Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

metrics: add cpu counters #26796

Merged
merged 9 commits into from
Mar 23, 2023
153 changes: 153 additions & 0 deletions metrics/counter_float64.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,153 @@
package metrics

import (
"sync"
)

// CounterFloat64 holds a float64 value that can be incremented and decremented.
type CounterFloat64 interface {
Clear()
Value() float64
Dec(float64)
Inc(float64)
Snapshot() CounterFloat64
}

// GetOrRegisterCounterFloat64 returns an existing CounterFloat64 or constructs and registers
// a new StandardCounterFloat64.
func GetOrRegisterCounterFloat64(name string, r Registry) CounterFloat64 {
if nil == r {
r = DefaultRegistry
}
return r.GetOrRegister(name, NewCounterFloat64).(CounterFloat64)
}

// GetOrRegisterCounterFloat64Forced returns an existing CounterFloat64 or constructs and registers a
// new CounterFloat64 no matter the global switch is enabled or not.
// Be sure to unregister the counter from the registry once it is of no use to
// allow for garbage collection.
func GetOrRegisterCounterFloat64Forced(name string, r Registry) CounterFloat64 {
if nil == r {
r = DefaultRegistry
}
return r.GetOrRegister(name, NewCounterFloat64Forced).(CounterFloat64)
}

// NewCounterFloat64 constructs a new StandardCounterFloat64.
func NewCounterFloat64() CounterFloat64 {
if !Enabled {
return NilCounterFloat64{}
}
return &StandardCounterFloat64{value: 0.0}
}

// NewCounterFloat64Forced constructs a new StandardCounterFloat64 and returns it no matter if
// the global switch is enabled or not.
func NewCounterFloat64Forced() CounterFloat64 {
return &StandardCounterFloat64{value: 0.0}
}

// NewRegisteredCounterFloat64 constructs and registers a new StandardCounterFloat64.
func NewRegisteredCounterFloat64(name string, r Registry) CounterFloat64 {
c := NewCounterFloat64()
if nil == r {
r = DefaultRegistry
}
r.Register(name, c)
return c
}

// NewRegisteredCounterFloat64Forced constructs and registers a new StandardCounterFloat64
// and launches a goroutine no matter the global switch is enabled or not.
// Be sure to unregister the counter from the registry once it is of no use to
// allow for garbage collection.
func NewRegisteredCounterFloat64Forced(name string, r Registry) CounterFloat64 {
c := NewCounterFloat64Forced()
if nil == r {
r = DefaultRegistry
}
r.Register(name, c)
return c
}

// CounterFloat64Snapshot is a read-only copy of another CounterFloat64.
type CounterFloat64Snapshot float64

// Clear panics.
func (CounterFloat64Snapshot) Clear() {
panic("Clear called on a CounterFloat64Snapshot")
}

// Value returns the value at the time the snapshot was taken.
func (c CounterFloat64Snapshot) Value() float64 { return float64(c) }

// Dec panics.
func (CounterFloat64Snapshot) Dec(float64) {
panic("Dec called on a CounterFloat64Snapshot")
}

// Inc panics.
func (CounterFloat64Snapshot) Inc(float64) {
panic("Inc called on a CounterFloat64Snapshot")
}

// Snapshot returns the snapshot.
func (c CounterFloat64Snapshot) Snapshot() CounterFloat64 { return c }

// NilCounterFloat64 is a no-op CounterFloat64.
type NilCounterFloat64 struct{}

// Clear is a no-op.
func (NilCounterFloat64) Clear() {}

// Value is a no-op.
func (NilCounterFloat64) Value() float64 { return 0.0 }

// Dec is a no-op.
func (NilCounterFloat64) Dec(i float64) {}

// Inc is a no-op.
func (NilCounterFloat64) Inc(i float64) {}

// Snapshot is a no-op.
func (NilCounterFloat64) Snapshot() CounterFloat64 { return NilCounterFloat64{} }

// StandardCounterFloat64 is the standard implementation of a CounterFloat64 and uses the
// sync.Mutex package to manage a single float64 value.
type StandardCounterFloat64 struct {
mutex sync.Mutex
value float64
}

