Google Cloud SQL & AlloyDB

Q: Postgres 18 on Cloud SQL?

Check gcloud sql database-versions list - Google ships majors on their cadence.

Google Cloud offers Cloud SQL for PostgreSQL for standard managed instances and AlloyDB for PostgreSQL for higher-performance OLTP with separated compute and storage. Both support HA, automated backups, and read pools, with different failover models and pricing profiles.

Recipe

# gcloud: create Cloud SQL Postgres HA instance (conceptual)
gcloud sql instances create orders-prod \
  --database-version=POSTGRES_18 \
  --tier=db-custom-4-16384 \
  --region=us-central1 \
  --availability-type=REGIONAL \
  --storage-auto-increase

-- After connect via Cloud SQL Auth Proxy or private IP
SELECT version();
CREATE EXTENSION IF NOT EXISTS pg_stat_statements;

When to reach for this:

GCP-native apps on GKE or Cloud Run needing private IP SQL
Choosing AlloyDB vs Cloud SQL for latency-sensitive OLTP
Designing read pool for analytics traffic

Working Example

Regional HA Cloud SQL with read replica and Auth Proxy sidecar on GKE.

# kubernetes deployment excerpt - Cloud SQL Auth Proxy 2.x
containers:
  - name: cloud-sql-proxy
    image: gcr.io/cloud-sql-connectors/cloud-sql-proxy:2.14.0
    args:
      - "--structured-logs"
      - "--private-ip"
      - "my-project:us-central1:orders-prod"
  - name: api
    env:
      - name: DATABASE_URL
        value: "postgresql://app@127.0.0.1:5432/orders?sslmode=disable"

# Read replica instance for BI (async)
gcloud sql instances create orders-replica \
  --master-instance-name=orders-prod \
  --tier=db-custom-2-8192 \
  --region=us-central1

-- On replica
SELECT pg_is_in_recovery();
-- Route ORM read-only connection to replica IP via separate DSN

What this demonstrates:

Regional HA = primary + standby in region (automatic failover)
Auth Proxy handles IAM and TLS to Cloud SQL without static IP whitelist on laptop
Read replica DSN isolates reporting load

Deep Dive

Cloud SQL vs AlloyDB

Aspect	Cloud SQL PostgreSQL	AlloyDB
Positioning	General managed Postgres	High-performance PostgreSQL-compatible
Storage	Standard PD/SSD	AlloyDB intelligent storage
Read scaling	Read replicas	Read pool nodes (low lag)
Best for	Typical web OLTP	Demanding latency/throughput
Extensions	Google allowlist	Allowlist (check matrix)

HA Models

Zonal: single zone, no automatic HA failover
Regional: primary + standby zone, automatic failover (production default)
Maintenance windows: declare; test failover quarterly

Networking

Private IP via VPC peering (recommended)
Public IP disabled unless legacy requirement
Auth Proxy for IAM DB auth and secure tunnel from CI/GKE

Gotchas

Public IP with authorized networks 0.0.0.0/0 - internet exposure. Fix: private IP + proxy only.
Single-zone prod instance - AZ outage is downtime. Fix: availability-type=REGIONAL.
Read replica lag ignored - stale BI dashboards. Fix: monitor replication lag metric; cap acceptable staleness.
AlloyDB cost surprise - CPU+storage+read nodes add up. Fix: POC with realistic workload before commit.
Extension parity - postgis, vector versions differ Cloud SQL vs AlloyDB. Fix: verify before architecture sign-off.
Connection from Cloud Run without proxy - VPC connector required for private IP. Fix: serverless VPC access or proxy sidecar pattern.

Alternatives

Alternative	Use When	Don't Use When
AlloyDB	Need Google scale-out read pool	Simple small DB, cost-sensitive
Cloud SQL zonal dev	Cheapest GCP sandbox	Production SLA
Spanner	Global strong consistency	Need pure Postgres SQL/extensions
Self-hosted on GCE	Full extension control	Want managed patching

FAQs

Postgres 18 on Cloud SQL?

Check gcloud sql database-versions list - Google ships majors on their cadence.

IAM database auth?

Supported with Auth Proxy and connector libraries - short-lived tokens similar AWS IAM.

Point-in-time recovery?

Transaction log retention configurable - align with RPO policy.

Cross-region DR?

Read replica in second region; promote manually or automate via runbook.

pgvector?

Check extension list per version; test HNSW parameters on staging AlloyDB.

Maintenance notifications?

Enable in console; fail over test after maintenance in staging first.

Cloud SQL Insights?

Query insights UI complements pg_stat_statements for top SQL.

AlloyDB columnar engine?

For analytics acceleration - not replacement for warehouse; evaluate scan patterns.

Quota limits?

Instances per project, connection limits per tier - plan PgBouncer early.

Export to GCS?

Automated backup to bucket for long-term retention beyond default window.

Managed Postgres Basics - shared responsibility
AWS RDS & Aurora Postgres - AWS comparison
Neon & Supabase - serverless alternatives
Cloud Selection ADR - decision matrix
Read Replicas for Reporting - read routing

Stack versions: This page was written for PostgreSQL 18.4 (stable 18, maintenance 17), pgvector 0.8+, PgBouncer 1.x, Patroni 3.x, and PostGIS 3.5+.

