Kubernetes Virtual Kubelet with ACI

Azure Container Instances (ACI) provide a hosted environment for running containers in Azure. When using ACI, there is no need to manage the underlying compute infrastructure, Azure handles this management for you. When running containers in ACI, you are charged by the second for each running container.

The Azure Container Instances provider for the Virtual Kubelet configures an ACI instance as a node in any Kubernetes cluster. When using the Virtual Kubelet ACI provider, pods can be scheduled on an ACI instance as if the ACI instance is a standard Kubernetes node. This configuration allows you to take advantage of both the capabilities of Kubernetes and the management value and cost benefit of ACI.

This document details configuring the Virtual Kubelet ACI provider.

Current feature set

Virtual Kubelet's ACI provider relies heavily on the feature set that Azure Container Instances provide. Please check the Azure documentation accurate details on region availability, pricing and new features. The list here attempts to give an accurate reference for the features we support in ACI and the ACI provider within Virtual Kubelet.

features

Volumes: empty dir, github repo, Azure Files
Secure env variables, config maps
Bring your own virtual network (VNet)
Network security group support
Basic Azure Networking support within AKS virtual node
Exec support for container instances
Azure Monitor integration or formally known as OMS

Limitations

Using service principal credentials to pull ACR images (see workaround)
Liveness and readiness probes
Limitations with VNet
VNet peering
Argument support for exec
Init containers
Host aliases support

Prerequisites

Kubernetes cluster up and running (can be an AKS cluster or minikube) and that kubectl is already configured.
A Microsoft Azure account.
Install the Azure CLI.
Install the Kubernetes CLI.
Install the Helm CLI.

You may also use Azure cloud shell which has the above tools already installed.

Install the Azure CLI

Install az by following the instructions for your operating system. See the full installation instructions if yours isn't listed below.

MacOS

brew install azure-cli

Windows

Download and run the Azure CLI Installer (MSI).

Ubuntu 64-bit

Add the azure-cli repo to your sources:

AZ_REPO=$(lsb_release -cs)
echo "deb [arch=amd64] https://packages.microsoft.com/repos/azure-cli/ $AZ_REPO main" |
    sudo tee /etc/apt/sources.list.d/azure-cli.list

Run the following commands to install the Azure CLI and its dependencies:

sudo apt-get install apt-transport-https
curl -sL https://packages.microsoft.com/keys/microsoft.asc |
    gpg --dearmor |
    sudo tee /etc/apt/trusted.gpg.d/microsoft.gpg > /dev/null
sudo apt-get update && sudo apt-get install azure-cli

Install the Kubernetes CLI

Install kubectl by running the following command:

az aks install-cli

Install the Helm 3.x CLI

Helm is a tool for installing pre-configured applications on Kubernetes. Install helm for macOS, Windows, or Linux via binary releases or package managers or check the detailed Helm install guide for more options including building from source.

Cluster and Azure Account Setup

Now that we have all the tools, we will set up your Azure account to work with ACI.

Configure your Azure account

First let's identify your Azure subscription and save it for use later on in the quickstart.

Run az login and follow the instructions in the command output to authorize az to use your account
List your Azure subscriptions:
```
az account list -o table
```

Copy your subscription ID and save it in an environment variable:

Bash

export AZURE_SUBSCRIPTION_ID="<SubscriptionId>"

PowerShell

$env:AZURE_SUBSCRIPTION_ID = "<SubscriptionId>"

Enable ACI in your subscription:

az provider register -n Microsoft.ContainerInstance

Quick set up with AKS

Linux containers with Virtual Nodes

Azure Kubernetes Service has an efficient way of setting up virtual kubelet with the ACI provider with a feature called virtual node. You can easily install a virtual node that will deploy Linux workloads to ACI. The pods that spin out will automatically get private IPs and will be within a subnet that is within the AKS cluster's Virtual Network. Virtual Nodes is the recommended path for using the ACI provider on Linux AKS clusters.

To install virtual node in the Azure portal go here. To install virtual node in the Azure CLI go here.