// Clear sets the counter to zero.
func (c *StandardCounterFloat64) Clear() {
c.mutex.Lock()
defer c.mutex.Unlock()
c.value = 0.0
}

// Value returns the current value.
func (c *StandardCounterFloat64) Value() float64 {
c.mutex.Lock()
defer c.mutex.Unlock()
return c.value
}

// Dec decrements the counter by the given amount.
func (c *StandardCounterFloat64) Dec(v float64) {
c.mutex.Lock()
defer c.mutex.Unlock()
c.value -= v
}

// Inc increments the counter by the given amount.
func (c *StandardCounterFloat64) Inc(v float64) {
c.mutex.Lock()
defer c.mutex.Unlock()
c.value += v
}

// Snapshot returns a read-only copy of the counter.
func (c *StandardCounterFloat64) Snapshot() CounterFloat64 {
return CounterFloat64Snapshot(c.Value())
}
77 changes: 77 additions & 0 deletions metrics/counter_float_64_test.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,77 @@
package metrics

import "testing"

func BenchmarkCounterFloat64(b *testing.B) {
c := NewCounterFloat64()
b.ResetTimer()
for i := 0; i < b.N; i++ {
c.Inc(1.0)
}
}

func TestCounterFloat64Clear(t *testing.T) {
c := NewCounterFloat64()
c.Inc(1.0)
c.Clear()
if count := c.Value(); count != 0 {
t.Errorf("c.Value(): 0 != %v\n", count)
}
}

func TestCounterFloat64Dec1(t *testing.T) {
c := NewCounterFloat64()
c.Dec(1.0)
if count := c.Value(); count != -1.0 {
t.Errorf("c.Value(): -1.0 != %v\n", count)
}
}

func TestCounterFloat64Dec2(t *testing.T) {
c := NewCounterFloat64()
c.Dec(2.0)
if count := c.Value(); count != -2.0 {
t.Errorf("c.Value(): -2.0 != %v\n", count)
}
}

func TestCounterFloat64Inc1(t *testing.T) {
c := NewCounterFloat64()
c.Inc(1.0)
if count := c.Value(); count != 1.0 {
t.Errorf("c.Value(): 1.0 != %v\n", count)
}
}

func TestCounterFloat64Inc2(t *testing.T) {
c := NewCounterFloat64()
c.Inc(2.0)
if count := c.Value(); count != 2.0 {
t.Errorf("c.Value(): 2.0 != %v\n", count)
}
}

func TestCounterFloat64Snapshot(t *testing.T) {
c := NewCounterFloat64()
c.Inc(1.0)
snapshot := c.Snapshot()
c.Inc(1.0)
if count := snapshot.Value(); count != 1.0 {
t.Errorf("c.Value(): 1.0 != %v\n", count)
}
}

func TestCounterFloat64Zero(t *testing.T) {
c := NewCounterFloat64()
if count := c.Value(); count != 0 {
t.Errorf("c.Value(): 0 != %v\n", count)
}
}

func TestGetOrRegisterCounterFloat64(t *testing.T) {
r := NewRegistry()
NewRegisteredCounterFloat64("foo", r).Inc(47.0)
if c := GetOrRegisterCounterFloat64("foo", r); c.Value() != 47.0 {
t.Fatal(c)
}
}
7 changes: 7 additions & 0 deletions metrics/exp/exp.go
Original file line number Diff line number Diff line change
Expand Up @@ -100,6 +100,11 @@ func (exp *exp) publishCounter(name string, metric metrics.Counter) {
v.Set(metric.Count())
}

func (exp *exp) publishCounterFloat64(name string, metric metrics.CounterFloat64) {
v := exp.getFloat(name)
v.Set(metric.Value())
}

