there are too many words and articles about Kubernetes CSI but none of them explains why Kubernetes needs to CSI. It’s already possible and easy to mount a stroage volume to the host/node, why does Kubernetes need to long and scenic routes to connect storage? Docker (or whatever) can already use a storage volume like a local volume. Could you please help me clarify?
1 Answers
to abstract the storage type and implementation details from the k8s resources that require storage into few abstractions pv, pvc, and storage class. in other words, think of k8s as your multi-cloud, multi-platform workload engine. storage can be public cloud storage, can be on-premise NetApp storage or anything else that does not exist right now, instead of overloading and writing storage specific code into kubernetes, you just need to the storage to know how to interfaces to kubernetes, that is one advantage, the other is that for end users ops or dev, the storage specifics are abstracted and they deal with it through k8s abstractions of pv, pvc, and storage classes