DevDayAug2022

Notes:

All the following steps are tested on k8s version 1.18.x
$ is used to present command prompt wherever required

Pre-requisite

Install kubectl

Install kubectl (v1.18.0) (as shown [here] https://kubernetes.io/docs/tasks/tools/#kubectl)

For MacOS

curl -LO "https://dl.k8s.io/release/v1.18.0/bin/darwin/amd64/kubectl"
chmod +x kubectl
./kubectl version --client
mv kubectl /usr/local/bin/kubectl

Create an alias ktl for kubectl

Setup Kops cluster on AWS

Setup and configure AWS CLI and use correct profile for your AWS account NOTE: This setup might incur cost. Please evaluate first. The kops setup that is created is following HA setup.

Ensure your AWS user has following permissions

(Follow Kops documentation here)

AmazonEC2FullAccess
AmazonRoute53FullAccess
AmazonS3FullAccess
IAMFullAccess
AmazonVPCFullAccess
AmazonSQSFullAccess
AmazonEventBridgeFullAccess

Create Terraform s3 state bucket

Note: Make sure to update the bucket name. It has to be globally unique Also, created by tag is added to identify which user created the resources. This is useful in case of shared AWS account (Company account)

export TF_S3_BUCKET_NAME=k8s-devday-2022-tfstate-bucket

cd tf-s3-state-bucket
terraform init
terraform plan -out tfplan -var="tf_s3_bucket_name=$TF_S3_BUCKET_NAME" -var="created_by_tag=prashantk"
terraform apply tfplan

Create bucket for KOPS state

export KOPS_STATE_BUCKET=devday-2022-kops-cluster-state

cd kops-state-bucket
terraform init -backend-config="bucket=$TF_S3_BUCKET_NAME" -backend-config=../backend.hcl
terraform plan -out tfplan -var="kops_state_bucket_name=$KOPS_STATE_BUCKET" -var="created_by_tag=prashantk"
terraform apply tfplan

Create kops cluster

Install kops binary as kops_1.18.2 (version 1.18.2) Create a public-private key pair (if not already created) This key can be used to ssh into the instances.

For MacOS

curl -Lo kops_1.18.2 https://github.com/kubernetes/kops/releases/download/v1.18.2/kops-darwin-amd64
chmod +x kops_1.18.2
sudo mv kops_1.18.2 /usr/local/bin/kops_1.18.2

NOTE: Make sure the cluster name ends with .k8s.local to use gossip based setup.

cd kops_cluster_tf
export CLUSTER_NAME=devday2022.cluster.k8s.local
export KOPS_STATE_STORE=s3://devday-2022-kops-cluster-state
kops_1.18.2 create cluster \
    --name ${CLUSTER_NAME} \
    --cloud=aws \
    --node-count 2 \
    --zones us-east-2a,us-east-2b,us-east-2c \
    --master-zones us-east-2a,us-east-2b,us-east-2c \
    --topology private \
    --networking calico \
    --ssh-public-key=~/.ssh/id_rsa.pub \
    --bastion="true" \
    --cloud-labels="created_by=prashantk,usage=k8sDevDayTalk2022" \
    --yes

Note: For production cluster, further improvements has to be done e.g. Restricting k8s API to specific CIDR or created internal ELB etc.

Validate cluster and access the cluster

Validate the cluster state

kops_1.18.2 validate cluster --name $CLUSTER_NAME --wait 10m

Validate kubectl access

ktl get nodes
NAME                                            STATUS   ROLES    AGE   VERSION
ip-172-20-105-15.ap-south-1.compute.internal    Ready    master   20m   v1.18.20
ip-172-20-112-185.ap-south-1.compute.internal   Ready    node     18m   v1.18.20
ip-172-20-45-107.ap-south-1.compute.internal    Ready    master   19m   v1.18.20
ip-172-20-50-30.ap-south-1.compute.internal     Ready    node     18m   v1.18.20
ip-172-20-86-189.ap-south-1.compute.internal    Ready    master   19m   v1.18.20

Allow access to port range from bastion to nodes

Create additional security groups to allow access to k8s nodes from bastion node on port ranged used by k8s for NodePort service.

Copy the node and bastion security group ids from AWS Console (Most likely named as nodes.devday2022.cluster.k8s.local and bastion.devday2022.cluster.k8s.local)

cd kops-k8s-node-bastion-access
terraform init -backend-config="bucket=$TF_S3_BUCKET_NAME" -backend-config=../backend.hcl
terraform plan -out tfplan -var="created_by_tag=prashantk" -var="node_security_group_id=sg-0feb31a6a945b15ed" -var="bastion_security_group_id=sg-069f7ca60a437c75f"
terraform apply tfplan

Note: The correct way to do this is to attach additional SGs through KOps configuration.