Manual set-up

Create a Resource Group for ACI

To use Azure Container Instances, you must provide a resource group. Create one with the az cli using the following command.

export ACI_REGION=eastus
az group create --name aci-group --location "$ACI_REGION"
export AZURE_RG=aci-group

Create a service principal

This creates an identity for the Virtual Kubelet ACI provider to use when provisioning resources on your account on behalf of Kubernetes. If you are provisioning Virtual Kubelet on AKS, please note: 1) This step is optional 2) The service principal will expire within a year unless updated.

Create a service principal with RBAC enabled for the quickstart:

   az ad sp create-for-rbac --name virtual-kubelet-quickstart -o table

Save the values from the command output in environment variables:

Bash

export AZURE_TENANT_ID=<Tenant>
export AZURE_CLIENT_ID=<AppId>
export AZURE_CLIENT_SECRET=<Password>

PowerShell

$env:AZURE_TENANT_ID = "<Tenant>"
$env:AZURE_CLIENT_ID = "<AppId>"
$env:AZURE_CLIENT_SECRET = "<Password>"

Deployment of the ACI provider in your cluster

Run these commands to deploy the virtual kubelet which connects your Kubernetes cluster to Azure Container Instances.

export VK_RELEASE=virtual-kubelet-latest

Grab the public master URI for your Kubernetes cluster and save the value.

export MASTER_URI="$(kubectl cluster-info | awk '/Kubernetes control plane/{print $7}' | sed "s,\x1B\[[0-9;]*[a-zA-Z],,g")"

If your cluster is an AKS cluster:

export RELEASE_NAME=virtual-kubelet
export VK_RELEASE=virtual-kubelet-latest
export NODE_NAME=virtual-kubelet
export CHART_URL=https://github.com/virtual-kubelet/azure-aci/raw/master/charts/$VK_RELEASE.tgz

# Linux Virtual Node
helm install "$RELEASE_NAME" "$CHART_URL" \
  --set provider=azure \
  --set providers.azure.targetAKS=true \
  --set providers.azure.masterUri=$MASTER_URI \
  --set nodeName=$NODE_NAME

# Windows Virtual Node
helm install "$RELEASE_NAME" "$CHART_URL" \
  --set provider=azure \
  --set "nodeOsType=Windows" \
  --set providers.azure.targetAKS=true \
  --set providers.azure.masterUri=$MASTER_URI \
  --set nodeName="${NODE_NAME}-win"

For any other type of Kubernetes cluster:

RELEASE_NAME=virtual-kubelet
NODE_NAME=virtual-kubelet
CHART_URL=https://github.com/virtual-kubelet/azure-aci/raw/master/charts/$VK_RELEASE.tgz

# Linux Virtual Node
helm install "$RELEASE_NAME" "$CHART_URL" \
  --set provider=azure \
  --set rbac.install=true \
  --set providers.azure.targetAKS=false \
  --set providers.azure.aciResourceGroup=$AZURE_RG \
  --set providers.azure.aciRegion=$ACI_REGION \
  --set providers.azure.tenantId=$AZURE_TENANT_ID \
  --set providers.azure.subscriptionId=$AZURE_SUBSCRIPTION_ID \
  --set providers.azure.clientId=$AZURE_CLIENT_ID \
  --set providers.azure.clientKey=$AZURE_CLIENT_SECRET \
  --set providers.azure.masterUri=$MASTER_URI \
  --set nodeName=$NODE_NAME

# Windows Virtual Node
helm install "$RELEASE_NAME" "$CHART_URL" \
  --set provider=azure \
  --set rbac.install=true \
  --set "nodeOsType=Windows" \
  --set providers.azure.targetAKS=false \
  --set providers.azure.aciResourceGroup=$AZURE_RG \
  --set providers.azure.aciRegion=$ACI_REGION \
  --set providers.azure.tenantId=$AZURE_TENANT_ID \
  --set providers.azure.subscriptionId=$AZURE_SUBSCRIPTION_ID \
  --set providers.azure.clientId=$AZURE_CLIENT_ID \
  --set providers.azure.clientKey=$AZURE_CLIENT_SECRET \
  --set providers.azure.masterUri=$MASTER_URI \
  --set nodeName="${NODE_NAME}-win"

