nebula/stats.go
Wade Simmons 06372e12f1 stats: add ability to set static prometheus labels
This changes lets you define static Prometheus labels in the config.
This can be useful for tagging your stats with the custom labels you
need for your environment.

    stats:
      type: prometheus
      labels:
        myLabelOne: value1
        myLabelTwo: value2
2024-03-27 14:37:23 -04:00

139 lines
3.9 KiB
Go

package nebula
import (
"errors"
"fmt"
"log"
"net"
"net/http"
"runtime"
"strconv"
"time"
graphite "github.com/cyberdelia/go-metrics-graphite"
mp "github.com/nbrownus/go-metrics-prometheus"
"github.com/prometheus/client_golang/prometheus"
"github.com/prometheus/client_golang/prometheus/promhttp"
"github.com/rcrowley/go-metrics"
"github.com/sirupsen/logrus"
"github.com/slackhq/nebula/config"
)
// startStats initializes stats from config. On success, if any further work
// is needed to serve stats, it returns a func to handle that work. If no
// work is needed, it'll return nil. On failure, it returns nil, error.
func startStats(l *logrus.Logger, c *config.C, buildVersion string, configTest bool) (func(), error) {
mType := c.GetString("stats.type", "")
if mType == "" || mType == "none" {
return nil, nil
}
interval := c.GetDuration("stats.interval", 0)
if interval == 0 {
return nil, fmt.Errorf("stats.interval was an invalid duration: %s", c.GetString("stats.interval", ""))
}
var startFn func()
switch mType {
case "graphite":
err := startGraphiteStats(l, interval, c, configTest)
if err != nil {
return nil, err
}
case "prometheus":
var err error
startFn, err = startPrometheusStats(l, interval, c, buildVersion, configTest)
if err != nil {
return nil, err
}
default:
return nil, fmt.Errorf("stats.type was not understood: %s", mType)
}
metrics.RegisterDebugGCStats(metrics.DefaultRegistry)
metrics.RegisterRuntimeMemStats(metrics.DefaultRegistry)
go metrics.CaptureDebugGCStats(metrics.DefaultRegistry, interval)
go metrics.CaptureRuntimeMemStats(metrics.DefaultRegistry, interval)
return startFn, nil
}
func startGraphiteStats(l *logrus.Logger, i time.Duration, c *config.C, configTest bool) error {
proto := c.GetString("stats.protocol", "tcp")
host := c.GetString("stats.host", "")
if host == "" {
return errors.New("stats.host can not be empty")
}
prefix := c.GetString("stats.prefix", "nebula")
addr, err := net.ResolveTCPAddr(proto, host)
if err != nil {
return fmt.Errorf("error while setting up graphite sink: %s", err)
}
if !configTest {
l.Infof("Starting graphite. Interval: %s, prefix: %s, addr: %s", i, prefix, addr)
go graphite.Graphite(metrics.DefaultRegistry, i, prefix, addr)
}
return nil
}
func startPrometheusStats(l *logrus.Logger, i time.Duration, c *config.C, buildVersion string, configTest bool) (func(), error) {
namespace := c.GetString("stats.namespace", "")
subsystem := c.GetString("stats.subsystem", "")
listen := c.GetString("stats.listen", "")
if listen == "" {
return nil, fmt.Errorf("stats.listen should not be empty")
}
path := c.GetString("stats.path", "")
if path == "" {
return nil, fmt.Errorf("stats.path should not be empty")
}
pry := prometheus.NewRegistry()
var prr prometheus.Registerer = pry
labelsRaw := c.GetMap("stats.labels", nil)
if labelsRaw != nil {
labels := prometheus.Labels{}
for k, v := range labelsRaw {
labels[fmt.Sprintf("%v", k)] = fmt.Sprintf("%v", v)
}
prr = prometheus.WrapRegistererWith(labels, prr)
}
pClient := mp.NewPrometheusProvider(metrics.DefaultRegistry, namespace, subsystem, prr, i)
if !configTest {
go pClient.UpdatePrometheusMetrics()
}
// Export our version information as labels on a static gauge
g := prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: namespace,
Subsystem: subsystem,
Name: "info",
Help: "Version information for the Nebula binary",
ConstLabels: prometheus.Labels{
"version": buildVersion,
"goversion": runtime.Version(),
"boringcrypto": strconv.FormatBool(boringEnabled()),
},
})
prr.MustRegister(g)
g.Set(1)
var startFn func()
if !configTest {
startFn = func() {
l.Infof("Prometheus stats listening on %s at %s", listen, path)
http.Handle(path, promhttp.HandlerFor(pry, promhttp.HandlerOpts{ErrorLog: l}))
log.Fatal(http.ListenAndServe(listen, nil))
}
}
return startFn, nil
}