Now that the AKS cluster has been deployed, the next step to configure a GitOps management solution on our cluster, Flux in this case.
GitOps allows a team to author Kubernetes manifest files, persist them in their git repo, and have the Kubernetes manifest files automatically applied to a cluster as changes are made in the source repository. In this reference implementation, Flux is used to managing cluster-level configurations. This is distinct from workload-level concerns, which would be possible as well to manage via Flux, and would typically be done by additional Flux operators in the cluster. The namespace cluster-baseline-settings
will be used to provide a logical division of the cluster configuration from workload configuration. Example configureation that would be managed with Flux are:
- Cluster Role Bindings for the AKS-managed Azure AD integration
- AAD Pod Identity
- CSI driver and Azure KeyVault CSI Provider
- KeyVault Flex Volume
- The workload's namespace and its resource quotas
- Zero trust Kubernetes Network Policies within the backend namespace
- Kubernetes RBAC permissions for Azure Application Insights
- Resource Quotas for the workload namespaces
-
Install
kubectl
1.19 or newer. (kubctl
supports +/-1 Kubernetes version.)sudo az aks install-cli kubectl version --client
-
Get the cluster name.
export AKS_CLUSTER_NAME=$(az deployment group show --resource-group rg-shipping-dronedelivery -n cluster-stamp --query properties.outputs.aksClusterName.value -o tsv)
-
Get AKS
kubectl
credentials (as a user that has admin permissions to the cluster).In the Azure Active Directory Integration step, we placed our cluster under AAD group-backed RBAC. This is the first time we see this used.
az aks get-credentials
allows you to usekubectl
commands against your cluster. Without the AAD integration, you'd have to use--admin
here, which isn't what we want to happen. Here, you'll log in with a user added to the Azure AD security group used to back the Kubernetes RBAC admin role. Executing the firstkubectl
command below invokes the AAD login process to auth the user of your choice, which will then be checked against Kubernetes RBAC to act. The user you choose to log in with must be a member of the AAD group bound to thecluster-admin
ClusterRole. For simplicity could either use the "break-glass" admin user created in Azure Active Directory Integration (dronedelivery-admin
) or any user you assign to thecluster-admin
group assignment in youruser-facing-cluster-role-aad-group.yaml
file. If you skipped those steps, use--admin
to proceed, but proper AAD group-based RBAC access is a critical security function that you should invest time in setting up.az aks get-credentials -g rg-shipping-dronedelivery -n $AKS_CLUSTER_NAME
⚠️ At this point, two important steps are happening:- The
az aks get-credentials
command will fetch akubeconfig
context file containing references to your AKS cluster. - To actually use the cluster, you will need to authenticate. For that, run any
kubectl
commands, which at this stage will prompt you to authenticate against Azure Active Directory. For example, run the following command:
kubectl get nodes
Once the authentication happens successfully, some new items will be added to your
kubeconfig
file, such as anaccess-token
with an expiration period. For more information on how this process works in Kubernetes, see https://kubernetes.io/docs/reference/access-authn-authz/authentication/#openid-connect-tokens) - The
-
Create the cluster baseline settings namespace.
# Verify the user you logged in with has the appropriate permissions, which should result in a "yes" response. # If you receive "no" to this command, check which user you authenticated as and ensure they are # assigned to the Azure AD Group you designated for cluster admins. kubectl auth can-i create namespace -A kubectl create namespace cluster-baseline-settings
-
Import cluster management images to your container registry.
Public container registries are subject to faults such as outages (no SLA) or request throttling. Interruptions like these can be crippling for a system that needs to pull an image right now. To minimize the risks of using public registries, store all applicable container images in a registry that you control, such as the SLA-backed Azure Container Registry.
# Get your ACR cluster name export ACR_NAME=$(az deployment group show --resource-group rg-shipping-dronedelivery -n cluster-stamp --query properties.outputs.containerRegistryName.value -o tsv) # Import cluster management images hosted in public container registries az acr import --source docker.io/library/memcached:1.5.20 -n $ACR_NAME az acr import --source docker.io/fluxcd/flux:1.19.0 -n $ACR_NAME az acr import --source docker.io/weaveworks/kured:1.4.0 -n $ACR_NAME
-
Deploy Flux.
If you used your own fork of this GitHub repo, update the
flux.yaml
file to include reference to your own repo and change the URL below to point to yours as well. Also, since Flux will begin processing the manifests incluster-baseline-settings/
, now would be a good time to:- update the
<replace-with-an-aad-group-object-id-for-this-cluster-role-binding>
placeholder inuser-facing-cluster-role-aad-group.yaml
with the Object IDs for the Azure AD group(s) you created for management purposes. If you don't, the manifest will still apply, but AAD integration will not be mapped to your specific AAD configuration. - Update three
image
manifest references to your container registry instead of the default public container registry. See comment in each file for instructions.- update the two
image:
values influx.yaml
. - update the one
image:
values inkured-1.4.0-dockerhub.yaml
.
- update the two
⚠️ Deploying the flux configuration using theflux.yaml
file unmodified from this repo will configure the Flux deployment to take dependencies on public container registries. This configuration is generally okay for exploratory/testing but not suitable for production. Before production use, ensure all image references are from your container registry (as imported in the previous step) or another that you feel confident relying on.kubectl apply -f https://raw.githubusercontent.com/mspnp/aks-fabrikam-dronedelivery/main/cluster-baseline-settings/flux.yaml
- update the
-
Wait for Flux to be ready before proceeding.
kubectl wait --namespace cluster-baseline-settings --for=condition=ready pod --selector=app.kubernetes.io/name=flux --timeout=90s
Generally speaking, this will be the last time you should need to use kubectl
for day-to-day operations on this cluster (outside of break-fix situations). Between ARM for Azure Resource definitions and the application of manifests via Flux, all configuration activities can be performed without the need to use kubectl
. You will, however, see us use it for the upcoming workload deployment. This is because the SDLC component of workloads are not in scope for this reference implementation, as this is focused the infrastructure and baseline configuration.