↓Skip to main content
Hugo’s IT journal

Networking in Docker and Kubernetes: Port Expose and Pod Communication

k8s docker
Hugo
Author
Hugo
DevOps Engineer in London
Table of Contents

Docker
#

ubuntu@ip-172-31-6-200:~$ docker run --rm -it -P httpd

ubuntu@ip-172-31-6-200:~$ docker ps 
CONTAINER ID   IMAGE     COMMAND              CREATED          STATUS          PORTS                                     NAMES
6f6f53ee7dde   httpd     "httpd-foreground"   12 seconds ago   Up 10 seconds   0.0.0.0:32768->80/tcp, :::32768->80/tcp   dreamy_goodall
  • Docker automatically knows which ports a container needs due to the EXPOSE directive in the container’s Dockerfile. (httpd’s Dockerfile)
  • Removing EXPOSE from the Dockerfile will not publish the port automatically anymore when using the -P flag.
  • Howerver, containers can still reach ports within the container via its IP/alias.
  • Docker does not enforce any security policies between containers.

We can test it by:

  • Create a custom network
  • Start two containers, connected to the same network
  • Install net-tools on both containers
  • Listen to port 3000 on container u1
  • Connect to u1:3000 on container u2
  • Connection successful!

On host:

docker network create mynetwork

docker run -it --rm --name u1 --network mynetwork ubuntu
docker run -it --rm --name u2 --network mynetwork ubuntu

On both container:

apt update
apt install -y iproute2 netcat

On u1:

root@b7bfb1fa159e:/# while true; do nc -l 3000; done

On u2:

root@1bd1cee0ce53:/# nc b7bfb1fa159e 3000 -v
Connection to b7bfb1fa159e (172.18.0.2) 3000 port [tcp/*] succeeded!

Kubernetes
#

  • Pods can reach each other, even in different namespaces.
  • Containers only need to listen on a port inside, and there is not nessessary to specify the exposed port in the Dockerfile or mention the pod config.
  • The ports field in the container object is used to list the ports to expose from the container.
  • Not specifying a port here does not prevent that port from being exposed.
  • Any port that is listening on the default 0.0.0.0 address inside a container will be accessible from the network.
  • Modifying the ports array with a strategic merge patch may corrupt the data.
  • The ports field is more like a self-document purpose.

Reference: https://kubernetes.io/docs/reference/kubernetes-api/workload-resources/pod-v1/#ports

We can re-test it in k8s, starting the pods in different namespace.

On host:

kubectl create ns ns1
kubectl create ns ns2

kubectl run -n ns1 -it u1 --image=ubuntu
kubectl run -n ns2 -it u2 --image=ubuntu

repeat what we did in Docker - u2 is able to connect to u1:3000 with ip.