-
Notifications
You must be signed in to change notification settings - Fork 10.3k
Deploy to Azure Kubernetes Service (AKS)
CONTENT
- Specifications
- Previous Versions
- Pre-Requisites
- Create The Kubernetes Cluster in AKS
- Configure Kubectl
- Install Helm
- Install NGINX Ingress Controller
- Install eShopOnContainers Using Helm
- Known Behaviours
- Explore eShopOnContainers
- Optional / Customize The Deployment
We have upgraded K8s scripts to work on the latest versions of Kubernetes with AKS. Current dev
branch has been tested with the following specifications :
Component | Versions |
---|---|
Azure CLI | 2.9.1 |
kubectl | 1.22.5 |
Kubernetes Services | AKS |
Container Registry | ACS |
Kubernetes Version | 1.21.7 |
Kubernetes-Dashboard | 2.0 |
Helm | 3.8.0 |
Nginx Ingress controller | 1.1.1 |
Nginx | 4.0.15 |
You can still refer old version of kubernetes related scripts under k8s/archived directory. For more details, please visit archived page.
IMPORTANT: Directory k8s/archived
will not be supported in the future.
If you don't have powershell
on your machine you can get that from the official documentation
Note: If you want you can use bash
shell as well for the deployment.
If you already have an access to Azure Subscription that's good or else, signup for a free account
You can follow official documentation to complete the installation.
If you already don't have kubectl
tool in your local machine you can install that using this official documentation.
You can create the AKS cluster by using two ways:
-
A. Use Azure CLI: Follow a procedure using Azure CLI like here, but you need to make sure you of the following items :
-
enable RBAC with
--enable-rbac
-
enable application routing with
--enable-addons http_application_routing
inaz aks create
command.
-
enable RBAC with
-
B. Use Azure's portal
The following steps are using the Azure portal to create the AKS cluster:
-
Start the process by providing the general data, like in the following screenshot:
-
Select default node pool related settings:
-
Then important is to set authentication method and enable RBAC:
-
To Enable http routing make sure to check the checkbox "Http application routing" on "Networking" settings. For more info, read the documentation
You can use basic network settings since for a test you don't need integration into any existing VNET.
Note: In case if you have missed out on setting http routing during the creation of the cluster, you would need to run below command once the cluster is up and running.
az aks enable-addons -a http_application_routing -g <resource-group-name> -n <aks-cluster-name>
-
You can also enable monitoring:
Note: You can choose your preconfigured Azure Container Registry as well but for that in the "Authentication" configuration step you would need to choose an authentication method as "System-assigned managed identity". For e.g :
-
Finally once all validations are passed, create the cluster. It'll take a few minutes for it to be ready.
Once the cluster is up and running, you'll need to configure your local kubectl
tool to access AKS cluster. You can do that by running the following command.
az aks get-credentials --resource-group <resource-group-name> --name <k8s-cluster-name>
Note Check your kubectl
configuration to see if it's pointing to the correct cluster context. You can do that using the below command.
kubectl config get-contexts
Helm is the package manager for Kubernetes.
-
For detailed installation steps, please refer official documentation page.
-
After successful installation,
helm
version can be seen as per below :
Please note, helm 3
onwards Tiller Server component has been removed. So you wouldn't need to install client and server separately.
Ingress is an API object that allows access to your clustered services from the outside. It's like a reverse proxy, that can handle load balancing, TLS, virtual hosting and the like.
NGINX is the Ingress controller used for eShopOnContainers.
To install the NGINX Ingress controller, run the following command:
kubectl apply -f https://raw.githubusercontent.com/kubernetes/ingress-nginx/controller-v1.1.1/deploy/static/provider/cloud/deploy.yaml
Alternatively, you can also install using Helm. Refer to the Quick Start Guide for details.
Navigate to the folder deploy\k8s\nginx-ingress and run the following command:
kubectl apply -f .\local-cm.yaml
The above command will set the proxy-buffer size to 128k
which is required by the identity service.
Go to the deploy/k8s/helm folder in your local copy of the eShopOnContainers repo.
At this point you have two options for installing eShopOnContainers:
- Use the public images from DockerHub (eshop) with tag
linux-dev
-
Or Use your local images with tag
linux-latest
The easiest way to setup eshop on AKS is to use public images. For that you would require to run below scripts:
-
For Windows :
.\deploy-all.ps1 -externalDns aks -aksName <aks-cluster-name> -aksRg <resource-group-name> -imageTag linux-dev -useMesh $false
-
For Mac :
.\deploy-all-mac.ps1 -externalDns aks -aksName <aks-cluster-name> -aksRg <resource-group-name> -imageTag linux-dev -useMesh $false
If you have done some local changes in the eshop application and want to deploy that to the cluster, you can do that by building and publishing your local docker images to your private container registry.
