This document describes the setup of Azure Kubernetes Service (AKS) and Wolfram Application Server (WAS).
The following CLI tools are required to be installed on your local machine to complete the setup and installation:
-
Azure CLI - https://docs.microsoft.com/en-us/cli/azure/install-azure-cli?view=azure-cli-latest
-
Kubectl >=v1.27 - https://kubernetes.io/docs/tasks/tools/install-kubectl/
-
Docker v20.10 or newer - https://docs.docker.com/get-docker/
-
Docker Compose v1.28.6 or newer - https://docs.docker.com/compose/install/
- Cluster Name: WAS
- Region: eastus
- AMI Instance Type: Standard_D8s_v3
- Disk Size: 30GB
- Node Group scaling configuration: [Minimum size: 2, Maximum size: 10, Desired size: 2]
- Kubernetes Version: 1.27
To change any of the above defaults open Source/terraform/variables.tf and modify accordingly and save file.
Prerequisite: Make sure your Azure subscription has a minimum total region CPU value of 40 or higher to support the default configuration. See the following for more information on managing your Azure quotas: https://docs.microsoft.com/en-us/azure/azure-portal/supportability/regional-quota-requests.
Once your total region CPU of 40 or higher is confirmed available, authenticate with Azure:
az login --tenant <your-tenant-id>
Note: Your Tenant ID can be fetched from https://portal.azure.com/#blade/Microsoft_AAD_IAM/ActiveDirectoryMenuBlade/Overview
Step 1. Checkout the repository:
git clone https://github.com/WolframResearch/WAS-Kubernetes.git
Step 2. Change directory to Azure:
cd WAS-Kubernetes/EnvironmentSetup/Azure/
Step 3. Create resource group and store name
az group create -l <LOCATION> -n <RESOURCE_GROUP_NAME>
LOCATION: The region of the resource group will use, eastus
RESOURCE_GROUP_NAME: The name of the resource group, WAS-rg
Step 4. Create storage account and store account key
az storage account create --resource-group <RESOURCE_GROUP_NAME> --name <SAN> --sku Standard_LRS --encryption-services blob
**SAN**: The storage account name, there shouldn't be any space and characters between letters
az storage account keys list --resource-group <RESOURCE_GROUP_NAME> --account-name <SAN> --query '[0].value' -o tsv
The command will return the **SAN_ACCOUNT_KEY**
Step 5. Update the containers files with proper values
RESOURCE_GROUP_NAME:<resource group name, WAS-rg>
REGION:<location of the resource group, eastus>
SAN:<storage account name that created in step 3, azurewasa3280>
SAN_ACCOUNT_KEY:<storage account key that created in step3, dGVzdDEyMyMkwr3CvntbXX0uIQo=>
RESOURCEINFO_BUCKET:<blob name that WAS will create to store resources, was-resources-3280, name can be like given format>
NODEFILEINFO_BUCKET:<blob name that WAS will create to store nodefiles, was-nodefiles-3280, name can be like given format>
Step 6. Replace the tag with your tenant ID and run the following command to set up AKS and deploy WAS:
YOURTENANTID=<your-tenant-id> && rm -rf Source/terraform/tenant-id.config && echo $YOURTENANTID >> Source/terraform/tenant-id.config && mkdir -p ~/.kube && docker-compose up --build -d && clear && docker exec -it azure-setup-manager bash setup --create && sudo chown -R $USER ~/.kube
Example:
YOURTENANTID=QQQQQQQ-9810-46e1-bef2-hswqw56sf && rm -rf Source/terraform/tenant-id.config && echo $YOURTENANTID >> Source/terraform/tenant-id.config && mkdir -p ~/.kube && docker-compose up --build -d && clear && docker exec -it azure-setup-manager bash setup --create && sudo chown -R $USER ~/.kube
Note: This can take approximately 25 minutes to complete.
Step 7. Run the following command to retrieve your base URL and application URLs:
docker-compose up --build -d && clear && docker exec -it azure-setup-manager bash setup --endpoint-info
The output of this command will follow this pattern:
Base URL - Active Web Elements Server: http://<your-base-url>/
Resource Manager: http://<your-base-url>/resources/
Endpoints Manager: http://<your-base-url>/endpoints/
Nodefiles: http://<your-base-url>/nodefiles/
Endpoints Info: http://<your-base-url>/.applicationserver/info
Restart AWES: http://<your-base-url>/.applicationserver/kernel/restart
Step 8. After completion, run this command to shutdown the azure-setup-manager:
docker-compose down
Step 9. Get a license file from your Wolfram Research sales representative.
Step 10. This file needs to be deployed to WAS as a node file in the conventional location .Wolfram/Licensing/mathpass. From a Wolfram Language client, this may be achieved using the following code:
was = ServiceConnect["WolframApplicationServer", "http://<your-base-url>"];
ServiceExecute[was, "DeployNodeFile",
{"Contents"-> File["/path/to/mathpass"], "NodeFile" -> ".Wolfram/Licensing/mathpass"}]
Alternatively you may use the node files REST API to install the license file.
Note: In order to use the Wolfram Language functions, the WolframApplicationServer paclet must be installed and loaded. Run the following code:
PacletInstall["WolframApplicationServer"];
Needs["WolframApplicationServer`"]
Step 11. Restart the application using the restart API to enable your Wolfram Engines.
URL: http://<your-base-url>/.applicationserver/kernel/restart
The default credentials for this API are:
Username: applicationserver
Password: P7g[/Y8v?KR}#YvN
To change these, see the configuration documentation.
Note: Active Web Elements Server will restart and activate using the mathpass. Upon successful activation, the application shall start.
Your setup is now complete.
The following completely deletes everything including the kubernetes cluster, Wolfram Application Server and all resources:
Step 1. Change your directory to the directory containing docker-compose.yml directory and run the following command to destroy your AKS cluster and WAS:
docker-compose up --build -d && clear && docker exec -it azure-setup-manager bash setup --delete
Warning: All data will be destroyed.
Step 2. After completion, shutdown the azure-setup-manager by running the following command:
docker-compose down -v
When you run azure-setup-manager, you may encounter this error.
Creating azure-setup-manager ...
ERROR: for azure-setup-manager UnixHTTPConnectionPool(host='localhost', port=None): Read timed out. (read timeout=60)
You need to export these to fix that issue.
export DOCKER_CLIENT_TIMEOUT=120
export COMPOSE_HTTP_TIMEOUT=120