在kube.netes中，Cluster IP的用法可以替代Pod IP的用法吗？

Question

In K8S, there are cluster IP and pod IP and I understand the difference between two.

I have two questions.

1. Usage of Cluster IP can replace usage of Pod IP? The cluster IP is used for the.networking between the servers in the same cluster. Pod IP is also for the internal.network traffic. In our company's legacy service, the pod IP is used so it needs a server which manage the pod IP of every server. We are making the new service now. When connecting the service mesh in Micro Service Architecture (MSA) server which is scalable (by duplication of pods), can I use the cluster IP? If so, why the service mesh tools like ISTIO and Spring Cloud Eureka is required in MSA? The cluster IP can be used instead of these tools.

2. Why the pod IP is designed to be changed when the pod is recreated?

Answer 1

When you install Kube.netes, it installs a private.network which are used by the pods.

This is a class like 10.0.0.0/24. So your node in this case have 256 ipaddresses available to run your pods.

Those Ipaddresses are not accessible from outside your node directly.

Each time a Pod starts one of the available Ipaddresses is assigned to the pod. Thats also the so called clusterIP.

Normally you have more than one node. So on each node there is configured such a.network for the Pods.

To communicate between the pods clusterIps on different nodes a Network Fabric must be installed in the cluster. (ex. Flannel or Calico).

The Pods clusterIp you cant predict. It can change every time a Pod is recreated. To connect Pods between them create a service with type: ClusterIP.

Kube.netes has an internal DNS. You can access a pod like so:

servicename.namespace.svc.cluster.local:port