func (exp *exp) publishGauge(name string, metric metrics.Gauge) {
v := exp.getInt(name)
v.Set(metric.Value())
Expand Down Expand Up @@ -167,6 +172,8 @@ func (exp *exp) syncToExpvar() {
switch i := i.(type) {
case metrics.Counter:
exp.publishCounter(name, i)
case metrics.CounterFloat64:
exp.publishCounterFloat64(name, i)
case metrics.Gauge:
exp.publishGauge(name, i)
case metrics.GaugeFloat64:
Expand Down
2 changes: 2 additions & 0 deletions metrics/graphite.go
Original file line number Diff line number Diff line change
Expand Up @@ -67,6 +67,8 @@ func graphite(c *GraphiteConfig) error {
switch metric := i.(type) {
case Counter:
fmt.Fprintf(w, "%s.%s.count %d %d\n", c.Prefix, name, metric.Count(), now)
case CounterFloat64:
fmt.Fprintf(w, "%s.%s.value %f %d\n", c.Prefix, name, metric.Value(), now)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It should be called count?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Well, as I mentioned above, I decided to rename the value field to Value. Are you guys ok with that?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it should be called count in exported metrics because it's a counter. The internal method name can be Value or Count, it doesn't matter that much to me.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I could always change it back.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'll change it back to count. It is better to be consistent with the existing int64 counter.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@fjl Hey, I hope you don't think I was beig a jerk when I said "I could always change it back" right after you explicitly said that is what I should do. Maybe it came across that way. What happened was, your comment of "It should be called count ..." hadn't come up on my screen yet for some reason, so I was just kind of talking into the void. When I finally saw your comment, I was like "Oh oh. My comment probably didn't come across very well!"

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi, please don't worry! I think it's great that you are contributing this code. The reason why it's not merged yet, is that I wanted to test it on my node, but haven't gotten around to that yet.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ok, no rush.

case Gauge:
fmt.Fprintf(w, "%s.%s.value %d %d\n", c.Prefix, name, metric.Value(), now)
case GaugeFloat64:
Expand Down
10 changes: 10 additions & 0 deletions metrics/influxdb/influxdb.go
Original file line number Diff line number Diff line change
Expand Up @@ -141,6 +141,16 @@ func (r *reporter) send() error {
},
Time: now,
})
case metrics.CounterFloat64:
value := metric.Value()
pts = append(pts, client.Point{
Measurement: fmt.Sprintf("%s%s.value", namespace, name),
Tags: r.tags,
Fields: map[string]interface{}{
"value": value,
},
Time: now,
})
case metrics.Gauge:
ms := metric.Snapshot()
pts = append(pts, client.Point{
Expand Down
17 changes: 11 additions & 6 deletions metrics/influxdb/influxdbv2.go
Original file line number Diff line number Diff line change
Expand Up @@ -24,8 +24,6 @@ type v2Reporter struct {

client influxdb2.Client
write api.WriteAPI

cache map[string]int64
}

// InfluxDBWithTags starts a InfluxDB reporter which will post the from the given metrics.Registry at each d interval with the specified tags
Expand All @@ -39,7 +37,6 @@ func InfluxDBV2WithTags(r metrics.Registry, d time.Duration, endpoint string, to
organization: organization,
namespace: namespace,
tags: tags,
cache: make(map[string]int64),
}

rep.client = influxdb2.NewClient(rep.endpoint, rep.token)
Expand Down Expand Up @@ -86,17 +83,25 @@ func (r *v2Reporter) send() {
switch metric := i.(type) {
case metrics.Counter:
v := metric.Count()
l := r.cache[name]

measurement := fmt.Sprintf("%s%s.count", namespace, name)
fields := map[string]interface{}{
"value": v - l,
"value": v,
}

pt := influxdb2.NewPoint(measurement, r.tags, fields, now)
r.write.WritePoint(pt)

r.cache[name] = v
case metrics.CounterFloat64:
v := metric.Value()

measurement := fmt.Sprintf("%s%s.value", namespace, name)
fields := map[string]interface{}{
"value": v,
}

pt := influxdb2.NewPoint(measurement, r.tags, fields, now)
r.write.WritePoint(pt)

