Monitoring

• Metrics, tracing, and logging，Metrics、tracing、logging三个监控系统的区别和联系
• Observability 3 ways: logging metrics and tracing，上篇文章的Slides。

Performance benchmark

• “How NOT to Measure Latency” by Gil Tene，如何正确解读监控/压力测试结果

Metrics

Prometheus

基本概念：

• How does a Prometheus Counter work?，对于理解rate()函数至关重要
• Counting with Prometheus [I] ，上篇博客的关联Presentation
• rate()/increase() extrapolation considered harmful 关于rate()函数extrapolation（外推）算法的讨论
• How does a Prometheus Gauge work?，gauge类型的分析
• irate graphs are better graphs，irate提供了更即时的结果
• Avoid irate() in alerts
• Rate then sum, never sum then rate，rate在前sum在后
• Why are Prometheus histograms cumulative?，histogram类型的分析

几个使用技巧：

• Existential issues with metrics，使用metrics-based monitoring system的的注意事项
• Common query patterns in PromQL，几个常见的PromQL语句
• Composing range vector functions in PromQL，如何实现诸如这样的查询：最近1小时内，rate(x[5m])的最高值

运维技巧：

• Scaling and Federating Prometheus
• Dropping metrics at scrape time with Prometheus

Machine metrics

• Understanding Machine CPU usage，虽然是P8S的一篇博客，但是对于理解常见的几个CPU指标还是有用的

Tracing

TODO

Logging

TODO