PG Exporter

PG Exporter

Prometheus exporter for PostgreSQL metrics. Gives you complete insight on your favourate elephant!

Latest binaries & rpms can be found on release page. Supported pg version: PostgreSQL 9.4+ & Pgbouncer 1.8+. Default collectors definition is compatible with PostgreSQL 10,11,12,13.

Latest pg_exporter version: 0.3.1

Features

• Support both Postgres & Pgbouncer
• Flexible: Almost all metrics are defined in customizable configuration files in SQL style.
• Fine-grained execution control (Tags Filter, Facts Filter, Version Filter, Timeout, Cache, etc…)
• Dynamic Planning: User could provide multiple branches of a metric queries. Queries matches server version & fact & tag will be actually installed.
• Configurable caching policy & query timeout
• Rich metrics about pg_exporter itself.
• Auto discovery multi database in the same cluster (multiple database scrape TBD)
• Tested and verified in real world production environment for years (200+ Nodes)
• Metrics overhelming! Gives you complete insight on your favourate elephant!
• (Pgbouncer mode is enabled when target dbname is pgbouncer)

性能表现

对于极端场景（几十万张表与几万种查询），一次抓取最多可能耗费秒级的时长。

好在所有指标收集器都是可选关闭的，且pg_exporter 允许为收集器配置主动超时取消（默认100ms）

自监控

Exporter展示了监控系统组件本身的监控指标，包括：

• Exporter是否存活，Uptime，Exporter每分钟被抓取的次数
• 每个监控查询的耗时，产生的指标数量与错误数量。

Prometheus的配置

Prometheus的抓取频率建议采用10～15秒，并配置适当的超时。

演示或特殊情况也可以配置的更精细（例如2秒，5秒等）

单Prometheus节点可以支持几百个实例的监控，约几百万个时间序列 （Dell R740 64 Core / 400GB Mem/ 3TB PCI-E SSD）

更大规模的集群可以通过Prometheus级联、联邦或分片实现伸缩。例如为每一个数据库集群部署一个Prometheus，并使用上级Prometheus统筹抓取并计算衍生指标