case metrics.Gauge:
ms := metric.Snapshot()
Expand Down
11 changes: 11 additions & 0 deletions metrics/librato/librato.go
Original file line number Diff line number Diff line change
Expand Up @@ -107,6 +107,17 @@ func (rep *Reporter) BuildRequest(now time.Time, r metrics.Registry) (snapshot B
}
snapshot.Counters = append(snapshot.Counters, measurement)
}
case metrics.CounterFloat64:
if m.Value() > 0 {
measurement[Name] = fmt.Sprintf("%s.%s", name, "value")
measurement[Value] = m.Value()
measurement[Attributes] = map[string]interface{}{
DisplayUnitsLong: Operations,
DisplayUnitsShort: OperationsShort,
DisplayMin: "0",
}
snapshot.Counters = append(snapshot.Counters, measurement)
}
case metrics.Gauge:
measurement[Name] = name
measurement[Value] = float64(m.Value())
Expand Down
3 changes: 3 additions & 0 deletions metrics/log.go
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,9 @@ func LogScaled(r Registry, freq time.Duration, scale time.Duration, l Logger) {
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(" value: %f\n", metric.Value())
case Gauge:
l.Printf("gauge %s\n", name)
l.Printf(" value: %9d\n", metric.Value())
Expand Down
19 changes: 13 additions & 6 deletions metrics/metrics.go
Original file line number Diff line number Diff line change
Expand Up @@ -144,6 +144,9 @@ func CollectProcessMetrics(refresh time.Duration) {
cpuSysLoad = GetOrRegisterGauge("system/cpu/sysload", DefaultRegistry)
cpuSysWait = GetOrRegisterGauge("system/cpu/syswait", DefaultRegistry)
cpuProcLoad = GetOrRegisterGauge("system/cpu/procload", DefaultRegistry)
cpuSysLoadTotal = GetOrRegisterCounterFloat64("system/cpu/sysload/total", DefaultRegistry)
cpuSysWaitTotal = GetOrRegisterCounterFloat64("system/cpu/syswait/total", DefaultRegistry)
cpuProcLoadTotal = GetOrRegisterCounterFloat64("system/cpu/procload/total", DefaultRegistry)
cpuThreads = GetOrRegisterGauge("system/cpu/threads", DefaultRegistry)
cpuGoroutines = GetOrRegisterGauge("system/cpu/goroutines", DefaultRegistry)
cpuSchedLatency = getOrRegisterRuntimeHistogram("system/cpu/schedlatency", secondsToNs, nil)
Expand Down Expand Up @@ -172,13 +175,17 @@ func CollectProcessMetrics(refresh time.Duration) {
secondsSinceLastCollect := collectTime.Sub(lastCollectTime).Seconds()
lastCollectTime = collectTime
if secondsSinceLastCollect > 0 {
sysLoad := (cpustats[now].GlobalTime - cpustats[prev].GlobalTime) / secondsSinceLastCollect
sysWait := (cpustats[now].GlobalWait - cpustats[prev].GlobalWait) / secondsSinceLastCollect
procLoad := (cpustats[now].LocalTime - cpustats[prev].LocalTime) / secondsSinceLastCollect
sysLoad := cpustats[now].GlobalTime - cpustats[prev].GlobalTime
sysWait := cpustats[now].GlobalWait - cpustats[prev].GlobalWait
procLoad := cpustats[now].LocalTime - cpustats[prev].LocalTime
// Convert to integer percentage.
cpuSysLoad.Update(int64(sysLoad * 100))
cpuSysWait.Update(int64(sysWait * 100))
cpuProcLoad.Update(int64(procLoad * 100))
cpuSysLoad.Update(int64(sysLoad / secondsSinceLastCollect * 100))
cpuSysWait.Update(int64(sysWait / secondsSinceLastCollect * 100))
cpuProcLoad.Update(int64(procLoad / secondsSinceLastCollect * 100))
// increment counters (ms)
cpuSysLoadTotal.Inc(sysLoad)
cpuSysWaitTotal.Inc(sysWait)
cpuProcLoadTotal.Inc(procLoad)
Comment on lines +186 to +188
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Shouldn't these be divided by the time since last collect?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No they should not.

The gauges contain the fraction (percentage) of time in which the CPU was active.

The counters contain the total CPU time in seconds.

}

// Threads
Expand Down
Loading