Just wanted to double check if there is a situation where the actual number of serving replicas can exceed the number defined in .spec.replicas for a deployment.
Based on your strategy, while the rollout is in progress, you can have more pods than number of replicas.
.spec.strategy.rollingUpdate.maxSurge controls how many extra pods may be spun up during the deployment.
http://kubernetes.io/docs/user-guide/deployments/#strategy
Once the rollout is complete, the desired and current number of replicas will match.