If your cluster has RBAC disabled set rbac.install=false

Output:

NAME:   virtual-kubelet
LAST DEPLOYED: Thu Feb 15 13:17:01 2018
NAMESPACE: default
STATUS: DEPLOYED

RESOURCES:
==> v1/Secret
NAME                             TYPE    DATA  AGE
virtual-kubelet-virtual-kubelet  Opaque  3     1s

==> v1beta1/Deployment
NAME                             DESIRED  CURRENT  UP-TO-DATE  AVAILABLE  AGE
virtual-kubelet-virtual-kubelet  1        1        1           0          1s

==> v1/Pod(related)
NAME                                              READY  STATUS             RESTARTS  AGE
virtual-kubelet-virtual-kubelet-7bcf5dc749-6mvgp  0/1    ContainerCreating  0         1s


NOTES:
The virtual kubelet is getting deployed on your cluster.

To verify that virtual kubelet has started, run:

```bash
  kubectl --namespace=default get pods -l "app=virtual-kubelet-virtual-kubelet"

Create an AKS cluster with VNet

Run the following commands to create an AKS cluster with a new Azure virtual network. Also, create two subnets. One will be delegated to the cluster and the other will be delegated to Azure Container Instances.

Create an Azure virtual network and subnets

First, set the following variables for your VNet range and two subnet ranges within that VNet.

  export VNET_RANGE=10.0.0.0/8  
  export CLUSTER_SUBNET_RANGE=10.240.0.0/16
  export ACI_SUBNET_RANGE=10.241.0.0/16
  export VNET_NAME=myAKSVNet
  export CLUSTER_SUBNET_NAME=myAKSSubnet
  export ACI_SUBNET_NAME=myACISubnet
  export AKS_CLUSTER_RG=myresourcegroup
  export KUBE_DNS_IP=10.0.0.10

Run the following command to create a virtual network within Azure, and a subnet within that VNet. The subnet will be dedicated to the nodes in the AKS cluster.

az network vnet create \
--resource-group $AKS_CLUSTER_RG \
--name $VNET_NAME \
--address-prefixes $VNET_RANGE \
--subnet-name $CLUSTER_SUBNET_NAME \
--subnet-prefix $CLUSTER_SUBNET_RANGE

Create a subnet that will be delegated to just resources within ACI, note that this needs to be an empty subnet, but within the same VNet that you already created.

az network vnet subnet create \
    --resource-group $AKS_CLUSTER_RG \
    --vnet-name $VNET_NAME \
    --name $ACI_SUBNET_NAME \
    --address-prefix $ACI_SUBNET_RANGE

Create a service principal (OPTIONAL)

Create an Azure Active Directory service principal to allow AKS to interact with other Azure resources. You can use a pre-created service principal too.

az ad sp create-for-rbac -n "virtual-kubelet-sp" --skip-assignment

The output should look similar to the following.

{
  "appId": "bef76eb3-d743-4a97-9534-03e9388811fc",
  "displayName": "azure-cli-2018-08-29-22-29-29",
  "name": "http://azure-cli-2018-08-29-22-29-29",
  "password": "1d257915-8714-4ce7-xxxxxxxxxxxxx",
  "tenant": "72f988bf-86f1-41af-91ab-2d7cd011db48"
}

Save the output values from the command output in environment variables.

export AZURE_TENANT_ID=<Tenant>
export AZURE_CLIENT_ID=<AppId>
export AZURE_CLIENT_SECRET=<Password>

These values can be integrated into the az aks create as a field --service-principal $AZURE_CLIENT_ID.

Integrating Azure VNet Resource

If you want to integrate an already created Azure VNet resource with your AKS cluster than follow these steps. Grab the virtual network resource id with the following command:

az network vnet show --resource-group $AKS_CLUSTER_RG --name $VNET_NAME --query id -o tsv

Grant access to the AKS cluster to use the virtual network by creating a role and assigning it.

az role assignment create --assignee $AZURE_CLIENT_ID --scope <vnetId> --role "Network Contributor"

Create an AKS cluster with a virtual network

Grab the id of the cluster subnet you created earlier with the following command.

