Skip to main content

Metric types

Usefull links:

Counter​

  • Always increases or resets to 0
  • Common fuctions: rate(), increase(), irate(), resets()
  • Metrics mostly named as '_count' or '_total'
  • Examples: number of requests, responses, errors

rate() - the per-second average rate of increase of the time series in the range vector. rate(some_metric[5m]) - average of all data points divided by 300 (seconds in 5m)

irate() - based on the last two data points. irate should only be used when graphing volatile, fast-moving counters => over 1m no sense

increase() - the increase in the time series in the range vector. increase(http_requests_total[5m]) - the number of HTTP requests as measured over the last 5 minutes

resets() - the number of counter resets within the provided time range as an instant vector

Gauge​

  • Can either go up or down
  • Common fuctions: delta(), deriv()
  • Examples: number of pods in a cluster, number of events in an queue

delta() - the difference between the first and last value of each time series element in a range vector

deriv() - the per-second derivative of the time series in a range vector. Similar for rate(), but for gauges

Histogram​

  • Data based on buckets
  • Common fuctions: histogram_quantile()
  • Example: Latency

histogram_quantile() - counts quantile from Histogram metrics, can be combined with rate()

# count 0.9 quantile
histogram_quantile(0.9, sum(rate(request_duration_milliseconds_bucket[1m])) by (le,path))

# count bucket % over all
1 -
(
  sum(rate(request_duration_milliseconds_bucket{le="$slow_response_ms"}[1m])) by (path)
/
  sum(rate(request_duration_milliseconds_count[1m])) by (path)
)
> 0