Conversation

1) A week ago, we had Cloudflare errors on FTX. I promised a post-mortem. Here it is.
9
280
3) What happened, roughly: (a) we had a particularly type of logging turned on (b) do to a bug in a maintenance process, Cloudflare accidentally routed logs to one of their shut-down servers (c) this triggered their DDOS systems
1
40
4) How we reacted: (a) we quickly found out that Cloudflare was blocking users (b) after about 10 minutes, we began switching systems over to a different DNS provider (c) at the same time, we began talking with Cloudflare's team to diagnose the issue
Replying to
5) After about an hour, we fully switched FTX's DNS provider, and service was no longer interrupted. After a few hours, Cloudflare identified and patched the issue; a few hours later, we switched back to Cloudflare.
2
38
6) I'd like to thank the team for being super responsive and helpful, and patching this same-day. Going forward, we have more practice switching DNS providers and should be able to do this more quickly if there are future issues.
2
87
7) Finally I'd like to apologize to Cloudflare and to our developers for the extent to which I'm sure I've botched this explanation.
8
115