Deployments and Requirements

Pyplan offers several deployment options:

• Pyplan Enterprise SaaS — Pyplan Cloud offers a world-class analytics and planning platform without the complexities of installing and managing your own deployment. It shares security responsibilities with an industry-leading cloud infrastructure provider and valued partner, Amazon Web Services. For more details refer to Pyplan Cloud Architecture.
• On customer cloud — Pyplan can be deployed in the customer's cloud using the managed Kubernetes service of each provider:
  • AWS (Amazon EKS)
  • Azure (Azure Kubernetes Service)
  • Google Cloud (Google Kubernetes Engine)
  • Oracle (OCI Kubernetes Engine)

Building blocks on each cloud

The building blocks are the same everywhere; what changes is the cloud service that provides each one.

Building block	AWS	Azure	Google Cloud
Kubernetes cluster	Amazon EKS	Azure Kubernetes Service (AKS)	Google Kubernetes Engine (GKE)
Infrastructure as code	CloudFormation	Terraform	Terraform
Ingress	Application Load Balancer + AWS WAF	Gateway API (AKS application routing)	Gateway API (GKE Gateway controller)
TLS certificate	AWS Certificate Manager	Azure Key Vault	Certificate Manager
PostgreSQL 16	Amazon RDS	Azure Database for PostgreSQL Flexible Server	In the cluster, or Cloud SQL
Shared file storage (/data)	Amazon EFS	Azure Files Premium	Filestore
Secrets	AWS Secrets Manager	Azure Key Vault	Secret Manager
Container images	Amazon ECR	Azure Container Registry	Docker Hub or Artifact Registry

Common requirements

These apply to every cloud. Each provider's page lists what it needs on top of them.

• Kubernetes: a private cluster on a supported version — K8S API supported: v1.34, v1.35, v1.36.
• Two kinds of nodes, each with its own label:
  • Application nodes (env: Pyplan) for the API, UI, WebSocket, Celery workers, the AI component (MCP), Redis and Argo CD.
  • Engine nodes for the calculation engines. Each connected user gets an engine pod that is removed when the session ends, so this node pool scales with concurrent users, not licensed users, and can scale to zero.
• Pyplan resource namespace (pyplan by default).
• PostgreSQL 16 with max_connections set to 1000, reachable only from the cluster.
• Shared file system (ReadWriteMany) mounted at /data: user models and the daily database dumps. It must be backed up daily.
• Redis for the user sessions. It runs in the cluster with persistence on its own disk, so restarting it does not log users out.
• A DNS name with a valid TLS certificate. HTTP is redirected to HTTPS.
• Outbound Internet access to pull or import the container images during installation and updates, and to the AI provider if the Pyplan AI features are enabled.
• Administrative access through a CI/CD pipeline with workload identity federation (recommended). No long-lived cloud keys are stored in the pipeline.
• Argo CD for continuous deployment (recommended): it keeps the cluster in sync with a Git branch, so updating or rolling back Pyplan is a commit.
• Container images: pyplan_ui, pyplan_api, pyplan_engine (Arm or x86 builds), redis:alpine, busybox and bitnami/kubectl for maintenance tasks, and pyplan_ai if the AI integration is enabled.