go-pulse/metrics/log.go
turboboost55 7dc100714d
metrics: add cpu counters (#26796)
This PR adds counter metrics for the CPU system and the Geth process.
Currently the only metrics available for these items are gauges. Gauges are
fine when the consumer scrapes metrics data at the same interval as Geth
produces new values (every 3 seconds), but it is likely that most consumers
will not scrape that often. Intervals of 10, 15, or maybe even 30 seconds
are probably more common.

So the problem is, how does the consumer estimate what the CPU was doing in
between scrapes. With a counter, it's easy ... you just subtract two
successive values and divide by the time to get a nice, accurate average.
But with a gauge, you can't do that. A gauge reading is an instantaneous
picture of what was happening at that moment, but it gives you no idea
about what was going on between scrapes. Taking an average of values is
meaningless.
2023-03-23 14:13:50 +01:00

84 lines
3.0 KiB
Go

package metrics
import (
"time"
)
type Logger interface {
Printf(format string, v ...interface{})
}
func Log(r Registry, freq time.Duration, l Logger) {
LogScaled(r, freq, time.Nanosecond, l)
}
// Output each metric in the given registry periodically using the given
// logger. Print timings in `scale` units (eg time.Millisecond) rather than nanos.
func LogScaled(r Registry, freq time.Duration, scale time.Duration, l Logger) {
du := float64(scale)
duSuffix := scale.String()[1:]
for range time.Tick(freq) {
r.Each(func(name string, i interface{}) {
switch metric := i.(type) {
case Counter:
l.Printf("counter %s\n", name)
l.Printf(" count: %9d\n", metric.Count())
case CounterFloat64:
l.Printf("counter %s\n", name)
l.Printf(" count: %f\n", metric.Count())
case Gauge:
l.Printf("gauge %s\n", name)
l.Printf(" value: %9d\n", metric.Value())
case GaugeFloat64:
l.Printf("gauge %s\n", name)
l.Printf(" value: %f\n", metric.Value())
case Healthcheck:
metric.Check()
l.Printf("healthcheck %s\n", name)
l.Printf(" error: %v\n", metric.Error())
case Histogram:
h := metric.Snapshot()
ps := h.Percentiles([]float64{0.5, 0.75, 0.95, 0.99, 0.999})
l.Printf("histogram %s\n", name)
l.Printf(" count: %9d\n", h.Count())
l.Printf(" min: %9d\n", h.Min())
l.Printf(" max: %9d\n", h.Max())
l.Printf(" mean: %12.2f\n", h.Mean())
l.Printf(" stddev: %12.2f\n", h.StdDev())
l.Printf(" median: %12.2f\n", ps[0])
l.Printf(" 75%%: %12.2f\n", ps[1])
l.Printf(" 95%%: %12.2f\n", ps[2])
l.Printf(" 99%%: %12.2f\n", ps[3])
l.Printf(" 99.9%%: %12.2f\n", ps[4])
case Meter:
m := metric.Snapshot()
l.Printf("meter %s\n", name)
l.Printf(" count: %9d\n", m.Count())
l.Printf(" 1-min rate: %12.2f\n", m.Rate1())
l.Printf(" 5-min rate: %12.2f\n", m.Rate5())
l.Printf(" 15-min rate: %12.2f\n", m.Rate15())
l.Printf(" mean rate: %12.2f\n", m.RateMean())
case Timer:
t := metric.Snapshot()
ps := t.Percentiles([]float64{0.5, 0.75, 0.95, 0.99, 0.999})
l.Printf("timer %s\n", name)
l.Printf(" count: %9d\n", t.Count())
l.Printf(" min: %12.2f%s\n", float64(t.Min())/du, duSuffix)
l.Printf(" max: %12.2f%s\n", float64(t.Max())/du, duSuffix)
l.Printf(" mean: %12.2f%s\n", t.Mean()/du, duSuffix)
l.Printf(" stddev: %12.2f%s\n", t.StdDev()/du, duSuffix)
l.Printf(" median: %12.2f%s\n", ps[0]/du, duSuffix)
l.Printf(" 75%%: %12.2f%s\n", ps[1]/du, duSuffix)
l.Printf(" 95%%: %12.2f%s\n", ps[2]/du, duSuffix)
l.Printf(" 99%%: %12.2f%s\n", ps[3]/du, duSuffix)
l.Printf(" 99.9%%: %12.2f%s\n", ps[4]/du, duSuffix)
l.Printf(" 1-min rate: %12.2f\n", t.Rate1())
l.Printf(" 5-min rate: %12.2f\n", t.Rate5())
l.Printf(" 15-min rate: %12.2f\n", t.Rate15())
l.Printf(" mean rate: %12.2f\n", t.RateMean())
}
})
}
}