05 — Build and Push Docker Image

Script: 05_build_and_push.sh · Build, transfer, load — no registry needed

Builds your Docker image locally, transfers it to the VM, and loads it into k3s's container runtime — all without a container registry.

Required environment variables:

K3S_SSH_USER, K3S_SSH_HOSTNAME — SSH access to the VM
K3S_PROJECT_BASE_DIRECTORY — path to the Dockerfile
K3S_PROJECT_DOMAIN — your app's domain name
K3S_UNIQUE_DOCKER_IMAGE_NAME — image name (e.g., blog-farewell-cloud)
K3S_TEMP_BUILD_DIRECTORY — temp dir for the tar.gz

The Approach: No Registry Needed

Avoids registry accounts, private registry costs, and public image exposure. Perfect for single-server setups.

1. Build the Docker image

IMAGE_NAME="$K3S_UNIQUE_DOCKER_IMAGE_NAME:latest"
docker build --platform linux/amd64 -t $IMAGE_NAME $K3S_PROJECT_BASE_DIRECTORY

--platform linux/amd64 is important on Apple Silicon Macs (ARM). Your VM runs x86_64 — without this flag, the image would be ARM and wouldn't run on the VM.

2. Export and compress

docker save $IMAGE_NAME | gzip > $IMAGE_TAR_GZ

docker save exports all layers and metadata as a tar archive. Piping through gzip typically compresses a 500 MB image to ~150 MB.

3. Copy to the VM

scp $IMAGE_TAR_GZ $K3S_SSH_USER@$K3S_SSH_HOSTNAME:~

4. Load into containerd

ssh ... "gzip -d < ~/$K3S_UNIQUE_DOCKER_IMAGE_NAME.tar.gz | \
  sudo k3s ctr -n k8s.io images import -"

k3s uses containerd, not Docker. k3s ctr is the containerd CLI:

gzip -d — decompress
k3s ctr -n k8s.io images import - — import into containerd's image store
-n k8s.io — the containerd namespace where k3s looks for images

5. Verify

ssh ... "sudo k3s ctr -n k8s.io images ls | grep $K3S_UNIQUE_DOCKER_IMAGE_NAME"

⚠️ Why imagePullPolicy: Never? The Deployment in step 06 must set this to prevent Kubernetes from trying to pull from Docker Hub. The image exists only locally on the VM.