Skip to main content

Documentation Index

Fetch the complete documentation index at: https://notes.kodekloud.com/llms.txt

Use this file to discover all available pages before exploring further.

In this walkthrough, we’ll explore how to secure Pod-to-Pod and Pod-to-External traffic in Kubernetes using NetworkPolicies. You will learn to:
  1. Verify the default connectivity behavior
  2. Apply default-deny rules for egress and ingress
  3. Permit specific egress/ingress to selected Pods
  4. Validate the resulting network restrictions

​
1. Verify Default Connectivity

By default, Kubernetes allows all egress and ingress traffic between Pods (even across namespaces) and to the Internet.

​
1.1 Test External Connectivity

Exec into pod1 (in the default namespace) and ping an external endpoint: 
kubectl exec -it pod1 --container container1 -- ping -c 4 www.google.com
You should see successful responses: 
64 bytes from 142.250.125.103: icmp_seq=1 ttl=111 time=2.05 ms
...
4 packets transmitted, 4 received, 0% packet loss

​
1.2 Test Cross-Namespace Connectivity

List Pod IPs in kube-system and pick one (e.g. 192.168.121.187): 
kubectl get pods -n kube-system -o jsonpath='{range .items[*]}{.status.podIP}{"\n"}{end}'
From pod1, ping that IP: 
kubectl exec -it pod1 --container container1 -- ping -c 3 192.168.121.187
You should receive replies, confirming open egress/ingress.
By default, no NetworkPolicy is enforced, so all traffic flows freely.

​
2. Apply Default-Deny Egress

To block all outbound traffic from Pods in the default namespace, create a default-deny egress policy. 
# deny-egress.yaml
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: default-deny-egress
spec:
  podSelector: {}          # selects all Pods in default
  policyTypes:
    - Egress
Apply and verify: 
kubectl apply -f deny-egress.yaml
kubectl describe networkpolicy default-deny-egress
You should see policyTypes: [Egress] and no egress rules.

​
2.1 Validate Egress Blocking

Attempt to ping Google and a cross-namespace Pod—both should time out: 
kubectl exec -it pod1 -- ping -c 2 www.google.com
kubectl exec -it pod1 -- ping -c 2 192.168.121.187
No responses will be received.

​
3. Apply Default-Deny Ingress

Similarly, deny all inbound traffic to Pods in default: 
# deny-ingress.yaml
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: default-deny-ingress
spec:
  podSelector: {}
  policyTypes:
    - Ingress
Apply the policy: 
kubectl apply -f deny-ingress.yaml
From a Pod in kube-system, try to curl pod2 (NGINX): 
POD_IP=$(kubectl get pod pod2 -o jsonpath='{.status.podIP}')
kubectl run --rm -i test-client --image=centos --namespace=kube-system --restart=Never -- \
  curl --connect-timeout 1 http://$POD_IP
You should see a timeout.
Applying default-deny policies without specific allow rules can disrupt critical workloads. Always plan your policies carefully.

​
4. Allow Specific Egress and Ingress

Once Pods are isolated by default, define exceptions:
Policy NameDirectionAllowed Peer PodsPort
default-deny-egressEgressapp=nginx80
default-deny-ingressIngressapp=centos80

​
4.1 Permit Egress to NGINX Pods

Update deny-egress.yaml: 
spec:
  policyTypes:
    - Egress
  egress:
    - to:
        - podSelector:
            matchLabels:
              app: nginx
      ports:
        - protocol: TCP
          port: 80
Apply the updated policy: 
kubectl apply -f deny-egress.yaml

​
4.2 Permit Ingress from Management Pods

Update deny-ingress.yaml: 
spec:
  policyTypes:
    - Ingress
  ingress:
    - from:
        - podSelector:
            matchLabels:
              app: centos
      ports:
        - protocol: TCP
          port: 80
Apply the updated policy: 
kubectl apply -f deny-ingress.yaml

​
5. Verify Selective Connectivity

  1. Allowed: From pod1 → NGINX on port 80 
    kubectl exec -it pod1 -- curl --connect-timeout 1 http://$POD_IP
    You should see the NGINX welcome page.
  2. Blocked: From pod1 → NGINX on port 8080 
    kubectl exec -it pod1 -- curl --connect-timeout 1 http://$POD_IP:8080
    Connections on other ports will time out.

​
Recap

  • Kubernetes defaults to allow all ingress/egress traffic.
  • Default-deny policies lock down Pods by default.
  • Fine-tune communication by defining egress and ingress rules matching labels, ports, and namespaces.

Watch Video

Practice Lab