Deploy nginx ingress controller

Install the ingress controller with following command. (Install helm if required)

Dry run with debug flag

helm upgrade --install --dry-run --debug ingress-nginx ingress-nginx --repo https://kubernetes.github.io/ingress-nginx --namespace ingress-nginx \
  --create-namespace --version='<4'

Install nginx ingress controller

helm upgrade --install ingress-nginx ingress-nginx --repo https://kubernetes.github.io/ingress-nginx --namespace ingress-nginx \
  --create-namespace --version='<4'

Destroy cluster

The cluster can be destroyed as follows. Do it at the end of the session to ensure cost is not accumulated for unused clusters.

kops_1.18.2 delete cluster --name $CLUSTER_NAME --yes

Delete the state buckets created using terraform as well (terraform plan -out tfplan -destroy && terraform apply tfplan). Note: For bucket deletion you will have to add force_destroy = true on the bucket resource in terraform. This is required to allow terraform to delete non empty buckets. I did not add that in the code as someone might endup copied the code.

Session Contents

Example Application details

Refer Slides here

Pod Application access

Install customer Service

ktl apply -f resources/customer_pod.yaml
watch -n 5 "kubectl get pods"

Wait till pod is running

$ ktl get pods
NAME        READY   STATUS    RESTARTS   AGE
customers   1/1     Running   0          7m15s

How do I access the pod?

Access using pod SSH

$ ktl exec customers -- curl -s -S http://localhost:8080/customers/1
{
  "id": "CUST_0001",
  "name": "Customer Name",
  "phoneNumber": "+911233243335",
  "gender": "M"
}

(Install curl if not available. The image linux is debian i.e. apt-get update && apt-get -y install curl should work)

what do we need to access network application?

Refer Slides here

Pod IP

$ ktl get pods -o wide
NAME        READY   STATUS    RESTARTS   AGE     IP                NODE                                            NOMINATED NODE   READINESS GATES
customers   1/1     Running   0          3m31s   100.105.198.130   ip-172-20-112-185.ap-south-1.compute.internal   <none>           <none>

Pod Port

$ ktl get pods customers -o yaml
  ...
    ports:
    - containerPort: 8080
      protocol: TCP
  ...

Install supporting pod to access the pod

Run a supporting container to access the application.

$ ktl run nginx --image=nginx
$ ktl get pods -w
$ ktl get pods
NAME        READY   STATUS    RESTARTS   AGE
customers   1/1     Running   0          5m33s
nginx       1/1     Running   0          28s

Access the pod with the pod IP

$ podIP=$(ktl get pods customers -o json | jq -r ".status.podIP")
$ ktl exec nginx -- curl --no-progress-meter http://$podIP:8080/customers/1 | jq "."
{
  "id": "CUST_0001",
  "name": "Customer Name",
  "phoneNumber": "+911233243335",
  "gender": "M"
}

But I deploy multiple pods per service

Refer Slides here

Application as a deployment (instead of pods)

Remove pod

ktl delete pod customers

Deploy application as deployment with multiple replicas

$ ktl apply -f resources/customers_deployment.yaml
$ ktl get pods
NAME                         READY   STATUS    RESTARTS   AGE
customers-568c95b849-7jhh5   1/1     Running   0          8s
customers-568c95b849-x2q54   1/1     Running   0          8s
nginx                        1/1     Running   0          18h

Accessing application deployed as deployment

Get pod ips for deployment pods

$ ktl get pods -o wide
NAME                         READY   STATUS    RESTARTS   AGE     IP                NODE                                            NOMINATED NODE   READINESS GATES
customers-568c95b849-7rj67   1/1     Running   0          31s     100.112.44.132    ip-172-20-50-30.ap-south-1.compute.internal     <none>           <none>
customers-568c95b849-x5jqj   1/1     Running   0          31s     100.105.198.131   ip-172-20-112-185.ap-south-1.compute.internal   <none>           <none>
nginx                        1/1     Running   0          4m48s   100.112.44.131    ip-172-20-50-30.ap-south-1.compute.internal     <none>           <none>

Access application with any's ip address

podIP=100.97.47.70
ktl exec nginx -- curl --no-progress-meter http://$podIP:8080/customers/1 | jq "."

