Database Services
Domain 4 - Database Services. Questions test managed-database limits, HA and backup design, and choosing an engine for a workload.
Managed engines
Cloud DB for MySQL, MSSQL, PostgreSQL, MongoDB, Redis, plus CUBRID and Tibero for Korean enterprise workloads. Managed means no OS or root access: you configure through console/API, receive DNS endpoints, and rely on the platform for patching, backup and failover. SQL engines fit transactions; MongoDB fits document data; Redis fits caching, sessions and queues.
Design rules
- Place the database in a private subnet; allow only the application ACG on the engine port.
- Connect through the provided DNS endpoint, never a fixed IP, so failover keeps working.
- Configure automated backups with a retention window and run a restore test.
- Add read replicas for reporting traffic; never send writes to a replica.
- Resize vertically with a maintenance window; plan the downtime.
Worked example
A MySQL-backed commerce service needs HA and reporting. Reasoning: Cloud DB for MySQL in the Korea region with a standby for failover; private subnet; ACG allowing 3306 from app servers only; nightly automated backup with 14-day retention; a read replica for the BI dashboard; the application reads the endpoint from the console rather than hard-coding an address.
Traps
- Expecting SSH or root access to a managed instance.
- Assuming HA is on by default - it is a design choice.
- Confusing backup with HA: HA keeps you serving, backup restores data.
- Hard-coding an instance IP in application configuration.
- Trying to autoscale a database the way you autoscale stateless app servers.