az network vnet subnet show --resource-group $AKS_CLUSTER_RG --vnet-name $VNET_NAME --name $CLUSTER_SUBNET_NAME --query id -o tsv

Save the entire output starting with "/subscriptions/..." in the following environment variable.

export VNET_SUBNET_ID=<subnet-resource>

Use the following command to create an AKS cluster with the virtual network you've already created.

az aks create \
    --resource-group myResourceGroup \
    --name myAKSCluster \
    --node-count 1 \
    --network-plugin azure \
    --service-cidr 10.0.0.0/16 \
    --dns-service-ip $KUBE_DNS_IP \
    --docker-bridge-address 172.17.0.1/16 \
    --vnet-subnet-id $VNET_SUBNET_ID \
    --client-secret $AZURE_CLIENT_SECRET

Deploy Virtual Kubelet

Manually deploy the Virtual Kubelet, the following env. variables have already been set earlier. You do need to pass through the subnet you created for ACI earlier, otherwise the container instances will not be able to participate with the other pods within the cluster subnet.

Grab the public master URI for your Kubernetes cluster and save the value.

kubectl cluster-info
export MASTER_URI=<public uri>

Set the following values for the helm chart.

RELEASE_NAME=virtual-kubelet
NODE_NAME=virtual-kubelet
CHART_URL=https://github.com/virtual-kubelet/azure-aci/raw/master/charts/$VK_RELEASE.tgz

If your cluster is an AKS cluster:

helm install "$RELEASE_NAME" "$CHART_URL" \
  --set provider=azure \
  --set providers.azure.targetAKS=true \
  --set providers.azure.vnet.enabled=true \
  --set providers.azure.vnet.subnetName=$ACI_SUBNET_NAME \
  --set providers.azure.vent.subnetCidr=$ACI_SUBNET_RANGE \
  --set providers.azure.vnet.clusterCidr=$CLUSTER_SUBNET_RANGE \
  --set providers.azure.vnet.kubeDnsIp=$KUBE_DNS_IP \
  --set providers.azure.masterUri=$MASTER_URI \
  --set nodeName=$NODE_NAME

For any other type of cluster:

# the resource group where the virtual network is located in
export ACI_VNET_RESOURCE_GROUP=<resource group>

# the virtual network name where container will deploy to
export ACI_VNET_NAME=<virtual network name>
# subnet name where ACI will deploy to. Virtual Kubelet will automatically create subnet resource if it not exists
export ACI_SUBNET_NAME=<subnet name>
# subnet's IP range, for example 10.1.0.0/16. You don't need specific this system variable if subnet has been exists
export ACI_SUBNET_RANGE=<subnet name where ACI will run in>

helm install "$RELEASE_NAME" "$CHART_URL" \
  --set provider=azure \
  --set providers.azure.targetAKS=false \
  --set providers.azure.vnet.enabled=true \
  --set providers.azure.vnet.vnetResourceGroup=$ACI_VNET_RESOURCE_GROUP \
  --set providers.azure.vnet.vnetName=$ACI_VNET_NAME \
  --set providers.azure.vnet.subnetName=$ACI_SUBNET_NAME \
  --set providers.azure.vent.subnetCidr=$ACI_SUBNET_RANGE \
  --set providers.azure.vnet.kubeDnsIp=$KUBE_DNS_IP \
  --set providers.azure.tenantId=$AZURE_TENANT_ID \
  --set providers.azure.subscriptionId=$AZURE_SUBSCRIPTION_ID \
  --set providers.azure.aciResourceGroup=$AZURE_RG \
  --set providers.azure.aciRegion=$ACI_REGION \
  --set providers.azure.masterUri=$MASTER_URI
  --set providers.azure.clientId=$AZURE_CLIENT_ID \
  --set providers.azure.clientKey=$AZURE_CLIENT_SECRET \

Validate the Virtual Kubelet ACI provider

To validate that the Virtual Kubelet has been installed, return a list of Kubernetes nodes using the kubectl get nodes command.

kubectl get nodes

Output:

NAME                                        STATUS    ROLES     AGE       VERSION
virtual-kubelet-aci-linux                   Ready     agent     2m        v1.13.1
aks-nodepool1-39289454-0                    Ready     agent     22h       v1.12.6
