managed-databases-architect
Architect-level managed database selection and operation — choosing between relational, document, key-value, and analytical engines, sizing and IOPS, multi-AZ and read replicas, connection pooling, backup and point-in-time recovery, and the operational limits of the managed tier. Use when a database is a cloud service rather than a server you run.
- License 许可
- MIT
- 最近更新
- 2026-11-13
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- managed-databases-architect
- license
- MIT
- metadata
- { "crusor-path": "cloud-and-devops-architect/data-services-architect/managed-databases-architect", "crusor-domain": "cloud-and-devops-architect", "crusor-origin": "original", "crusor-sources": "docs.aws.amazon.com/AmazonRDS/latest/UserGuide; cloud.google.com/sql/docs; docs.aws.amazon.com/amazondynamodb/latest/developerguide", "crusor-triggers": "rds, aurora, dynamodb, managed database, read replica, point in time recovery, connection pooling, database sizing, cloud sql, cosmos db", "crusor-verified-on": "2026-10-30" }
- description
- Architect-level managed database selection and operation — choosing between relational, document, key-value, and analytical engines, sizing and IOPS, multi-AZ and read replicas, connection pooling, backup and point-in-time recovery, and the operational limits of the managed tier. Use when a database is a cloud service rather than a server you run.
- compatibility
- RDS/Aurora, Cloud SQL, Azure SQL, DynamoDB, Cosmos, Firestore
03 SKILL.md 正文
Managed Databases Architect
A managed database removes operations, not design. Schema, access pattern, and connection discipline are still yours, and in some ways more consequential because you cannot tune the box.
When to use
- Choosing a managed engine for a new service.
- Performance or cost problems on RDS/Aurora/Cloud SQL/DynamoDB.
- Designing backups, replicas, or failover.
When NOT to use
- Schema modelling detail → web
database-engineering-architect/*. - Self-managed clusters on Kubernetes →
kubernetes-architect(and question the choice).
Decision 1 — Engine
| Need | Engine |
|---|---|
| relational, joins, transactions, known schema | PostgreSQL (often Aurora/Cloud SQL) |
| relational, extreme read scale, storage growth | Aurora/Spanner-class |
| key-value at scale, predictable access by key | DynamoDB/Bigtable |
| document, evolving schema, per-document queries | MongoDB/Cosmos/Firestore |
| analytical OLAP | warehouse, not OLTP engine |
- Managed multi-model engines tempt you into one engine for everything; the access-pattern mismatch shows up at scale, not at launch.
- DynamoDB design is access-pattern-first: get the keys wrong and the fix is a rewrite, not an index.
Decision 2 — Sizing
- Size on the working set (fits in memory?) plus IOPS, not on row count.
- Provision IOPS explicitly on gp3-style storage; default burst credits run out under sustained load.
- Serverless/on-demand tiers are for spiky, low-duty workloads; sustained load is cheaper on provisioned capacity.
- Measure p95 query latency and buffer cache hit ratio; both predict the wall.
Decision 3 — Availability
- Multi-AZ for anything in the request path; understand the failover time (seconds to minutes) and that connections drop.
- Read replicas for read scale and for analytics isolation — not as a backup.
- Failover tested; an untested failover is a guess about the app's retry logic.
- Cross-region replicas for DR with a documented RPO and promotion runbook.
Decision 4 — Connections
- The app's pool size × instance count must be below the database's max connections. This is the single most common cloud outage.
- Use a proxy/pooler (RDS Proxy, PgBouncer, Cloud SQL connector) for serverless/bursty workloads.
- Timeouts: connect, statement, idle — all set; a missing one becomes a stuck service.
- Long transactions block vacuum and locks; cap and alert on them.
Decision 5 — Backups
- Automated backups with PITR enabled and retention matching the business.
- Snapshots copied to another region/account; test a restore, not just a backup.
- Know the restore time: a 1 TB database is not a 5-minute operation.
- Logical exports for reporting/portability alongside physical backups.
Decision 6 — Cost
- Idle multi-AZ instances in dev; schedule them off.
- Storage grows with WAL/backups — set retention explicitly.
- Reserved capacity for steady instances; on-demand for dev.
- The expensive surprise is usually cross-AZ data transfer and replicas nobody reads.
Failure modes at scale
- 500 app connections from 20 Lambdas; database refuses connections under load.
- Failover takes 90 s, app has no retry and no connection refresh.
- Read replica used for writes in a reporting job, silently failing writes.
- PITR off; a bad migration requires restoring a week-old snapshot.
- DynamoDB table with one hot partition key; throttling in production only.
Verification receipts
- Engine choice with the access pattern it serves.
- Connection budget: pool × instances vs max connections, with headroom.
- Failover drill: connection loss handled, recovery time measured.
- Restore drill from PITR to a timestamp, verified data intact.
- Cost breakdown: instance, storage, IOPS, replicas, transfer.
04 同仓库更多 Skills
interaction-patterns-architect
Architect-level interaction design in code — choosing established patterns over invented ones, feedback and latency behaviour, destructive-action safeguards, navigation and overlay conventions, and the state model behind every interactive surface. Use when a screen has to behave, not just look.
design-system-ops-architect
Architect-level operation of a design system in code — token pipelines and versioning, component API and variant policy, the catalogue as the contract, adoption metrics, and the governance that stops a system from forking into three. Use when a component library must survive several teams.
design-handoff-architect
Architect-level design handoff — what must be specified before implementation starts, tokens over raw values, responsive and content-limit contracts, state matrices, and the review loop that catches drift between design and code. Use when a design is about to become code.
content-microcopy-architect
Architect-level UI content — where strings live, voice and terminology consistency, error and empty-state copy as part of the interface, translation-ready authoring, and the content model a product needs before it multiplies screens. Use when the words in the interface matter as much as the pixels.
ui-ux-implementation-architect
Architect-level direction for turning a design into working UI — the handoff contract, component ownership versus screen-level layout, interaction and state conventions, design QA gates, and the review habits that keep an implementation faithful without frozen pixels. Use before any specific UI implementation skill.
test-strategy-architect
Architect-level test strategy — allocating assertions across unit, integration, contract and end-to-end layers, designing for testability through dependency boundaries, fixture and data strategy, and the flake and coverage policies that keep a suite trustworthy. Use when defining how a codebase is tested.