But accessing application with individual pod ip defeats the whole purpose of using deployment. Pods are ephemeral and can change IPs or scale out or scale in any time.

Adding a clusterIP service to access application with stable IP

Create service

$ ktl apply -f resources/customers_clusterip_service.yaml
$ ktl get services
NAME         TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)    AGE
customers    ClusterIP   100.64.222.177   <none>        8080/TCP   22s
kubernetes   ClusterIP   100.64.0.1       <none>        443/TCP    47h

Access customer service with stable IP and port

serviceIP=100.64.222.177
ktl exec nginx -- curl --no-progress-meter http://$serviceIP:8080/customers/1 | jq "."

Service Endpoints

An endpoint resource is created for every clusterIP service.

$ ktl get endpoints

Endpoint resource reflect the changes in the pod ips supported by the service. Delete a pod and see the changes are reflected or not.

$ ktl delete pod customers-568c95b849-4spdd
$ ktl describe endpoints customers

Service DNS & Well known port

$ ktl apply -f resources/customers_clusterip_WellknownPort.yaml
$ ktl get services
NAME         TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)   AGE
customers    ClusterIP   100.64.222.177   <none>        80/TCP    68m

Try accessing the application without the port (default http port 80 will be used)

$ serviceIP=100.64.222.177
$ ktl exec nginx -- curl --no-progress-meter http://$serviceIP/customers/1 | jq "."

What about Domain Name

Services are resolvable with service name inside k8s cluster. That means this should work.

ktl exec nginx -- curl --no-progress-meter http://customers/customers/1 | jq "."

How does DNS works in this case?

To add later.

But how do I access the service from outside?

How will we run the customer service using docker locally.

docker run -p 18080:8080 amitsadafule/customer-service:1.0.0

Access the application using link: http://localhost:18080/customers/1

Here we are doing a port mapping. Host port 18080 is mapped to port 8080 on the container.

Similarly, can be map container port to the k8s node port?

NotePort Service

ktl apply -f resources/customers_nodeport_service.yaml

Now the Port 30001 is mapped to the container pod on the k8s node and should be accessible from with in AWS as follows:

Refer bastion ssh steps here

ssh -A ubuntu@<bastion_dns>

k8sNodeIP=172.20.54.44
curl -s -S http://$k8sNodeIP:30001/customers/1 | jq "."

Can't really use single nodeIP

Refer Slides here

Load balancer Service

Instead of just using NodePort use load balancer service if provided by the cloud provider Create LB type service

ktl apply -f resources/customers_loadbalancer_service.yaml

Use dynamic Node Port

ktl apply -f resources/customers_loadbalancer_dynamicport.yaml

Install products service

ktl apply -f resources/products_deployment.yaml

Add service resource for the products service

ktl apply -f resources/products_clusterip_WellknownPort.yaml

Check products service connectivity

ktl exec nginx -- curl --no-progress-meter http://products/products/1 | jq "."
{
  "id": "PROD_0001",
  "name": "Product Name",
  "price": 10
}

Ingress Controller

Check ingress controller deployed

$ ktl get pods -n ingress-nginx
NAME                                        READY   STATUS    RESTARTS   AGE
ingress-nginx-controller-776cb4b9dd-4nm2c   1/1     Running   0          18h

Create ingress controllers for our services

Update customers service back to ClusterIP type service.

ktl apply -f resources/customers_clusterip_WellknownPort.yaml --force

Add the ingress resource for the customers service

ktl apply -f resources/customers_ingress.yaml

Add the ingress resource for the products service

ktl apply -f resources/products_ingress.yaml

Test connectivity by using ingress controller

Get ingress controller IPs address

$ ktl get pods -n ingress-controller -o wide
NAME                                        READY   STATUS    RESTARTS   AGE   IP               NODE                                          NOMINATED NODE   READINESS GATES
ingress-nginx-controller-776cb4b9dd-4nm2c   1/1     Running   0          21h   100.112.44.136   ip-172-20-50-30.ap-south-1.compute.internal   <none>           <none>

Send request to ingress controller pod IP instead of customers service or pod IP.

$ ingressControllerPod=100.112.44.136
$ ktl exec nginx -it -- curl --resolve test.ap-south-1.elb.amazonaws.com:80:$ingressControllerPod http://test.ap-south-1.elb.amazonaws.com/customers/1 | jq "."
{
  "id": "CUST_0001",
  "name": "Customer Name",
  "phoneNumber": "+911233243335",
  "gender": "M"
}

Now we are sending the request to the ingress controller Pod IP which is in turn sent to the customers service Pod.

