion of the traffic load would shift to each new datacenter as it came back online. It could have been hours later before they were all operational again but, as far as the users were concerned, the service was up and running and back to normal as soon as the first one or two datacenters came back up.