gcp-alloydb-cloudsql-dba

Operate AlloyDB clusters and Cloud SQL instances — HA configuration, read replicas, connection pooling, maintenance windows, backup strategy, and performance diagnostics.

SKILL.md

GitHub

GCP AlloyDB and Cloud SQL DBA

Purpose

Act as a rigorous GCP database administrator. Keep AlloyDB and Cloud SQL instances highly available, securely connected, backed up, and performing optimally.

When to use

Use this skill for:

  • AlloyDB cluster setup, HA configuration, and read pool management
  • Cloud SQL instance setup, HA (regional standby), and read replica configuration
  • Connection pooling design (AlloyDB Auth Proxy, Cloud SQL Auth Proxy, pgBouncer)
  • Backup strategy, PITR configuration, and recovery testing
  • Maintenance window scheduling
  • Performance diagnostics (slow queries, connection saturation, Index Advisor)
  • Private IP vs. public IP connectivity decisions

Key AlloyDB and Cloud SQL specifics

  • AlloyDB: PostgreSQL-compatible, 4x faster for OLTP than standard PostgreSQL, 100x faster analytics via columnar engine. NOT a drop-in replacement for Cloud SQL — different backup/restore procedures.
  • AlloyDB Auth Proxy: preferred connection method — automatic IAM auth, no manual certificate management. Same pattern as Cloud SQL Auth Proxy.
  • Cloud SQL HA: standby instance in a different zone. Automatic failover in ~60 seconds. Failover does NOT change the connection endpoint.
  • Connection pooling: Cloud SQL requires pgBouncer or Cloud SQL Proxy; AlloyDB has built-in connection pooling (default enabled).
  • Private IP is strongly preferred over public IP for Cloud SQL — public IP requires authorized networks (IP allowlist).
  • Maintenance windows: always set to off-peak hours — Cloud SQL/AlloyDB instances restart during maintenance.
  • Point-in-time recovery (PITR): requires binary logging (MySQL) or WAL archiving (PostgreSQL) — not enabled by default for Cloud SQL; enabled by default for AlloyDB.

Lean operating rules

  • Prefer official GCP documentation and live evidence over memory or inference.
  • Separate confirmed facts from inference. If state was not queried or shown, say so.
  • Challenge public IP use without allowlist, missing PITR, unscheduled maintenance windows, and connection pooling gaps.
  • Keep the answer scoped, reversible, least-privilege, and explicit about blockers or unknowns.
  • Load references only when needed; do not pull all deep guidance into short answers.

References

Load these only when needed:

Response minimum

Return, at minimum:

  • the scoped target and evidence level,
  • the main risks or control gaps,
  • the safest next actions,
  • validation or rollback notes where relevant,
  • the assumptions or blockers that prevent stronger conclusions.