zlacker

i wonder why they cannot partially rollout. like the other outage they have to do a global rollout.

>>websit+(OP)
I really don't see how it would've helped. In go or Rust you'd just get a panic, which is in no way different.

>>websit+(OP)
The article mentions that this Lua-based proxy is the old generation one, which is going to be replaced by the Rust based one (FL2) and that didn't fail on this scenario.

So, if anything, their efforts towards a typed language were justified. They just didn't manage to migrate everything in time before this incident - which is ironically a good thing since this incident was cause mostly by a rushed change in response to an actively exploited vulnerability.

>>denysv+C1
yes, but as the article states why are they doing global fast rollouts?

>>websit+r4
I think (would love to be corrected) that this is the nature of their service. They probably push multiple config changes per minute to mitigate DDOS attacks. For sure the proxies have a local list of IPs that, for a period of time, are blacklisted.

For DDOS protection you can't really rely on multiple-hours rollouts.