aks-nodepool1-39289454-1                    Ready     agent     22h       v1.12.6
aks-nodepool1-39289454-2                    Ready     agent     22h       v1.12.6

Schedule a pod in ACI

Create a file named virtual-kubelet-test.yaml and copy in the following YAML.

apiVersion: v1
kind: Pod
metadata:
  name: helloworld
spec:
  containers:
  - image: microsoft/aci-helloworld
    imagePullPolicy: Always
    name: helloworld
    resources:
      requests:
        memory: 1G
        cpu: 1
    ports:
    - containerPort: 80
      name: http
      protocol: TCP
    - containerPort: 443
      name: https
  dnsPolicy: ClusterFirst
  nodeSelector:
    kubernetes.io/role: agent
    beta.kubernetes.io/os: linux
    type: virtual-kubelet
  tolerations:
  - key: virtual-kubelet.io/provider
    operator: Exists
  - key: azure.com/aci
    effect: NoSchedule

Notice that Virtual-Kubelet nodes are tainted by default to avoid unexpected pods running on them, i.e. kube-proxy, other virtual-kubelet pods, etc. To schedule a pod to them, you need to add the toleration to the pod spec and a node selector:

  nodeSelector:
    kubernetes.io/role: agent
    beta.kubernetes.io/os: linux
    type: virtual-kubelet
  tolerations:
  - key: virtual-kubelet.io/provider
    operator: Exists
  - key: azure.com/aci
    effect: NoSchedule

Private registry

If your image is on a private registry, you need to add a kubernetes secret to your cluster and reference it in the pod spec.

  spec:
    containers:
    - name: aci-helloworld
      image: <registry name>.azurecr.io/aci-helloworld:v1
      ports:
      - containerPort: 80
    imagePullSecrets:
    - name: <K8 secret name>

Run the application with the kubectl create command.

kubectl create -f virtual-kubelet-test.yaml

Use the kubectl get pods command with the -o wide argument to output a list of pods with the scheduled node.

kubectl get pods -o wide

Notice that the helloworld pod is running on the virtual-kubelet node.

NAME                                            READY     STATUS    RESTARTS   AGE       IP             NODE
aci-helloworld-2559879000-8vmjw                 1/1       Running   0          39s       52.179.3.180   virtual-kubelet-aci-linux

If the AKS cluster was configured with a virtual network, then the output will look like the following. The container instance will get a private ip rather than a public one.

NAME                            READY     STATUS    RESTARTS   AGE       IP           NODE
aci-helloworld-9b55975f-bnmfl   1/1       Running   0          4m        10.241.0.4   virtual-kubelet-aci-linux

To validate that the container is running in an Azure Container Instance, use the az container list Azure CLI command.

az container list -o table

Output:

Name                             ResourceGroup    ProvisioningState    Image                     IP:ports         CPU/Memory       OsType    Location
-------------------------------  ---------------  -------------------  ------------------------  ---------------  ---------------  --------  ----------
helloworld-2559879000-8vmjw  myResourceGroup    Succeeded            microsoft/aci-helloworld  52.179.3.180:80  1.0 core/1.5 gb  Linux     eastus

Work around for the virtual kubelet pod

If your pod that's scheduled onto the Virtual Kubelet node is in a pending state please add this workaround to your Virtual Kubelet pod spec.

First, grab the logs from your virtual kubelet pod, with the following command. You can get the pod name from the kubectl get pods command

kubectl logs virtual-kubelet-virtual-kubelet-7bcf5dc749-6mvgp

Stream or pod watcher errors

If you see the following errors in the logs:

