danielfm / prometheus-for-developers Goto Github PK

I'm constantly asked about, given some use case, what metric types should be used, and how to extract useful information from those metrics.

One recent example of such question was how to measure cache efficiency.

One answer is that this can be implemented with two Counters: one for counting the cache hits, and other for counting the cache misses.

Some possible queries we might try on these metrics:

• Cache hits per second:

  sum(rate(<hit count>[5m]))

• Cache misses per second:

  sum(rate(<miss count>[5m]))

• Cache hit rate (hits / total):

  sum(rate(<hit count>[5m])) / (sum(rate(<hit count>[5m])) + sum(rate(<miss count>[5m])))

• Calculate montly uptime (i.e. for a Jenkins master instance):

  sum_over_time(jenkins_health_check_score[30d:1m]) / 43200

I imagine there are other common use cases and queries not currently covered in this document that we should add.

(If anyone has more examples of such use cases, please let me know)