Lots of great tidbits there from a guy who really knows his performance stuff (ex-Sun, now Netflix)
On Sunday, 14 December 2014, Vick Khera <vivek@khera.org> wrote:
On Sat, Dec 13, 2014 at 1:55 PM, Edson Carlos Ericksson Richter <edsonrichter@hotmail.com> wrote:
I've been searching in web for guidelines on OS (Linux) and PostgreSQL (9.3.5) active monitoring best practices.
Recent trends are more toward monitoring response latency by first establishing a baseline level of activity and latency, then alerting when those numbers get out of acceptable range.
There are some open source tools to collect and sort and report this way (see Kibana and Grafana and their underlying data stores). I've not seen alerting tools based on this that are non-commercial, though. Two services I know of are Ruxit and Circonus.
Personally I still use Nagios to tell my staff when things are down or not responding, but often that is too late to proactively fix things.
One thing that'd be really cool is to use the new binary JSON storage in the upcoming Pg release to store the time series data for use with Grafana... but then you'd have a chicken/egg problem with monitoring itself. :)