ERROR: logging before flag.Parse: E0914 00:02:01.546132       1 streamwatcher.go:109] Unable to decode an event from the watch stream: stream error: stream ID 181; INTERNAL_ERROR
time="2018-09-14T00:02:01Z" level=error msg="Pod watcher connection is closed unexpectedly" namespace= node=virtual-kubelet-myconnector-linux operatingSystem=Linux provider=azure

Then copy the master URI with cluster-info.

kubectl cluster-info

Output:

Kubernetes master is running at https://aksxxxx-xxxxx-xxxx-xxxxxxx.hcp.uksouth.azmk8s.io:443

Edit virtual kubelet deployment by first getting the deployment name.

kubectl get deploy

Output:

NAME                           DESIRED   CURRENT   UP-TO-DATE   AVAILABLE   AGE
virtual-kubelet-virtual-kubelet 1         1         1            1           5d
aci-helloworld                  1         1         1            0           12m

Edit the deployment.

kubectl edit deploy virtual-kubelet-virtual-kubelet

Add the following name and value to the deployment in the environment section. Use your copied AKS master URI.

--name: MASTER_URI
  value: https://aksxxxx-xxxxx-xxxx-xxxxxxx.hcp.uksouth.azmk8s.io:443

Taint deprecated errors

If you see the following errors in the logs:

Flag --taint has been deprecated, Taint key should now be configured using the VK_TAINT_KEY environment variable

Then edit your virtual kubelet deployment by first grabbing the deployment name.

kubectl get deploy

Output:

NAME                           DESIRED   CURRENT   UP-TO-DATE   AVAILABLE   AGE
virtual-kubelet-virtual-kubelet 1         1         1            1           5d
aci-helloworld                  1         1         1            0           12m

Edit the virtual kubelet deployment.

kubectl edit deploy virtual-kubelet-virtual-kubelet

Add the following as an environment variable within the deployment.

--name: VK_TAINT_KEY
  value: azure.com/aci

Also, delete the following argument in your pod spec:

- --taint
  - azure.com/aci

Remove the Virtual Kubelet

You can remove your Virtual Kubelet node by deleting the Helm deployment. Run the following command:

helm uninstall virtual-kubelet

If you used Virtual Nodes, can follow the steps here to remove the add-on

Name		Name	Last commit message	Last commit date
Latest commit History 472 Commits
.circleci		.circleci
charts		charts
client		client
cmd/virtual-kubelet		cmd/virtual-kubelet
e2e		e2e
hack		hack
helm		helm
provider		provider
.dockerignore		.dockerignore
.gitignore		.gitignore
Dockerfile		Dockerfile
LICENSE		LICENSE
Makefile		Makefile
README.md		README.md
go.mod		go.mod
go.sum		go.sum

License

telotlik/azure-aci

Folders and files

Latest commit

History

Repository files navigation

Kubernetes Virtual Kubelet with ACI

Table of Contents

Current feature set

features

Limitations

Prerequisites

Install the Azure CLI

MacOS

Windows

Ubuntu 64-bit

Install the Kubernetes CLI

Install the Helm 3.x CLI

Cluster and Azure Account Setup

Configure your Azure account

Quick set up with AKS

Linux containers with Virtual Nodes

Manual set-up

Create a Resource Group for ACI

Create a service principal

Deployment of the ACI provider in your cluster

Create an AKS cluster with VNet

Create an Azure virtual network and subnets

Create a service principal (OPTIONAL)

Integrating Azure VNet Resource

Create an AKS cluster with a virtual network

Deploy Virtual Kubelet

Validate the Virtual Kubelet ACI provider

Schedule a pod in ACI

Private registry

Work around for the virtual kubelet pod

Stream or pod watcher errors

Taint deprecated errors

Remove the Virtual Kubelet

About

Resources

License

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages