Deploy Apache Superset on Azure. Use when deploying Superset for BI/data visualization with PostgreSQL backend.
npx skills add https://github.com/microsoft/agentic-journeys --skill superset-azure
Deploy Apache Superset data visualization platform on Azure Kubernetes Service.
> Complexity Note: Superset is the most complex deployment in this project due to psycopg2 requirements and AKS architecture. Deploy time: ~15-20 minutes.
Require Azure CLI, Azure Developer CLI 1.28.0 or later, and Node.js LTS or later. Verify az version, azd version, and node --version before generating infrastructure. The host must not need kubectl or Helm. Installation options for Windows, Mac, and Linux are in ../../../docs/tool-installation.md.
Generate the AKS post-provision workflow as infra-superset/hooks/postprovision.js and reference it directly from azure.yaml. The hook must attach the Kubernetes manifests and a remote deployment script to az aks command invoke. Run Helm and kubectl inside Azure. Invoke az and azd with argument arrays. On Mac and Linux, call each executable directly. On Windows, use the static PowerShell runner and JSON environment payload defined by the container-apps-deployment skill so Azure CLI .cmd shims aren't launched directly. Reject double quotes for every Windows target and additional shell metacharacters or CR/LF for .cmd/.bat; use attached scripts for complex remote commands. Do not generate a Bash-only host hook.
Start the long deployment command with --no-wait, parse the returned command ID, and poll az aks command result. Require provisioningState to equal Succeeded and exitCode to equal 0. Use a separate short AKS run command to read the ingress IP so URL discovery does not depend on long-command log truncation.
Write generated Kubernetes Secret values to a mode-0600 temporary manifest as base64 data. Attach the temporary bundle to AKS run command, do not print the values, delete the remote Secret manifest immediately after kubectl apply, and remove the local bundle in finally after success or failure.
The hook owns SUPERSET_SECRET_KEY and SUPERSET_ADMIN_PASSWORD. On a clean environment, generate cryptographically random values for either missing setting, persist each with azd env set, never print the values, and reuse existing values on reruns. A first deployment must not depend on undocumented manual secret setup.
This skill provides Superset-specific configuration only. Infrastructure (Bicep, azure.yaml, K8s manifests) should be generated fresh each time by the official azure-prepare → azure-validate → azure-deploy pipeline. Do NOT rely on pre-existing infra code.
ALWAYS set AZURE_SUBSCRIPTION_ID explicitly before running azd up. Read it with az account show --query id -o tsv, then pass the returned value to azd env set AZURE_SUBSCRIPTION_ID <subscription-id>. Do not use Bash command substitution when the host OS is unknown.
Without this, azd and Azure MCP tools will fail silently or produce incomplete deployments.
📖 See ../config/postgresql-avm-defaults.md for all PostgreSQL AVM gotchas (publicNetworkAccess, passwordAuth, HA, password pinning). Without these, Superset will fail with "authentication failed" or "connection timeout".
Superset-specific: Pin POSTGRES_PASSWORD, SUPERSET_SECRET_KEY, and SUPERSET_ADMIN_PASSWORD in the azd environment. Generate them with Node's crypto.randomBytes() or another cryptographically secure platform API. Do not require openssl, which is not installed by default on Windows.
module aksCluster 'br/public:avm/res/container-service/managed-cluster:0.9.0' = {
params: {
disableLocalAccounts: false // Default requires AAD — fails without it
primaryAgentPoolProfiles: [
{ name: 'system', availabilityZones: [] } // westus doesn't support AZ
]
}
}
# 1. Register providers (one-time per subscription)
az provider register --namespace Microsoft.ContainerService
az provider register --namespace Microsoft.DBforPostgreSQL
az provider register --namespace Microsoft.OperationalInsights
# 2. Create environment
azd env new my-superset-env
# 3. Set required variables (replace placeholders with collected/generated values)
azd env set AZURE_SUBSCRIPTION_ID "<subscription-id>"
azd env set AZURE_LOCATION "westus"
azd env set POSTGRES_PASSWORD "<generated-secret>"
azd env set SUPERSET_SECRET_KEY "<generated-secret>"
azd env set SUPERSET_ADMIN_PASSWORD "<generated-secret>"
# 4. Deploy (~15-20 minutes)
azd up
# 5. Access Superset
azd env get-value SUPERSET_URL
# Login: admin / value returned by azd env get-value SUPERSET_ADMIN_PASSWORD
Deployment time breakdown:
| File | Purpose |
|------|---------|
| config/environment-variables.md | All Superset environment variables |
| config/health-probes.md | Health probe timing for Superset startup |
| troubleshooting.md | Common issues and solutions |
Apache Superset is a modern data exploration and visualization platform. It requires:
graph TB
LB["Load Balancer<br/>(Public IP)"]
subgraph AKS["AKS Cluster"]
NGINX["NGINX Ingress Controller"]
SVC["Superset Service<br/>(ClusterIP:80)"]
subgraph POD["Superset Deployment"]
INIT["Init Container (migrate)"]
MAIN["Main Container (web · port 8088)"]
CM["ConfigMap (config.py)"]
VOL["emptyDir (psycopg2)"]
end
end
PG["PostgreSQL Flexible Server<br/>(Azure Managed PaaS)"]
LB --> NGINX --> SVC --> POD
POD --> PG
The official Superset image does NOT include psycopg2 for PostgreSQL. Without it, Superset falls back to SQLite. See references/psycopg2-installation.md for the full solution.
