Distributed Domain Destruction - Adventures in building distributed systems

Didn't learn much in this talk, but really enjoyed it, and had a good laugh :D

Good talk too listen to. Maybe the speed of the speaking combined with the absence of something to read made the presentation a bit hard to follow

I enjoyed the presentation and the attitude of the speaker. I really found it amusing. It brought back the memories from wc 2014, ansible talk and "what the f** is juju?!" response to audience question. Epic.

On a side note, consider using API throttling as standard method for avoiding "barcode-hitting-f5-continuously" DDoS attack. :P

While stories were maybe too specific in some cases, I love hearing them instead of 'best practices' people tend to read and then just spread around without having personal experience there!

BTW. Since there are so many issues with supervisor & upstart, any suggestions what to use or limiting number of workers is the only cure?

An exceptional talk, right from the trenches, where not everything is super and shiny and pink, but some real world scenarios from a large scale project and how to deal with things when they go wrong.

As always, Vranac puts together a great talk. Mandatory reading (viewing?) for the young'uns before they wander into distributed systems unprepared.

@Mario that solution is another 10 slides added to the talk :) the problem is a bit more complex than that.

@Miro as stated in the "Finite expanse of the unopening wallet" make sure principals are not too cheap, and getting more smaller instances instead of few big ones will help out sort this situation a lot.

Ofc ymmv depending if there are external factors limiting the number of instances.

The situation I described with Daemons is an extreme one, albeit real, as the limits of them are now widely known, and we had some limitations in play as well.