-
Setup a container registry. In this scenario
ACS
has been used.Alternatively, you can use Azure CLI to set up ACS
-
Go to the
src
directory of your local repo and below command :docker-compose build
-
Login to your container registry.
docker login <container-registry-name>.azurecr.io
-
Then tag those images and publish them to the
ACS registry
.docker tag eshop/mobileshoppingagg:linux-latest <acs-registry-name>.azurecr.io/eshop/mobileshoppingagg:linux-lates docker tag eshop/ordering.signalrhub:linux-latest <acs-registry-name>.azurecr.io/eshop/ordering.signalrhub:linux-lates .... ....
docker push <acs-registry-name>.azurecr.io/eshop/mobileshoppingagg:linux-latest docker push <acs-registry-name>.azurecr.io/eshop/ordering.signalrhub:linux-latest .... ....
-
After completion, your container registry should look like below :
-
To use your own images instead of the public ones, you have to pass following additional parameters to the
deploy-all.ps1
script:-
registry
: Login server for the Docker registry -
dockerUser
: User login for the Docker registry -
dockerPassword
: User password for the Docker registry
For e.g :
- Windows :
.\deploy-all.ps1 -externalDns aks -aksName <aks-cluster-name> -aksRg <resource-group> -imageTag linux-latest -registry <container-registry>.azurecr.io -dockerUser <container-registry-user-name> -dockerPassword <container-registry-user-password> -useMesh $false
- Mac :
.\deploy-all-mac.ps1 -externalDns aks -aksName <aks-cluster-name> -aksRg <resource-group> -imageTag linux-latest -registry <container-registry>.azurecr.io -dockerUser <container-registry-user-name> -dockerPassword <container-registry-user-password> -useMesh $false
-
Once the script is run, you should see following output when using kubectl get deployment
:
NAME READY UP-TO-DATE AVAILABLE AGE
eshop-apigwmm 1/1 1 1 29d
eshop-apigwms 1/1 1 1 29d
eshop-apigwwm 1/1 1 1 29d
eshop-apigwws 1/1 1 1 29d
eshop-basket-api 1/1 1 1 30d
eshop-basket-data 1/1 1 1 30d
eshop-catalog-api 1/1 1 1 30d
eshop-identity-api 1/1 1 1 30d
eshop-keystore-data 1/1 1 1 30d
eshop-locations-api 1/1 1 1 30d
eshop-marketing-api 1/1 1 1 30d
eshop-mobileshoppingagg 1/1 1 1 30d
eshop-nosql-data 1/1 1 1 30d
eshop-ordering-api 1/1 1 1 30d
eshop-ordering-backgroundtasks 1/1 1 1 30d
eshop-ordering-signalrhub 1/1 1 1 30d
eshop-payment-api 1/1 1 1 30d
eshop-rabbitmq 1/1 1 1 30d
eshop-sql-data 1/1 1 1 30d
eshop-webhooks-api 1/1 1 1 30d
eshop-webhooks-web 1/1 1 1 30d
eshop-webmvc 1/1 1 1 30d
eshop-webshoppingagg 1/1 1 1 30d
eshop-webspa 1/1 1 1 30d
eshop-webstatus 1/1 1 1 30d
Every public service is exposed through its own ingress resource, as you can see if using kubectl get ing
:
eshop-apigwmm eshop.<your-guid>.<region>.aksapp.io <public-ip> 80 4d
eshop-apigwms eshop.<your-guid>.<region>.aksapp.io <public-ip> 80 4d
eshop-apigwwm eshop.<your-guid>.<region>.aksapp.io <public-ip> 80 4d
eshop-apigwws eshop.<your-guid>.<region>.aksapp.io <public-ip> 80 4d
eshop-identity-api eshop.<your-guid>.<region>.aksapp.io <public-ip> 80 4d
eshop-webhooks-api eshop.<your-guid>.<region>.aksapp.io <public-ip> 80 4d
eshop-webhooks-web eshop.<your-guid>.<region>.aksapp.io <public-ip> 80 4d
eshop-webmvc eshop.<your-guid>.<region>.aksapp.io <public-ip> 80 4d
eshop-webspa eshop.<your-guid>.<region>.aksapp.io <public-ip> 80 4d
eshop-webstatus eshop.<your-guid>.<region>.aksapp.io <public-ip> 80 4d
Note: Ingresses are automatically configured to use the public DNS of the AKS provided by the "https routing" addon.
One step more is needed: we need to configure the nginx ingress controller that AKS has to allow larger headers. This is because the headers sent by the identity server exceed the size configured by default. Fortunately this is very easy to do. Just type (from the /k8s/helm
folder):
kubectl apply -f aks-httpaddon-cfg.yaml
Then you can restart the pod that runs the nginx controller. Its name is addon-http-application-routing-nginx-ingress-controller-<something>
and runs on kube-system
namespace. So run a kubectl get pods -n kube-system
find it and delete with kubectl delete pod <pod-name> -n kube-system
.
