Providers and regions
Production exposure, implemented provider support, topology, storage, and multi-region boundaries.
Provider availability is resolved at runtime. Active database rows in the catalog are necessary but not sufficient: Parix also requires an enabled provider, a configured gateway region, regional pricing, and a purchasable size/topology/storage combination.
The dashboard create form and GET /api/v1/catalog/create are authoritative for the environment you are using.
Current production exposure
The checked-in prd Worker configuration currently exposes:
| Provider | Create regions | Production state |
|---|---|---|
| Google Cloud | asia-southeast1 (Singapore), us-central1 (Iowa) | Enabled |
| AWS | None | AWS_PROVIDER_ENABLED is false in prd |
AWS is implemented and enabled in local/development configuration for ap-southeast-1 and us-east-1. Its presence in code, seed data, gateway maps, or development does not make it selectable in production.
The seed also contains GCP us-east1 and europe-west1, plus inactive AWS region rows. They are filtered out when the environment lacks an enabled gateway route or provider switch.
Topology by plan
| Plan | Customer topology choice |
|---|---|
| Developer | None. Parix places the project on a compatible shared cell. |
| Dedicated Single Node | 1 replica (single). |
| Production HA | 3 replicas (ha). |
| Production 6 | 6 replicas (ha-6). |
| Enterprise | Contract-defined and not self-service. |
The current GCP and implemented AWS provisioners accept 1, 3, and 6 nodes. The create UI maps those node counts to the dedicated plan codes above.
Storage support
Storage availability is a provider, region, topology, and cluster-size intersection. Always use the options returned by the current catalog.
Google Cloud in production
| Catalog tier | Current boundary |
|---|---|
SSD | Provider backend is persistent disk. The generic configuration range is 10–4,096 GB per replica, subject to a priced catalog combination. |
NVMe | Local NVMe is enabled only for 3- and 6-replica topologies, PX-10 or PX-20, with fixed 375 or 750 GB choices. |
Hyperdisk Premium | Implemented in catalog/pricing code but disabled in the current production configuration. |
For persistent disk, post-create storage changes can only increase capacity. GCP topology and storage changes that require a new shape use a fresh backup and restore into the target cluster during controlled downtime.
Local NVMe is host-local. Post-create storage tier, storage size, topology, and cluster-size changes are blocked in v1.
AWS implementation boundary
When AWS is enabled in a non-production environment:
| Catalog tier | Implemented boundary |
|---|---|
SSD | Provider backend is EBS. The generic configuration range is 10–4,096 GB per replica, subject to catalog pricing. |
NVMe | Local NVMe is limited to 3- and 6-replica topologies. Fixed options vary by PX size from 10 GB through 400 GB. |
AWS topology and EBS storage changes are currently blocked until the AWS restore-backed migration path supports them. AWS local NVMe also blocks post-create storage, topology, and cluster-size changes in v1.
Cluster sizes
The active seeded dedicated sizes are PX-10, PX-20, PX-40, and PX-80. PX-5 and PX-160 remain inactive in the create catalog. A size appears only when an active provider/topology variant and regional price can be resolved.
Do not copy a size or price from seed SQL into automation. Resolve the catalog when constructing the request.
Cross-region HA beta
Cross-region HA is implemented behind account-level feature and configuration flags. It is not a default production capability.
When enabled, the current selection rules require:
- a 3- or 6-replica Production configuration
- 2–3 total regions, including the primary
- all regions on the same provider
- every region to be present in the create catalog and feature allowlist
- the selected storage tier to be available in every chosen region
- no more regions than replicas
GCP has the automated isolated multi-region path. The AWS implementation uses shared networking and is unavailable in production while the AWS provider switch is off.
Do not describe a single-region Production database as cross-region HA. The ordinary 3- and 6-replica catalog configurations are single-cluster topologies unless an enabled create request includes additional regions.
Query the runtime catalog
curl "https://parix.io/api/v1/catalog/create" \
-H "x-api-key: $PARIX_API_KEY"The credential must be organization-wide and include db:write. If a provider or region is absent, treat it as unavailable in that environment even if another environment or a checked-in seed contains it.