Hello community, We are currently attempting to provision downstream RKE2 clusters via the native OpenStack Node Driver in Rancher Provisioning v2. Our private cloud environment utilizes infrastructure flavors with zero local root disk, which mandates that all instances must be created as volume-backed servers. While the driver authenticates and maps flavors and network assets correctly, it returns a generic `Internal Server Error` during the `Creating OpenStack instance…` phase. Because the native node driver config options do not appear to offer a way to specify a bootable block volume allocation during the initial compute creation request, the endpoint rejects the deployment.
# Actual Log Output The driver successfully executes authentication, resolves resource mapping IDs, and handles keypairs, but experiences a terminal failure when issuing the raw server block creation request: ```text Creating machine… (node-openstack-pool) Authenticating… map[ApplicationCredentialId: ApplicationCredentialName: AuthUrl:https://<YOUR_CLEANED_OPENSTACK_ENDPOINT>:5000/v3 CaCert: DomainId: DomainName:Default Insecure:true TenantDomainID: TenantDomainName:Default TenantID:<YOUR_SCRUBBED_TENANT_ID> TenantName:<YOUR_SCRUBBED_PROJECT_NAME> UserDomainID: UserDomainName:Default UserId: Username:<YOUR_SCRUBBED_USERNAME>] (node-openstack-pool) Found flavor id using its name map[ID:<YOUR_SCRUBBED_FLAVOR_ID> Name:<YOUR_SCRUBBED_FLAVOR_NAME>] (node-openstack-pool) Found floating IP pool id using its name map[ID:<YOUR_SCRUBBED_POOL_ID> Name:<YOUR_SCRUBBED_POOL_NAME>] (node-openstack-pool) Loading Key Pair <YOUR_SCRUBBED_KEYPAIR_NAME> (node-openstack-pool) Loading Private Key from /path/to/machine/files/id_rsa (node-openstack-pool) Creating OpenStack instance… map[FlavorId:<YOUR_SCRUBBED_FLAVOR_ID> ImageId:<YOUR_SCRUBBED_IMAGE_ID>] (node-openstack-pool) Creating machine… error creating machine: error in driver during machine creation: Internal Server Error failureReason: CreateError
Expected Behavior
The node driver should provide native backend schema keys (such as boot_from_volume, volume_size, and volume_type) and expose corresponding input fields in the Cluster Management UI layout. This would allow the initial server creation mapping payload to request a volume-backed server initialization sequence, enabling compatibility with zero-disk enterprise private cloud flavors.
Environment Details
-
Rancher Engine: Modern Provisioning v2 Architecture
-
Downstream Kubernetes Version: v1.30.2+rke2r1 / v1.35.4+rke2r1
-
Infrastructure Context: Private Cloud endpoint enforcing mandatory block-storage boot volumes.
Has anyone successfully tackled this behavior or developed a configuration workaround to pass volume backing instructions through the machine driver wrapper without reverting entirely to external Terraform workflows? Thank you for any insight!