Google Cloud provider¶
The Google Cloud provider lets ORB acquire, track, and release compute capacity backed by Google Compute Engine, either as a Managed Instance Group (MIG) or as standalone VM instances. It reuses the same template, request, and machine model as the AWS, Kubernetes, and Azure providers, so callers of the CLI, REST API, SDK, and HostFactory plugin do not need to special-case Google Cloud.
The provider supports two provisioning shapes (provider_api values):
provider_api |
Compute Engine resource(s) | Start/stop | Max instances |
|---|---|---|---|
MIG |
Instance template + Managed Instance Group (regional or zonal) | no | 1000 |
SingleVM |
A single standalone VM instance | yes | 1 |
See Handlers for how to pick between them.
Install¶
The Google Cloud provider lives behind an optional install extra so operators who only target AWS, Kubernetes, or Azure do not pay for the Google Cloud SDKs.
This pulls in google-cloud-compute and google-auth.
Quick start¶
1. Authenticate¶
The Google Cloud provider authenticates exclusively with Application Default Credentials (ADC) — there is no separate ORB-level credential configuration. See Authentication for the supported ADC sources and troubleshooting.
2. Configure the provider¶
{
"providers": [
{
"name": "gcp_main",
"type": "gcp",
"config": {
"project_id": "my-gcp-project",
"region": "us-central1"
}
}
]
}
The full set of fields lives in Configuration reference.
3. Create a template¶
This emits a template targeting the MIG handler. Set machine_type,
project_id, region/zones, and a source image, then save it. See
Configuration reference for the full
field list and Handlers for per-handler requirements.
4. Request capacity¶
5. Track and release¶
Releases delete the named standalone VM(s) for SingleVM. For MIG,
releasing every member of a group deletes the Managed Instance Group and
its backing instance template; releasing a subset deletes only those
managed instances.
AWS concepts mapped to Google Cloud¶
| AWS concept | Google Cloud equivalent |
|---|---|
| EC2 instance | Compute Engine instance |
| Auto Scaling Group / EC2 Fleet | Managed Instance Group (MIG), regional or zonal |
| Amazon Machine Image (AMI) | source_image, or source_image_family + source_image_project |
| Instance type | machine_type, e.g. e2-standard-4 |
| Spot / On-Demand | price_type ("ondemand" or "spot") |
| Availability Zone | Google Cloud zone, e.g. us-central1-a (zones field) |
| EC2 key pair | Not supported — Google Cloud does not use named SSH key pairs (key_name is rejected). |
| VPC subnet | network / subnetwork |
| IAM instance profile | Attached service account (service_account_email, service_account_scopes) |
| EC2 user data | Not modelled by the current template; use image-baked startup scripts. |
Key differences to keep in mind:
- Google Cloud has exactly one credential identity per ORB process: Application Default Credentials. There is no equivalent to AWS multi-profile or Azure managed-identity-selection — if you need to target more than one Google Cloud project or account, run separate ORB deployments with distinct ADC sources.
- Start/stop is only supported for
SingleVM.MIG-managed instances follow group policy; ORB'sstart/stopoperations against a MIG return a warning and perform no action. See Provider capabilities. - Google Cloud templates reject
key_name(named SSH key pairs) outright — Google Cloud does not have that concept at the Compute Engine API level.
What is in this section¶
- Configuration reference - every
GCPProviderConfigandGCPTemplatefield, with worked examples. - Handlers - MIG and SingleVM; when to pick each.
- Authentication - Application Default Credentials and troubleshooting.