Add Azure and GCP Terraform support with 32 GB RAM defaults across all clouds.

[PondiB]

No description provided.

[Copilot]

Pull request overview

This PR introduces first-class, multi-cloud Terraform stacks for provisioning an OpenEOcraft host on AWS, Azure, and Google Cloud with defaults targeting ≥32 GB RAM and a consistent bootstrap flow via a shared startup script.

Changes:

• Added new Terraform stacks + documentation for Azure and GCP.
• Standardized bootstrap via terraform/shared/cloud_init.tftpl across clouds, and updated AWS defaults to 32 GB RAM class instances.
• Updated top-level docs and .gitignore to reflect the new multi-cloud Terraform layout.

Reviewed changes

Copilot reviewed 25 out of 28 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
terraform/shared/cloud_init.tftpl	Shared bootstrap script for Docker/NVIDIA/workspace mount across clouds
terraform/README.md	New multi-cloud Terraform entrypoint documentation
terraform/gcp/versions.tf	Adds Terraform + Google provider constraints
terraform/gcp/variables.tf	GCP stack configuration knobs (RAM/GPU defaults, networking CIDRs, etc.)
terraform/gcp/main.tf	Provisions GCE instance, firewall rules, optional persistent disk, startup script
terraform/gcp/outputs.tf	Exposes instance identifiers, public IP, API URL, SSH command
terraform/gcp/terraform.tfvars.example	Example default variables for GCP
terraform/gcp/terraform.gpu.tfvars.example	Example GPU variables for GCP
terraform/gcp/terraform.cpu.tfvars.example	Example CPU-only fallback variables for GCP
terraform/gcp/README.md	GCP-specific usage and verification documentation
terraform/gcp/.terraform.lock.hcl	Provider lockfile for GCP stack
terraform/azure/versions.tf	Adds Terraform + AzureRM provider constraints
terraform/azure/variables.tf	Azure stack configuration knobs (RAM/GPU defaults, networking CIDRs, etc.)
terraform/azure/main.tf	Provisions RG/VNet/NSG/VM, optional managed disk, cloud-init
terraform/azure/outputs.tf	Exposes VM identifiers, public IP, API URL, SSH command
terraform/azure/terraform.tfvars.example	Example default variables for Azure
terraform/azure/terraform.gpu.tfvars.example	Example GPU variables for Azure
terraform/azure/terraform.cpu.tfvars.example	Example CPU-only fallback variables for Azure
terraform/azure/README.md	Azure-specific usage and verification documentation
terraform/azure/.terraform.lock.hcl	Provider lockfile for Azure stack
terraform/aws/variables.tf	Updates AWS defaults to 32 GB class instance + 30 GB container memory
terraform/aws/terraform.tfvars.example	Updates AWS example vars to match new defaults
terraform/aws/terraform.gpu.tfvars.example	Updates AWS GPU example vars to match new defaults
terraform/aws/terraform.cpu.tfvars.example	Updates AWS CPU fallback to 32 GB RAM minimum
terraform/aws/README.md	Updates AWS docs to reflect new defaults and quota guidance
terraform/aws/main.tf	Switches AWS user_data to shared cloud_init.tftpl
README.md	Points users to the new multi-cloud Terraform docs
.gitignore	Ignores Azure/GCP terraform.tfvars like AWS already did

Files not reviewed (2)

• terraform/azure/.terraform.lock.hcl: Language not supported
• terraform/gcp/.terraform.lock.hcl: Language not supported

Comments suppressed due to low confidence (1)

terraform/shared/cloud_init.tftpl:53

• The workspace mount logic is unsafe in a multi-cloud script: after DEVICE="${workspace_device}", the loop falls back to picking the first non-root disk via lsblk (see later in this block). On Azure/GCP the first disk is often the boot disk (e.g. /dev/sda), so the script can mount (or even format, if blkid returns nothing for the whole disk) the root disk as the workspace, risking data loss/unbootable VMs. On AWS Nitro it can also pick instance-store NVMe disks instead of the intended EBS volume.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.