Serverless vs GKE decision in GCP - CLI Comparison

Introduction

Choosing between serverless and Google Kubernetes Engine (GKE) helps you decide how to run your app. Serverless lets you run code without managing servers, while GKE gives you control over containers and infrastructure.

When you want to quickly run small apps or functions without worrying about servers or scaling.

When you need full control over containerized apps and want to manage scaling and updates yourself.

When your app has unpredictable traffic and you want automatic scaling without manual setup.

When you have complex apps that need multiple containers working together with custom networking.

When you want to reduce operational work and pay only for the exact compute time your code uses.

Commands

Deploys a container image to Cloud Run, which is a serverless platform. This runs your app without managing servers and scales automatically.

Terminal

gcloud run deploy my-serverless-app --image=gcr.io/my-project/my-image --region=us-central1 --platform=managed

Expected Output

Deploying container to Cloud Run service [my-serverless-app] in project [my-project] region [us-central1]

Done.

Service [my-serverless-app] revision [my-serverless-app-00001-abc] has been deployed and is serving traffic at https://my-serverless-app-abc.a.run.app

→

--image - Specifies the container image to deploy

→

--region - Sets the region where the service runs

→

--platform - Chooses the managed serverless platform

Grants your Google account admin rights on the GKE cluster so you can manage resources.

Terminal

kubectl create clusterrolebinding cluster-admin-binding --clusterrole cluster-admin --user $(gcloud config get-value account)

Expected Output

clusterrolebinding.rbac.authorization.k8s.io/cluster-admin-binding created

Creates a GKE cluster with 3 nodes in the specified zone. This cluster lets you run and manage containerized apps with control over infrastructure.

Terminal

gcloud container clusters create my-gke-cluster --zone us-central1-a --num-nodes 3

Expected Output

Creating cluster my-gke-cluster in us-central1-a...

Created [https://container.googleapis.com/v1/projects/my-project/zones/us-central1-a/clusters/my-gke-cluster].

To inspect the contents of your cluster, go to: https://console.cloud.google.com/kubernetes/workload_/gcloud/us-central1-a/my-gke-cluster?project=my-project

kubeconfig entry generated for my-gke-cluster

NAME             LOCATION       MASTER_VERSION  MASTER_IP       MACHINE_TYPE   NODE_VERSION  NUM_NODES  STATUS

my-gke-cluster   us-central1-a  1.26.8-gke.100  35.233.123.45   e2-medium      1.26.8-gke.100 3          RUNNING

→

--zone - Specifies the zone for the cluster

→

--num-nodes - Sets the initial number of nodes

Lists the nodes in your GKE cluster to verify the cluster is ready and nodes are running.

Terminal

kubectl get nodes

Expected Output

NAME                                       STATUS   ROLES    AGE   VERSION

my-gke-cluster-default-pool-1a2b3c4d-5e6f RUNNING  <none>   2m    v1.26.8-gke.100

Key Concept

If you remember nothing else from this pattern, remember: serverless is best for simple, auto-scaling apps without server management, while GKE is best when you need full control over containerized apps and infrastructure.

Common Mistakes

Trying to deploy a complex multi-container app directly to Cloud Run without splitting it into services.

Cloud Run is designed for single container services and does not support multi-container pods like Kubernetes.

Use GKE for multi-container apps or split your app into multiple Cloud Run services.

Creating a GKE cluster without setting up proper permissions for kubectl access.

Without cluster-admin binding, you cannot manage resources on the cluster.

Run the clusterrolebinding command to grant your user admin rights on the cluster.

Deploying serverless apps in a region far from your users causing latency.

Serverless apps scale automatically but latency depends on region proximity.

Choose a region close to your users when deploying Cloud Run services.

Summary

Deploy serverless apps with 'gcloud run deploy' for easy scaling and no server management.

Create GKE clusters with 'gcloud container clusters create' to run complex containerized apps with control.

Use 'kubectl' commands to manage and verify your GKE cluster nodes and permissions.