Professional Cloud Database EngineerDemo

Cloud Spanner is a horizontally scalable, strongly consistent ACID database suited to durable ecommerce transaction data such as shopping carts. A regional Spanner instance has a 99.99% monthly uptime SLA, satisfying the less-than-five-minutes-per-month availability target, and Spanner scales compute and storage to support high write throughput. Multi-region configurations can provide even higher availability when required.

A regional Cloud SQL HA configuration provides a primary and standby across separate zones, enabling automatic failover and preserving availability for writes. Cross-region read replicas in us-east1 and us-west1 place read capacity nearer to applications in those regions, which improves read latency while the HA primary handles writes in its configured region.

Cloud Spanner multi-region configurations replicate data across regions and maintain availability during regional failures. Spanner supports customer-managed encryption keys (CMEK) in Cloud KMS, which lets the customer control key location, lifecycle, permissions, and rotation. Cloud KMS symmetric encryption keys use AES-256.

Database Migration Service conversion workspaces convert Oracle schema and supported code objects, including stored procedures, to PostgreSQL syntax. A continuous Database Migration Service job performs the initial data backfill and then uses change data capture (CDC) to keep AlloyDB synchronized until cutover, which reduces downtime. Legacy conversion workspaces are older, more limited, use Ora2Pg-based conversion, and are not recommended for migrations.

AlloyDB automated backups can be scheduled daily and retained for seven days. Enabling continuous backup and recovery provides point-in-time recovery and an RPO of zero, which is more stringent than a maximum six-hour data-loss tolerance; it also supports recovery from a recent point in time for the required recovery objective.

Evicted keys indicate that the instance has reached its maxmemory limit. Evictions can remove frequently used cached data and degrade application performance; increasing the instance’s memory capacity reduces memory pressure and lowers eviction frequency. Google Cloud recommends scaling up when cache capacity is insufficient for the workload.

Cloud Spanner PITR uses stale reads at a timestamp before corruption to surgically recover a portion of a database, after which the recovered results are written back. For significant or whole-database corruption, create a backup at the required recovery timestamp and restore it to a new database. Backup and restore is the appropriate full-database recovery method rather than export/import for this scenario.

Cloud SQL self-service maintenance is intended when an update is needed before the next scheduled maintenance event. When it is initiated on a primary instance with read replicas, Cloud SQL applies the target maintenance version to all read replicas and then updates the primary instance. Scheduled maintenance cannot be rescheduled when less than 24 hours remain before its originally scheduled time.

Cloud SQL high-availability instances require automated backups and point-in-time recovery (PITR) to be enabled. Automated backups provide scheduled recoverable backup points, while PITR supports recovery to a selected point using transaction logs.

A Cloud SQL high-availability configuration uses a primary and standby in separate zones within the same region. Google recommends placing read replicas in a different zone from both the primary and standby so that the replica can continue operating if either of those zones has an outage.

Cloud Scheduler can run on a defined schedule and invoke a Cloud Function that initiates Firestore’s managed export operation. This serverless pattern is the documented approach for scheduled Firestore exports and avoids operating a virtual machine or building an unnecessary data-processing pipeline.

Bare Metal Solution for Oracle lets organizations run Oracle databases, including Oracle Real Application Clusters (RAC), on certified bare-metal infrastructure while retaining existing Oracle capabilities, tools, and operational practices. This supports a lift-and-shift migration with minimal application changes.

For a Cloud SQL instance managed by Terraform, deletion_protection = true is a Terraform-level safeguard that prevents Terraform from deleting the instance. If a plan requires a destroy-and-recreate operation, Terraform blocks it until deletion protection is explicitly disabled.

A Persistent Disk can be increased while attached to a running VM. For an ext4-formatted non-boot data disk, run resize2fs after increasing the disk size so the filesystem uses the added capacity; the VM does not need to be restarted.

A Dataflow pipeline using DynamoDBIO and BigtableIO provides managed, scalable data movement from DynamoDB to Bigtable. Streaming records in near real time keeps the target closely synchronized during migration, which supports a short cutover window and avoids the operational effort and consistency gap of a one-time file export or a custom transfer program. BigtableIO supports batch and streaming pipeline operations with Dataflow.

Cloud SQL for PostgreSQL cross-region replicas support planned regional migration with minimal downtime by allowing replication to catch up before promotion. In Enterprise Plus, a cross-region replica can be designated for advanced disaster recovery and failed over during a regional failure; failover testing validates that recovery capability.

Cloud Spanner distributes data and write load by primary-key ranges. Putting a high-cardinality attribute first in a composite primary key helps distribute writes across more key ranges, whereas a low-cardinality leading attribute can concentrate them. Bit-reversed sequential primary-key values are distributed evenly across the key space, avoiding the hotspot behavior caused by monotonic auto-incrementing values.

Cloud Spanner provides ACID read-write transactions for globally distributed transactional applications. Its multi-regional and dual-regional configurations have a 99.999% monthly uptime objective, satisfying the required availability level.

Cloud Spanner Autoscaler's linear scaling method is designed for workloads that change gradually or experience a few large peaks. It automatically calculates and adds the compute capacity required to keep utilization below the configured threshold, supporting continued growth and holiday concurrency spikes.

Customer-managed encryption keys let an organization manage encryption-key lifecycle activities in Cloud KMS and use keys located in the required region. Cloud SQL supports CMEK when creating the database instance, while encryption at rest and in transit is provided by the service.

Data dump parallelism accelerates the initial migration load, but it increases load on the source database. Using an Amazon RDS MySQL read replica as the migration source isolates that work from the production primary, and setting Initial Load Parallelism to Maximum maximizes initial-load throughput. Google Cloud also documents using a read replica when locking or disrupting the source is not acceptable.

Cloud Spanner is a fully managed, horizontally scalable relational database that provides SQL semantics and transactional consistency at global scale, making it suitable for an application that requires relational data management at hundreds-of-terabytes scale.

Firestore bundles can package the daily news data for download and client-side loading. Serving those bundles through a CDN caches and replicates the static content close to users globally, reducing repeated Firestore reads, delivery cost, and latency.

A custom app profile using multi-cluster routing routes traffic to an available cluster in the replicated Bigtable instance. When one cluster becomes unavailable, Bigtable automatically fails over requests to another available cluster. Single-cluster routing cannot provide that automatic cluster failover.