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.
Nine out of ten slow queries are caused by missing indexes or indexes that are not being used. This guide covers PG index types, composite indexes and the leftmost-prefix rule, EXPLAIN ANALYZE, common causes of unusable indexes, and when not to add an index.
Redis is far more than a cache, yet many developers never get beyond SET/GET. This article covers its core data structures, production patterns for caching, distributed locks, rate limiting, and leaderboards, persistence and high-availability choices, and classic pitfalls including cache penetration, breakdown, avalanches, big keys, and expiration.
Model vendor keys, object storage credentials, and signing keys can't be rotated by hand. This article lays out an end-to-end rotation flow: asset inventory, dual-credential switching, dynamic refresh, revocation verification, and leak response.
Single QPS rate limiting cannot stop long-context and batch tasks from squeezing inference pools. This article designs tiered quotas, atomic accounting, fair queuing, burst allowance, and failure degradation to protect multi-tenant AI platforms.
When a single database can no longer keep up, what is the right move? This article lays out the scaling path: optimize vertically first (indexes, queries, read/write splitting), then horizontally (sharding). Covers shard key selection, cross-shard queries, distributed transactions, and why you should not rush to shard — most projects never need it.
Go's goroutine-and-channel model is simple and powerful, but misuse can cause leaks, deadlocks, and races. This guide covers lightweight goroutines and scheduling, channel synchronization and communication, select multiplexing, worker pools, fan-out, pipelines, and common concurrency traps.
When a server fails and SSH leaves you staring at an unfamiliar terminal, this Linux survival kit gives developers everyday file and text commands, process and resource diagnostics (top/htop/lsof/df), network troubleshooting (curl/ss/netstat), pipes and redirection, and a practical incident workflow: check resources, then processes, then logs.
AI requests naturally carry high-dimensional information such as model, tenant, prompt version, and toolchain. This article covers metric label layering, cardinality budgeting, OpenTelemetry attribute trimming, Prometheus acceptance criteria, and over-limit rollback.
SSH is more than remote login. Its tunneling and port forwarding provide a developer's Swiss Army knife for reaching private services, creating secure proxies, and traversing bastion hosts. This article covers local, remote, and dynamic forwarding, ProxyJump, SSH key management, and ssh_config.
Do not want to deploy Elasticsearch? PostgreSQL's built-in full-text search is enough at small and medium scale. This guide covers tsvector and tsquery, tokenization and dictionaries, GIN indexing, ts_rank ranking, Chinese tokenization with zhparser or pg_jieba, and hybrid vector search.
Is your service running blind, with users reporting incidents first? This guide covers practical observability: structured JSON logs and request IDs, log levels and sampling, RED/USE metrics, alert design without fatigue, and a fast incident workflow using metrics, traces, and logs.
Singleflight can coalesce concurrent origin fetches, but it can't stop a stale rebuilder from overwriting the new value after invalidation. This article uses target generations, rebuild leases, conditional publication, and a bounded stale policy to close cache stampede and late-write windows.