BYOCloud cloud resources

​

Goal

​

Prerequisites

• A provisioned BYOCloud node
• A provider that supports the resource type you want to use

​

Workflow

​

Current provider support

• AWS: EBS volumes, EBS volume snapshots, VPC/subnet/AZ metadata, Elastic IPs
• DigitalOcean: block storage volumes, volume snapshots, region/network metadata, Reserved IPs
• Hetzner: cloud volumes, location metadata, Floating IPs; attached-volume snapshots are intentionally unsupported
• Azure: managed disks, disk snapshots, resource-group/VNet/subnet metadata, static Public IPs

​

Node-scoped inventory and refresh

• Inventory records provider resource ID, region or placement, provider state, StackShift ownership, safe-to-delete state, last synchronization, and operation state.
• Refresh queries only the selected node and never triggers global reconciliation.
• Use refresh after a provider incident, quota change, or emergency provider-console action.

stackshift byoc refresh NODE_UUID -o json
stackshift byoc resources NODE_UUID -o json

​

Volumes

• Create after the node is active with a name, positive size, mount path, and filesystem; the default filesystem is ext4.
• Hetzner volumes must be at least 10 GB.
• Detach preserves the provider volume and its snapshots.
• Delete is refused while dependent snapshots exist or the volume is not safe to delete.

stackshift byoc volume-create NODE_UUID --data '{
  "name": "production-data",
  "size_gb": 80,
  "mount_path": "/var/lib/stackshift-data",
  "filesystem": "ext4"
}' -o json

stackshift byoc volume-detach VOLUME_UUID -o json
stackshift byoc volume-delete VOLUME_UUID --yes -o json

​

Snapshots

• DigitalOcean, AWS, and Azure support volume snapshots; Hetzner attached-volume snapshots are not advertised.
• Wait for available before relying on a snapshot for recovery.
• Deleting a volume never silently deletes dependent snapshots.

stackshift byoc snapshot-create VOLUME_UUID --data '{"name":"before-schema-migration"}' -o json
stackshift byoc snapshot-delete SNAPSHOT_UUID --yes -o json

​

Stable public IP and reboot

• Allocate one supported provider-native stable address and inspect provider ID, address, assignment, and last sync.
• A stable workload IP never exposes agent TCP 8443; agent access remains WireGuard-only.
• Reboot makes the node unschedulable until overlay, certificate, heartbeat, runtime, Node Guard, and Caddy recover.

stackshift byoc static-ip NODE_UUID -o json
stackshift byoc static-ip-release NODE_UUID --yes -o json
stackshift byoc reboot NODE_UUID --yes -o json

​

Safe deletion saga

• Deletion returns 409 with structured blockers for projects, stacks, databases, volumes, and snapshots.
• After blockers are cleared, StackShift removes add-ons and networks in dependency order.
• The provider instance must be confirmed absent before identity revocation and node soft deletion.
• Provider failures remain persisted and retryable; resources are never marked deleted optimistically.

stackshift byoc node-delete NODE_UUID --yes -o json

​

Terraform contract

• Resources: stackshift_byoc_provider_connection, stackshift_byoc_node, stackshift_byoc_volume, stackshift_byoc_snapshot, and stackshift_byoc_static_ip.
• Terraform derives mutation idempotency keys from immutable resource identity and persists them when omitted, so process recovery repeats the same provider request.
• Node, volume, snapshot, and stable-IP resources wait for provider-confirmed terminal state.
• Import volumes and snapshots with NODE_ID:RESOURCE_ID.

​

Expected result

​

Common failures

​

Related guides