Talk with our engineers by joining the new Onehouse community Slack!
Deploying Onehouse
Deploy Onehouse on Google Cloud
On Google Cloud, Onehouse runs inside your own Google Cloud project and can provision the Kubernetes cluster for you. You provide the VPC, the Cloud Storage buckets, and the service accounts Onehouse uses; Onehouse creates the GKE cluster and its node pools, installs the data plane, and operates it from then on. Your data stays in your GCS buckets, and compute is billed to your Google Cloud account at your own contract rates.
Projects, the control plane, and connectivity work the same way on every deployment path; see How Onehouse deploys.
What you provide
| Requirement | Detail |
|---|---|
| VPC | Node subnets with outbound access on port 443 through Cloud NAT or equivalent |
| GCS buckets | A platform bucket for platform metadata and job artifacts, and one or more lakehouse buckets in the cluster’s region for your tables |
| Storage access | Private access to Cloud Storage from the node subnet, so data traffic does not pass through NAT |
| Source reachability | VPC peering if your Kafka brokers or databases live in another network |
The full list, including address space and outbound access, is in Environment requirements.
What Onehouse provisions
- The Kubernetes cluster and node pools. Onehouse creates and manages the GKE cluster and the node pools for each workload type. Do not change node counts on these pools by hand; Onehouse manages capacity through pod scheduling.
- Instance types. The
oh-generalfamily maps toe2-standardmachines. Project admins can add custom instance types. See Instance types. - Attached storage. Block storage for shuffle and disk spill is provisioned automatically and sized from the instance.
- Identities. Service accounts with scoped custom roles: one the Onehouse control plane uses to manage the cluster, and one workloads use to read and write your buckets. Individual clusters can run as their own identity to scope storage access per team.
Google Cloud native integrations
| Area | What works on Google Cloud |
|---|---|
| Sources | Google Cloud Storage with event notifications, Kafka and Confluent, and database CDC through any Kafka-compatible endpoint or LakeLog |
| Catalogs | Dataproc Metastore and BigQuery / BigLake, plus Hive Metastore, DataHub, Databricks Unity Catalog, and the Snowflake catalogs |
| Schema registry | Google Managed schema registry, alongside Confluent and file-based registries |
| Open engines | Trino and Flink clusters run against a Dataproc Metastore catalog. See Open engines. |
| Secrets | Bring-your-own-secrets: credentials stay in Google Cloud Secret Manager in your project, and Onehouse reads only the secrets you opt in by name prefix |
Onboarding
- Start the setup wizard in the Onehouse console, under Connections > Cloud Accounts. It gives you the request ID the customer stack needs.
- Prepare the VPC and buckets described above.
- Run the customer stack. A Terraform module creates the service accounts and their roles.
- Onehouse provisions the cluster. The Onehouse install wizard completes the link: Onehouse creates the GKE cluster and node pools in your project, installs the data plane, and the agent connects.
From then on, Onehouse installs, upgrades, and operates the platform remotely. See Operations and management.