การค้นหาบริการและ DNS ใน K8s

Welcome to Service Discovery & DNS in K8s! In a dynamic Kubernetes cluster, Pods are constantly created, destroyed, and moved, leading to ever-changing IP addresses.

How do applications running in one Pod find and communicate with applications in another Pod or Service?

This is where Service Discovery comes in. It's how services find each other without needing to know their specific, ephemeral IP addresses.

You might already know about DNS (Domain Name System) on the internet. It translates human-readable names like google.com into IP addresses.

Kubernetes uses a similar concept internally. Instead of relying on unstable Pod IPs, K8s assigns stable DNS names to its Services.

This allows your applications to connect to other services using simple, consistent names.

Every Kubernetes cluster comes with its own DNS server, typically CoreDNS (or sometimes kube-dns in older versions).

CoreDNS runs as a Pod within your cluster and is responsible for resolving all internal Kubernetes service names to their corresponding cluster IP addresses.

It's automatically configured for you when you set up your cluster.

When a Pod starts, Kubernetes automatically injects DNS configuration into it.

• Each Pod's /etc/resolv.conf file is updated to point to the cluster's CoreDNS Service IP.
• It also includes search paths, allowing you to use shorter service names.

This means any application inside a Pod can immediately use the cluster's DNS for name resolution.

When you create a Kubernetes Service (e.g., a ClusterIP Service), CoreDNS automatically creates a DNS record for it.

The most common format for a Service's DNS name within its own namespace is simply: <service-name>.

For example, a service named my-web-app in the default namespace can be reached by other Pods in the same namespace using just my-web-app.

Let's say you have a deployment hello-app and a service hello-service in the default namespace. Any other Pod in default can reach it via hello-service.

Here's a simple example of a Deployment and Service:

After applying the YAML from the previous scene, you can test service discovery. First, create a temporary Pod (e.g., debug-pod) in the same namespace.

Then, exec into debug-pod and try to ping the service:

• kubectl run debug-pod --image=busybox --restart=Never --rm -it --command -- sh
• Inside the pod: ping hello-service

You should see the hello-service ClusterIP being resolved!

What if your services are in different namespaces? Kubernetes uses a Fully Qualified Domain Name (FQDN) format:

<service-name>.<namespace-name>.svc.cluster.local

For convenience, if the Pod's namespace is in its DNS search path (which it usually is), you can often use a shorter form:

<service-name>.<namespace-name>

This allows clear and unambiguous communication across different parts of your application.

Let's imagine a backend-service in the prod namespace. From a Pod in the default namespace, you'd access it like this:

Kubernetes DNS handles the resolution, directing traffic to the correct service, even across namespaces.

This provides strong isolation while still enabling communication where needed.

Great job! In this lesson, you learned how Kubernetes handles service discovery and DNS resolution for internal communication:

• Pods have ephemeral IPs, requiring a stable discovery mechanism.
• CoreDNS is the cluster's internal DNS server.
• Services get stable DNS names, resolvable by other Pods.
• Pods are automatically configured to use the cluster's DNS.
• You can access services in the same namespace by name or in other namespaces using FQDNs.

This powerful system ensures your applications can always find each other reliably!