Great talk. Some very helpful tips for users running K8s on AWS. However it does seem like the whole termination policy of spot requests makes it complex to manage - if the cost savings are high then I could see a use case but I also see a lot of additional code for managing something that could be automatic. In a nutshell, spot requests seem to bring a new layer of complexity and i always find others who are hesitant to use them. But it's great to see how it's worked out for your team. After watching this I will most likely consider spot request for managing K8s workloads.