This is devastating for Yandex. What does the SRE recovery plan even look like? “Raise shields”?
"We should be able to handle a long term net-split" was always part of my emergency plan for every service I worked on. (ie. Imagine all Comms between the Americas and the rest of the world cut for weeks)
Pretty likely to happen if there is a major war or political upheaval to require the internet be cut in two. I didn't require the service had no downtime, but that it was reasonably possible to restore service within a few hours.
Things contributing to that are you should have either live data or a backup in similar geo location to the user. Both the data and the crypto keys. Same for the code. And you should build your service simple enough that someone else who has never seen it before can deploy it and get things at least mostly working. Things like terraform are good for that - and it dramatically reduces employee training too.
Activate the failover datacenters in orbit :)
Put 90 percent energy into the shields!
Build data centers in politically stable countries.
Sure you cannot bribe anyone, actually have to pay taxes and can't install gas turbines but... they don't get blown up.
Same for AWS Bahrain, data centers are dual use technologies now and valid military targets.
I'd like to see their business continuity plan lmao
I used to say "What happens if Hurricane Sandy 2 hits?" e.g. a whole region goes down.
Now I can add "AWS Bahrain" and "Yandex" to that list.
Issue recommendation for management: exit the Ukrainian market to reduce physical cybersecurity risks.