Time: 15:40 - 16:30
PostgreSQL provides many metrics for monitoring and documents them exhaustively. But what do these metrics mean? How do you use them to answer the questions you have? What questions should you even be asking?
I'll break down how we built a comprehensive dashboard using
Prometheus and Grafana. It included PostgreSQL global metrics as
well as per-table metrics and eventually even per-query metrics from
pg_stat_statements. It also integrated system level metrics and
Rails application level metrics.
We can see the kinds of problems this level of instrumentatation allowed us to identify and resolve. We can also see the limitations of this kind of high level instrumentation and the kinds of problems it's not well suited for debugging.
The state of the art in this area still leaves much to be desired, particularly with integration between the database metrics and metrics from application and system sources. We'll look at the advanced tools that propose to solve this and future development that PostgreSQL will need to support them.