TL;DR: Install to emptyDir volume with --target=/psycopg2-lib, set PYTHONPATH=/psycopg2-lib in both init and main containers.
| Variable | Description | Example |
|----------|-------------|---------|
| SQLALCHEMY_DATABASE_URI | PostgreSQL connection string | postgresql://USER:PASS@HOST:5432/DB?sslmode=require |
| SUPERSET_SECRET_KEY | Flask secret key (required) | 32+ char random string |
| SUPERSET_CONFIG_PATH | Path to config file | /app/pythonpath/superset_config.py |
| PYTHONPATH | Include psycopg2 location | /psycopg2-lib |
See config/environment-variables.md for full details.
Critical: Azure PostgreSQL requires sslmode=require in the connection string.
See references/kubernetes-manifests.md for complete Deployment, ConfigMap, and Ingress patterns.
See config/health-probes.md for liveness, readiness, and startup probe configuration. Key values: /health on port 8088, initialDelaySeconds: 90 for liveness (Superset is slow to start).
| Component | CPU Request | CPU Limit | Memory Request | Memory Limit |
|-----------|-------------|-----------|----------------|--------------|
| Superset Web | 250m | 1000m | 512Mi | 2Gi |
| Init Container | (inherits) | (inherits) | (inherits) | (inherits) |
⚠️ CPU Gotcha: Standard_DS2_v2 (2 vCPU) only has ~500m available after AKS system pods. Set CPU request to 250m (not 500m) or the pod will be stuck in Pending with "Insufficient cpu". CPU limit can stay at 1000m for bursting.
See troubleshooting.md for detailed fixes. Most common: psycopg2 import errors (install to /psycopg2-lib with PYTHONPATH), SQLite fallback (check superset_config.py ConfigMap), and SSL connection errors (add ?sslmode=require).
Use these Azure MCP Server tools for Superset deployments:
| Tool | When to Use |
|------|-------------|
| azure_deploy_plan | Generate a deployment plan — use params: target=AKS, provisioning_tool=AZD |
| azure_bicep_schema | Get latest schemas for Microsoft.ContainerService/managedClusters and Microsoft.DBforPostgreSQL/flexibleServers |
| azure_deploy_iac_guidance | AKS-specific Bicep best practices — use resource_type=aks |
| azure_deploy_app_logs | Fetch Log Analytics logs post-deployment to troubleshoot pod CrashLoopBackOff or init failures |
| azure_deploy_architecture | Generate Mermaid architecture diagrams for the Superset AKS deployment |
Before verifying: ensure PostgreSQL has a firewall rule, AKS run command is available, remote Helm installed NGINX Ingress, ConfigMap has superset_config.py, K8s secret has SQLALCHEMY_DATABASE_URI + SUPERSET_SECRET_KEY + ADMIN_PASSWORD, and PYTHONPATH=/psycopg2-lib is set. See troubleshooting.md for the full verification checklist.
For testing only (change in production):
adminADMIN_PASSWORD env var| Resource | Monthly Cost |
|----------|--------------|
| AKS Cluster (2x Standard_D2s_v3) | ~$100-150 |
| PostgreSQL Flexible Server (B1ms) | ~$15 |
| Load Balancer | ~$20 |
| Total | ~$135-185/month |
Note: Superset on AKS is more expensive than Container Apps deployments (n8n, Grafana). Consider Container Apps if AKS features aren't required.
azd down --force --purge
Note: Teardown takes 5-10 minutes (AKS + PostgreSQL deletion is slow).
After deployment completes, run the checked-in verifier from journeys/superset:
node ../../.github/scripts/verify-superset.mjs
The verifier must use az aks command invoke for pod status, logs, and in-pod checks. It must not invoke a local kubectl or Helm binary.
For automated browser login, use #username, #password, and the resilient submit selector input[type="submit"], button[type="submit"]. Superset 4.1.1 renders a Flask-AppBuilder submit input; other versions may render a button. Verify successful navigation to /superset/welcome/.
Take microsoft/superset-azure from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.