I’ve taken down production at Sky before, a revenue-generating component.
Specifically, the component for processing and fulfilling orders from skystore.com.
We were in the process of migrating this component from eu-west to eu-central and unbeknownst to me I thought we were ready to go ahead with deleting the cluster in eu-west… we definitely weren’t.
The repo was already configured to deploy to eu-central and all references to eu-west had been deleted. There were no references to eu-west anywhere, apart from the remaining k8s production pods.
Luckily we have a docker registry that an engineer on our team knew how to use to find a deployment of the production pods on eu-west so we could rectify my mistake.
Production was down for a good 20 hours or so in which we had amassed 25+ orders worth about £300 on the queue.
Definitely a learning experience. Made me realise it’s not a bad thing to slow down from time to time and understand the consequences of my actions.