Check the same for products service

$ ingressControllerPod=100.112.44.136
$ ktl exec nginx -it -- curl --resolve test.ap-south-1.elb.amazonaws.com:80:$ingressControllerPod http://test.ap-south-1.elb.amazonaws.com/products/1 | jq "."
{
  "id": "PROD_0001",
  "name": "Product Name",
  "price": 10
}

How traffic can be sent to ingress pod?

A load balancer service can be added for the ingress controller. Such a service is already created for ingress deployment

$ ktl get services -n ingress-nginx
NAME                                 TYPE           CLUSTER-IP       EXTERNAL-IP                                                                PORT(S)                      AGE
ingress-nginx-controller             LoadBalancer   100.71.89.35     a17676a4bcc284ea5a83aa3eed5bb866-1255914040.ap-south-1.elb.amazonaws.com   80:30060/TCP,443:31577/TCP   22h

That means nginx controller pods are reachable at address: http://a17676a4bcc284ea5a83aa3eed5bb866-1255914040.ap-south-1.elb.amazonaws.com

Let's try to access customers and product endpoint from load balancer DNS.

$ curl -s http://a17676a4bcc284ea5a83aa3eed5bb866-1255914040.ap-south-1.elb.amazonaws.com/customers/1 | jq "."
{
  "id": "CUST_0001",
  "name": "Customer Name",
  "phoneNumber": "+911233243335",
  "gender": "M"
}

$ curl -s http://a17676a4bcc284ea5a83aa3eed5bb866-1255914040.ap-south-1.elb.amazonaws.com/products/1 | jq "."
{
  "id": "PROD_0001",
  "name": "Product Name",
  "price": 10
}

Slides

Refer Slides here The event details are here The video recording of this event is uploaded on youtube here

Additional sections

Switching kubectl context and namespaces

List all contexts

ktl config get-contexts

or list just context names

ktl config get-contexts -o name

Switch context

ktl config use-context kind-kind18

kind-kind18 will become the current context.

List all namespaces

ktl get namespaces

Switch current kubectl context namespace

ktl config set-context --current --namespace=kube-system

kube-system will become current namespace for current context.

ktl get pods

Will provide pods running in current selected namespace for current context.

TIP: It is recommended to install kubectx to make it easy to switch between contexts and namespaces. Also install kube-ps1 to add context and namespace information to commandline.

Kops Bastion access

(Steps taken from kops documentation here: https://kops.sigs.k8s.io/bastion/)

Get the kops dns name for the bastion host

kops_1.18.2 toolbox dump -ojson | grep 'bastion.*elb.amazonaws.com'

Copy the DNSName.

Add key to ssh agent (this is same key we used while cluster creation)

$ ssh-add ~/.ssh/id_rsa

Verify if key is added

$ ssh-add -L

SSH into bastion node

ssh -A ubuntu@<bastion_elb_a_record>
ssh ubuntu@<master_ip>

Name		Name	Last commit message	Last commit date
Latest commit History 25 Commits
kopsSetup		kopsSetup
resources		resources
.gitignore		.gitignore
README.md		README.md

prashantkalkar/devDayAug2022

Folders and files

Latest commit

History

Repository files navigation

DevDayAug2022

Pre-requisite

Install kubectl

For MacOS

Setup Kops cluster on AWS

Ensure your AWS user has following permissions

Create Terraform s3 state bucket

Create bucket for KOPS state

Create kops cluster

For MacOS

Validate cluster and access the cluster

Allow access to port range from bastion to nodes

Deploy nginx ingress controller

Destroy cluster

Session Contents

Example Application details

Pod Application access

Install customer Service

How do I access the pod?

Access using pod SSH

what do we need to access network application?

Pod IP

Pod Port

Install supporting pod to access the pod

Access the pod with the pod IP

But I deploy multiple pods per service

Application as a deployment (instead of pods)

Remove pod

Deploy application as deployment with multiple replicas

Accessing application deployed as deployment

Adding a clusterIP service to access application with stable IP

Service Endpoints

Service DNS & Well known port

What about Domain Name

How does DNS works in this case?

But how do I access the service from outside?

NotePort Service

Can't really use single nodeIP

Load balancer Service

Use dynamic Node Port

Install products service

Ingress Controller

Create ingress controllers for our services

Test connectivity by using ingress controller

How traffic can be sent to ingress pod?

Slides

Additional sections

Switching kubectl context and namespaces

List all contexts

Switch context

List all namespaces

Switch current kubectl context namespace

Kops Bastion access

About

Resources

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages