# Infrastructure as cattle, not pets This theory in [[Site Reliability Engineering]] is born out of the [[Distributed computing]] revolution: it states that [[Containerization|containerized]] components and services should be thought of as transient, temporary, and ultimately replacable. Where previous practices may have involved, for example, logging into a server and manually configuring it for optimal [[Performance]], this theory looks at this one-off configuration as a warning sign. Instead, engineers should include such configurations in the container itself, so that any changes made will be adopted the next time the container is rebuilt. ![[GitHub’s Engineering Team Has Moved to Codespaces the GitHub Blog#^62c0ee]] In [[Kubernetes]], one of the defining characteristics of pods is that they are [[Kubernetes#Pods are ephemeral|ephemeral]]: easily replaced.