Google Cloud Data Platform Engineering, Migration & Managed Operations

Values are illustrative. Each real change is recorded this way from measured workload evidence, applied first in a non-production instance, and scheduled for a maintenance window when the flag restarts the instance.

Google Cloud Data Platform Engineering also covers the MySQL and SQL Server editions of Cloud SQL, with engine depth from our PostgreSQL support and MySQL support teams.

SQL · pg_stat_statements on Cloud SQL or AlloyDB

-- Cloud SQL / AlloyDB for PostgreSQL: top statements by total execution time
-- (pg_stat_statements column names for PostgreSQL 13 and later)
SELECT queryid,
       calls,
       ROUND(total_exec_time::numeric / 1000, 1)        AS total_s,
       ROUND(mean_exec_time::numeric, 2)                AS mean_ms,
       ROUND(100.0 * shared_blks_hit
             / NULLIF(shared_blks_hit + shared_blks_read, 0), 1) AS cache_hit_pct,
       rows,
       LEFT(query, 100)                                 AS query_head
FROM pg_stat_statements
ORDER BY total_exec_time DESC
LIMIT 20;

In Google Cloud data platform engineering, the same statement ranking runs on both services, which makes it the fairest test when deciding between them: we replay the top statements against a restored copy on each and compare latency, throughput and cost per transaction.

Rollback is planned before cutover, not after: the source stays intact, and for PostgreSQL targets we can set up reverse logical replication so the old primary can take traffic back. The Database Migration Service documentation covers supported sources and versions, which we confirm per engagement.

Bash · gcloud: verify, confirm, promote, validate

#!/usr/bin/env bash
# Cutover gate for a Database Migration Service continuous job (for example PostgreSQL to AlloyDB)
set -euo pipefail
: "${PROJECT_ID:?set PROJECT_ID}" "${REGION:?set REGION}" "${MIGRATION_JOB:?set MIGRATION_JOB}"

# 1. Verify before: the job is RUNNING and in the CDC phase
gcloud database-migration migration-jobs describe "${MIGRATION_JOB}" \
  --region="${REGION}" --project="${PROJECT_ID}" \
  --format="value(state,phase)"            # expected: RUNNING  CDC

# 2. Confirmation gate: application writes stopped, replication lag at zero in Cloud Monitoring
read -r -p "Writes stopped and lag at 0 for ${MIGRATION_JOB}? Type the job name to promote: " answer
if [[ "${answer}" != "${MIGRATION_JOB}" ]]; then
  echo "Aborted. No change made."; exit 1
fi

# 3. Promote the destination. The source stops replicating to it from this point.
gcloud database-migration migration-jobs promote "${MIGRATION_JOB}" \
  --region="${REGION}" --project="${PROJECT_ID}"

# 4. Validate after: the job reports completion; then run row-count and checksum reconciliation
gcloud database-migration migration-jobs describe "${MIGRATION_JOB}" \
  --region="${REGION}" --project="${PROJECT_ID}" \
  --format="value(state,phase)"            # expected: COMPLETED once promotion finishes