Prometheus Queries

notes.md

• Get rid of pushgateway
• Saturation - Load testing to determine upper threshold (up 10 rq/sec)
• Imminent
• Should we use summeries
  • Historical
  • There might be situation where you don't care about aggregation
• Historgrams
  • Use up to 10 bukcets
  • Some use cases if you don't know the
  • Aggregations, quantile, timewindow at query time
  • Problem is that you need to choose buckets up from with latency expectations
  • AppDex - Good latency, acceptanceable , bad latency aggregated
  • Even simpler, two buckets

USE - Utilization Saturation Errors RED - requests errors durations LEST - Latency Errors Saturation Traffic

prometheus.md

• List the number of timeseries per metric name/job sort_desc(count by (__name__, job) ({__name__=~".+"}) )

• Sum the number of nginx connections per instances sum by (instance)(nginx_http_connections{instance=~"185.+"})

• Rule for DST IF ( (month() => 3 and day_of_month() => 26) or (month() <= 10 and day_of_month() < 29) ) and hour() == 10 and minute() >= 35