DevOps、Docker、Kubernetes、可观测性
This collection features articles and community discussions in the 云原生 space, focusing on real-world engineering scenarios, tool selection, practical paths, and risk boundaries.
The page is continuously updated and serves as a learning roadmap, technology selection reference, and technical review entry point.
Some data is naturally semi-structured, including configurations, events, and nested objects. PostgreSQL JSONB combines relational structure with document flexibility in one database. This guide covers storage, access, operators, GIN acceleration, and when to use JSONB instead of relational columns.
Redis is more than a cache. In distributed systems it provides locks, rate limiting, messaging with pub/sub and Streams, leaderboards, and session storage. This article explains each use case and its key reliability concerns, including safe lock release, Stream consumer groups, and accurate rate limiting.
Microservices are not a silver bullet; a bad decomposition hurts more than a monolith. This article explains the core patterns: service boundaries, interservice communication (synchronous vs. asynchronous), API gateways, service discovery, distributed transactions, observability, and, most importantly, when to use microservices (most teams should not).
Data is life; lose it and you are done. This article explains PostgreSQL high availability and disaster recovery: streaming replication, synchronous vs. asynchronous replication, physical backups with pg_basebackup and WAL archiving, logical replication, the 3-2-1 backup strategy, and recovery drills that help databases withstand failures.
Kafka is more than a message queue; it is an event streaming platform and a foundation for logs, stream processing, and event-driven systems. This article explains Kafka's topic, partition, offset, and consumer-group model, its high-throughput design, ordering and Exactly-Once guarantees, and when to use it.
On the cloud it's easy to 'get more expensive as you use it' — idle resources, over-provisioning, forgotten services, traffic fees. Covers cost optimization levers: compute (reserved/spot/right-sizing), storage (tiering/cleanup), network (egress/CDN), bill monitoring + alerts, and the 'FinOps' culture making costs visible and accountable.
Creating a new database connection for every request can bring PostgreSQL down as concurrency rises. This article explains why PostgreSQL connections are expensive, application-side pooling, PgBouncer, transaction mode, and how to calculate connection limits.
A confidential VM or GPU label is not automatically trustworthy. This article uses RATS roles, EAT freshness, composite measurements, and policy-based key release to turn confidential inference into a verifiable control.
Consistency is unavoidable in distributed systems. This article explains CAP tradeoffs, how Raft reaches consensus, the choice between eventual and strong consistency, Raft elections and log replication, and how primary-replica, Quorum, and CRDT systems navigate different consistency levels.
A successful write to the configuration center does not mean every instance downloaded, validated, and applied the configuration. Restarts and late ACKs can also create a false majority. Freeze the membership snapshot, configuration digest, and health window before atomically promoting the active generation.
CI/CD isn't just 'auto-run tests and deploy' — its design determines delivery speed and quality. Covers pipeline design principles: speed (parallelism/caching/incremental), layering (lint→test→build→deploy), environment isolation (dev/staging/prod), security (gates/manual approval), observability, and the 'small batches, fast flow' engineering culture.
Changing a database schema is high-risk: production data cannot be lost and the system cannot stop. This article covers migration tools (Prisma migrate/Flyway/Liquibase), forward/backward compatible migration strategies, zero-downtime migrations (expand → migrate → contract), rollbacks, and the pitfalls of changing schema on large tables.