Recipe

# gcloud: create Cloud SQL Postgres HA instance (conceptual)
gcloud sql instances create orders-prod \
  --database-version=POSTGRES_18 \
  --tier=db-custom-4-16384 \
  --region=us-central1 \
  --availability-type=REGIONAL \
  --storage-auto-increase

-- After connect via Cloud SQL Auth Proxy or private IP
SELECT version();
CREATE EXTENSION IF NOT EXISTS pg_stat_statements;

When to reach for this:

GCP-native apps on GKE or Cloud Run needing private IP SQL
Choosing AlloyDB vs Cloud SQL for latency-sensitive OLTP
Designing read pool for analytics traffic

Working Example

Regional HA Cloud SQL with read replica and Auth Proxy sidecar on GKE.

# kubernetes deployment excerpt - Cloud SQL Auth Proxy 2.x
containers:
  - name: cloud-sql-proxy
    image: gcr.io/cloud-sql-connectors/cloud-sql-proxy:2.14.0
    args:
      - "--structured-logs"
      - "--private-ip"
      - "my-project:us-central1:orders-prod"
  - name: api
    env:
      - name: DATABASE_URL
        value: "postgresql://app@127.0.0.1:5432/orders?sslmode=disable"

# Read replica instance for BI (async)
gcloud sql instances create orders-replica \
  --master-instance-name=orders-prod \
  --tier=db-custom-2-8192 \
  --region=us-central1

-- On replica
SELECT pg_is_in_recovery();
-- Route ORM read-only connection to replica IP via separate DSN

What this demonstrates:

Regional HA = primary + standby in region (automatic failover)
Auth Proxy handles IAM and TLS to Cloud SQL without static IP whitelist on laptop
Read replica DSN isolates reporting load

Deep Dive

Cloud SQL vs AlloyDB

Aspect	Cloud SQL PostgreSQL	AlloyDB
Positioning	General managed Postgres	High-performance PostgreSQL-compatible
Storage	Standard PD/SSD	AlloyDB intelligent storage
Read scaling	Read replicas	Read pool nodes (low lag)
Best for	Typical web OLTP	Demanding latency/throughput
Extensions	Google allowlist	Allowlist (check matrix)

HA Models

Zonal: single zone, no automatic HA failover
Regional: primary + standby zone, automatic failover (production default)
Maintenance windows: declare; test failover quarterly

Networking

Private IP via VPC peering (recommended)
Public IP disabled unless legacy requirement
Auth Proxy for IAM DB auth and secure tunnel from CI/GKE

Gotchas

Public IP with authorized networks 0.0.0.0/0 - internet exposure. Fix: private IP + proxy only.
Single-zone prod instance - AZ outage is downtime. Fix: availability-type=REGIONAL.
Read replica lag ignored - stale BI dashboards. Fix: monitor replication lag metric; cap acceptable staleness.
AlloyDB cost surprise - CPU+storage+read nodes add up. Fix: POC with realistic workload before commit.
Extension parity - postgis, vector versions differ Cloud SQL vs AlloyDB. Fix: verify before architecture sign-off.
Connection from Cloud Run without proxy - VPC connector required for private IP. Fix: serverless VPC access or proxy sidecar pattern.

Alternatives

Alternative	Use When	Don't Use When
AlloyDB	Need Google scale-out read pool	Simple small DB, cost-sensitive
Cloud SQL zonal dev	Cheapest GCP sandbox	Production SLA
Spanner	Global strong consistency	Need pure Postgres SQL/extensions
Self-hosted on GCE	Full extension control	Want managed patching

FAQs

Postgres 18 on Cloud SQL?

Check gcloud sql database-versions list - Google ships majors on their cadence.

IAM database auth?

Supported with Auth Proxy and connector libraries - short-lived tokens similar AWS IAM.

Point-in-time recovery?

Transaction log retention configurable - align with RPO policy.

Cross-region DR?

Read replica in second region; promote manually or automate via runbook.

pgvector?

Check extension list per version; test HNSW parameters on staging AlloyDB.

Maintenance notifications?

Enable in console; fail over test after maintenance in staging first.

Cloud SQL Insights?

Query insights UI complements pg_stat_statements for top SQL.

AlloyDB columnar engine?

For analytics acceleration - not replacement for warehouse; evaluate scan patterns.

Quota limits?

Instances per project, connection limits per tier - plan PgBouncer early.

Export to GCS?

Automated backup to bucket for long-term retention beyond default window.

Managed Postgres Basics - shared responsibility
AWS RDS & Aurora Postgres - AWS comparison
Neon & Supabase - serverless alternatives
Cloud Selection ADR - decision matrix
Read Replicas for Reporting - read routing

Stack versions: This page was written for PostgreSQL 18.4 (stable 18, maintenance 17), pgvector 0.8+, PgBouncer 1.x, Patroni 3.x, and PostGIS 3.5+.

Google Cloud SQL & AlloyDB

Search Documentation

Google Cloud SQL & AlloyDB