PostgreSQL

Managed PostgreSQL 15, 16, and 17 with automated backups, point-in-time recovery, and zero-config internal connectivity.

Creating a PostgreSQL instance

From + Service → Databases, add PostgreSQL. Pick version (15, 16, or 17), disk size, and plan. It appears on the canvas in seconds.

Open another service, go to Variables, and attach the database. DATABASE_URL is injected as a secret. You can also wire it in stackblaze.yaml with fromDatabase.

Connection strings

StackBlaze exposes two connection strings per database: an internal one for services in the same project, and an external one for local tooling like Postico or TablePlus.

Internal (recommended)

Services in the same environment talk over the private network using the database service name. Traffic stays inside the environment.

Connection string (internal)

postgresql://user:password@[service-name]:5432/dbname

External (for local tools)

The external connection string is available in the dashboard under Settings → Connection. It uses SSL and routes traffic through our secure proxy. Use this for GUI clients like TablePlus, Postico, or DBeaver, not for production app connections.

Connection string (external, SSL required)

postgresql://user:password@db.stackblaze.cloud:5432/dbname?sslmode=require

Running migrations

The safest pattern is to run migrations as a pre-deploy command so they complete before new traffic reaches your updated service. Configure this under Settings → Deploy → Pre-deploy command.

Prisma

Pre-deploy command

npx prisma migrate deploy

Node.js / npm script

Pre-deploy command

npm run migrate

Django

Pre-deploy command

python manage.py migrate

One-off from CI

Queue a build after your own migrate job, or set a pre-deploy command on the app so StackBlaze runs it with the same env as the workload (including DATABASE_URL). There is no stackblaze run CLI.

Tip

Always run migrations before the new version of your code starts serving traffic. StackBlaze’s pre-deploy command runs inside an ephemeral container with the same environment variables as your service, including DATABASE_URL.

Point-in-time recovery (PITR)

On Pro and Enterprise plans, StackBlaze continuously archives WAL (Write-Ahead Log) segments to object storage alongside periodic base backups. This lets you restore your database to any moment within your plan’s retention window, not just the last daily snapshot.

To initiate a PITR restore, navigate to your database in the dashboard, click Backups → Restore to point in time, select a timestamp, and choose whether to restore in-place or to a new instance.

Plan	PITR	Retention window
Free	No	—
Starter	No	3 days (snapshots)
Pro	Yes	7 days
Enterprise	Yes	30 days

Automated backups

All plans include automated daily snapshots taken during off-peak hours (typically 02:00–04:00 UTC). Backups are encrypted at rest and stored separately from your primary data.

To restore from a snapshot, go to Backups in the database dashboard, select the snapshot, and click Restore. You can restore to a new instance without affecting the running database.

Supported extensions

The following popular extensions are available on all plans and can be enabled without any configuration:

Extension	Use case
postgis	Geospatial data and queries
pgvector	Vector similarity search for AI/ML workloads
pg_trgm	Fuzzy text search using trigrams
uuid-ossp	UUID generation functions
pg_stat_statements	Query performance tracking
btree_gin	GIN indexes for B-tree operators
hstore	Key-value store within a column
citext	Case-insensitive text type

Enable an extension

CREATE EXTENSION IF NOT EXISTS pgvector;
CREATE EXTENSION IF NOT EXISTS postgis;

Scaling

Vertical scaling

To increase CPU or memory, upgrade the database plan from the dashboard under Settings → Plan. StackBlaze performs a rolling StatefulSet update with a brief connection interruption (typically under 30 seconds). Schedule vertical scaling during off-peak hours.

Read replicas

Read replicas are available on Enterprise plans. They use PostgreSQL streaming replication and are typically under 1 second behind the primary. Connect read-heavy workloads (analytics queries, reporting) to the replica connection string to reduce load on the primary.

Under the hood

PostgreSQL runs as a Kubernetes StatefulSet with a dedicated PersistentVolumeClaim backed by SSD storage. A ClusterIP Service provides a stable internal DNS name ([service-name]) that survives pod restarts and rescheduling. Connection pooling via PgBouncer is available as an add-on on Pro and Enterprise plans.

Connecting with common ORMs

Prisma

prisma/schema.prisma

datasource db {
  provider = "postgresql"
  url      = env("DATABASE_URL")
}

Drizzle ORM

db.ts

import { drizzle } from 'drizzle-orm/postgres-js'
import postgres from 'postgres'

const client = postgres(process.env.DATABASE_URL!)
export const db = drizzle(client)

SQLAlchemy (Python)

database.py

from sqlalchemy import create_engine

engine = create_engine(
    os.environ["DATABASE_URL"],
    pool_size=5,
    max_overflow=10,
    pool_pre_ping=True,
)