Note: If running in a bash shell you can type:
kubectl delete pod $(kubectl get pod -l app=addon-http-application-routing-nginx-ingress -n kube-system -o jsonpath="{.items[0].metadata.name}) -n kube-system
After a while, when all services are running OK, you should get to see something like this:
- WebStatus:
http://[eshop.<your-guid>.<region>.aksapp.io]/webstatus
- WebMVC:
http://[eshop.<your-guid>.<region>.aksapp.io]/webmvc
- WebSPA:
http://[eshop.<your-guid>.<region>.aksapp.io]/
To delete eShop related deployments you can use this command:
helm uninstall $(helm ls --filter eshop -q)
Before performing actual delete if you want to perform a dry run, below command is useful.
helm uninstall $(helm ls --filter eshop -q) --dry-run
In order NOT to get errors in the Kubernetes dashboard, you'll need to set the following service-account steps.
Here you can see the errors you might see:
- Because the cluster is using RBAC, you need to grant needed rights to the Service Account
kubernetes-dashboard
with this kubectl command:
kubectl create clusterrolebinding kubernetes-dashboard -n kube-system --clusterrole=cluster-admin --serviceaccount=kube-system:kubernetes-dashboard
Now, just run the Azure CLI command to browse the Kubernetes Dashboard:
az aks browse --resource-group pro-eshop-aks-helm-linux-resgrp --name pro-eshop-aks-helm-linux
There is the possibility to install eShopOnContainers ready to run with the Linkerd Service Mesh. To use Linkerd, you must complete the following steps:
- Install Linkerd on your cluster. The process is described in the Linkerd installation documentation. Steps 0 trough 3 need to be done.
- Then install eShopOnContainers as per the procedure described above, but using this command:
.\deploy-all.ps1 -externalDns aks -aksName <aks-cluster-name> -aksRg <resource-group-name> -imageTag linux-dev -useMesh $true
Once eShop is installed you can check that all non-infrastructure pods have two containers:
Now you can use the command linkerd dashboard
to show the mesh and monitor all the connections between eShopOnContainer pods.
The mesh monitors all HTTP connections (including gRPC), but don't monitor RabbitMQ or any other connection (SQL, Mongo, ...)
For more information read the Resiliency and Service Mesh page in the wiki.
The -externalDns
parameter controls the DNS bounded to ingresses. You can pass a custom DNS (like my.server.com
), or the aks
value to autodiscover the AKS DNS. For autodiscover to work you also need to pass which AKS is, using the -aksName
and -aksRg
parameters.
Autodiscovering works using Azure CLI under the hood, so ensure that Azure CLI is logged and pointing to the right subscription.
If you don't pass any external DNS at all, ingresses isn't bound to any DNS, and you have to use public IP to access the resources.
If you want to use external resources, use -deployInfrastructure $false
to not deploy infrastructure containers. However you still have to manually update the scripts to provide your own configuration (see next section).
The file inf.yaml
contains the description of the infrastructure used. That file is documented so take a look at it to understand all of its entries. If using external resources you need to edit this file according to your needs. You'll need to edit:
-
inf.sql.host
with the host name of the SQL Server -
inf.sql.common
entries to provide your SQL user, password.Pid
is not used when using external resources (it is used to set specific product id for the SQL Server container). -
inf.sql.catalog
,inf.sql.ordering
,inf.sql.identity
: To provide the database names for catalog, ordering and identity services -
mongo.host
: With the host name of the Mongo DB -
mongo.locations
,mongo.marketing
with the database names for locations and marketing services -
redis.basket.constr
with the connection string to Redis for Basket Service. Note thatredis.basket.svc
is not used when using external services -
redis.keystore.constr
with the connection string to Redis for Keystore Service. Note thatredis.keystore.svc
is not used when using external services -
eventbus.constr
with the connection string to Azure Service Bus andeventbus.useAzure
totrue
to use Azure service bus. Note thateventbus.svc
is not used when using external services
Using Azure storage for catalog (and marketing) photos is not directly supported, but you can accomplish it by editing the file k8s/helm/catalog-api/templates/configmap.yaml
. Search for lines:
catalog__PicBaseUrl: http://{{ $webshoppingapigw }}/api/v1/c/catalog/items/[0]/pic/
And replace it for:
catalog__PicBaseUrl: http://<url-of-the-storage>/
In the same way, to use Azure storage for the marketing service, have to edit the file k8s/helm/marketing-api/templates/configmap.yaml
and replacing the line:
marketing__PicBaseUrl: http://{{ $webshoppingapigw }}/api/v1/c/catalog/items/[0]/pic/
by:
marketing__PicBaseUrl: http://<url-of-the-storage>/
- System requirements
- Development setup
- Databases & containers
- Architecture
- Application
- Code
- Logging and Monitoring
- Tests