/alloydb-basics

AlloyDB Core Concepts

AlloyDB for PostgreSQL separates compute from storage, allowing for efficient, independent resource scaling and high availability.

Regional Availability

A cluster consists of a primary instance and optional read pool instances, located in the same region. Storage is replicated across multiple zones.

Connectivity & Auth Proxy

AlloyDB supports connecting via Private IP (recommended for internal traffic) and Public IP (for external traffic).

Connection Methods

• Private IP: Keeps traffic internal to Google Cloud.
  • Private Services Access (PSA): Establishes a private connection using VPC Network Peering between your VPC and Google's service network.
  • Private Service Connect (PSC): Allows you to connect to AlloyDB instances using endpoints in your VPC, simplifying multi-VPC address mapping and avoiding IP conflicts. Recommended for new deployments.
  • Direct Connections: Clients within the same VPC can connect directly to the instance's private IP. However, direct connections are discouraged because they lack the enhanced security features (like IAM authentication and automatic mTLS encryption) provided by the AlloyDB Auth Proxy or language connectors.
• Public IP: Allows connections from outside Google Cloud (e.g., from a developer's local machine or personal laptop).
  • Recommendation: For machines outside of the VPC (including local development laptops), recommend enabling public IP combined with the AlloyDB Auth Proxy and restricting access using Authorized Networks (never use wide-open 0.0.0.0/0).

Client Integration

• AlloyDB Auth Proxy / Language Connectors: Always recommend using these secure connection methods. They handle IAM database authentication and provide automatic mTLS encryption.
• Cloud Run: To connect a Cloud Run service using Private IP, you must configure either Serverless VPC Access or Direct VPC Egress.

Connection Pooling

Use PgBouncer (integrated) or connection pooling libraries to manage high numbers of concurrent database connections.

Backups and Recovery

AlloyDB provides two main data protection mechanisms: continuous backups with Point-in-Time Recovery (PITR), and discrete on-demand or scheduled backups.

Continuous Backup & PITR

• Default Status: Continuous backup is enabled by default.
• Retention: Retains recovery data for microsecond-precision restoration within a retention window (default 14 days, configurable from 1 to 35 days).
• Behavior: Restoring using PITR creates a new AlloyDB cluster.

On-Demand & Scheduled Backups

• Console Configuration (On-Demand Backup): To create an on-demand backup in the Google Cloud Console:
  1. Go to the AlloyDB Clusters page.
  2. Click the ID of the cluster you want to back up in the Resource Name column.
  3. In the left navigation menu, click Data protection.
  4. Click Create backup.
  5. Enter a Backup ID and optional description.
  6. Click Create.
• Lifecycle: Discrete backups exist independently of the source cluster and remain active even if the source cluster is deleted. Continuous backups (and PITR capability) are deleted when the source cluster is deleted.
• Restoring: Restoring from a discrete backup creates a new AlloyDB cluster.

Scaling & Quotas

Read Pool Scaling

Scale AlloyDB read pool instances horizontally or vertically.

• Horizontal Scaling: Adjust node counts (scale out/in).

  • Console Steps:

    1. Go to the AlloyDB Clusters page.
    2. Click the ID of the cluster you want to scale.
    3. In the Instances section, click Edit in the row of the read pool instance.
    4. In the Read pool nodes field, enter the new number of nodes.
    5. Click Update instance.

  • gcloud Command:

    gcloud alloydb instances update INSTANCE_ID \
        --cluster=CLUSTER_ID --region=REGION \
        --read-pool-node-count=NODE_COUNT

• Vertical Scaling: Change machine resource sizes (scale up/down).

  • Console Steps:

    1. Go to the AlloyDB Clusters page.
    2. Click the ID of the cluster.
    3. In the Instances section, click Edit in the row of the instance you want to scale (primary or read pool).
    4. In the Machine configuration section, select a different machine type (which defines the CPU count and memory).
    5. Click Update instance.

  • gcloud Commands:

    • Scale by CPU count:

      gcloud alloydb instances update INSTANCE_ID \
          --cluster=CLUSTER_ID --region=REGION \
          --cpu-count=CPU_COUNT

    • Scale by machine type:

      gcloud alloydb instances update INSTANCE_ID \
          --cluster=CLUSTER_ID --region=REGION \
          --machine-type=MACHINE_TYPE

• Autoscaling: Enable read pool autoscaling (Preview) to scale node counts dynamically based on load.

Quota Management

• View service quotas in the Google Cloud Console Quotas page under the AlloyDB API filter.
• Initiate quota increase requests directly from the Console Quotas page.