Merge pull request #16 from prometheus/migration/next-client-golang

Migration/next client golang
This commit is contained in:
Björn Rabenstein 2014-06-30 17:23:41 +02:00
commit eb17b5fc9d
14 changed files with 459 additions and 302 deletions

View file

@ -8,12 +8,11 @@ import (
)
var (
attributes = prometheus.NewGauge()
attributes *prometheus.GaugeVec
)
type attributesCollector struct {
registry prometheus.Registry
config Config
config Config
}
func init() {
@ -22,22 +21,33 @@ func init() {
// Takes a config struct and prometheus registry and returns a new Collector exposing
// labels from the config.
func NewAttributesCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewAttributesCollector(config Config) (Collector, error) {
c := attributesCollector{
config: config,
registry: registry,
config: config,
}
registry.Register(
"node_attributes",
"node_exporter attributes",
prometheus.NilLabels,
attributes,
labelNames := []string{}
for l := range c.config.Attributes {
labelNames = append(labelNames, l)
}
gv := prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Name: "attributes",
Help: "The node_exporter attributes.",
},
labelNames,
)
collector, err := prometheus.RegisterOrGet(gv)
if err != nil {
return nil, err
}
attributes = collector.(*prometheus.GaugeVec)
return &c, nil
}
func (c *attributesCollector) Update() (updates int, err error) {
glog.V(1).Info("Set node_attributes{%v}: 1", c.config.Attributes)
attributes.Set(c.config.Attributes, 1)
attributes.Reset()
attributes.With(c.config.Attributes).Set(1)
return updates, err
}

View file

@ -1,11 +1,9 @@
// Exporter is a prometheus exporter using multiple Factories to collect and export system metrics.
package collector
import (
"github.com/prometheus/client_golang/prometheus"
)
const Namespace = "node"
var Factories = make(map[string]func(Config, prometheus.Registry) (Collector, error))
var Factories = make(map[string]func(Config) (Collector, error))
// Interface a collector has to implement.
type Collector interface {
@ -13,6 +11,12 @@ type Collector interface {
Update() (n int, err error)
}
// TODO: Instead of periodically call Update, a Collector could be implemented
// as a real prometheus.Collector that only gathers metrics when
// scraped. (However, for metric gathering that takes very long, it might
// actually be better to do them proactively before scraping to minimize scrape
// time.)
type Config struct {
Attributes map[string]string `json:"attributes"`
}

View file

@ -18,35 +18,119 @@ import (
const (
procDiskStats = "/proc/diskstats"
diskSubsystem = "disk"
)
type diskStat struct {
name string
metric prometheus.Metric
documentation string
}
var (
ignoredDevices = flag.String("diskstatsIgnoredDevices", "^(ram|loop|(h|s|xv)d[a-z])\\d+$", "Regexp of devices to ignore for diskstats.")
diskLabelNames = []string{"device"}
// Docs from https://www.kernel.org/doc/Documentation/iostats.txt
diskStatsMetrics = []diskStat{
{"reads_completed", prometheus.NewCounter(), "The total number of reads completed successfully."},
{"reads_merged", prometheus.NewCounter(), "The number of reads merged. See https://www.kernel.org/doc/Documentation/iostats.txt"},
{"sectors_read", prometheus.NewCounter(), "The total number of sectors read successfully."},
{"read_time_ms", prometheus.NewCounter(), "the total number of milliseconds spent by all reads."},
{"writes_completed", prometheus.NewCounter(), "The total number of writes completed successfully."},
{"writes_merged", prometheus.NewCounter(), "The number of writes merged. See https://www.kernel.org/doc/Documentation/iostats.txt"},
{"sectors_written", prometheus.NewCounter(), "The total number of sectors written successfully."},
{"write_time_ms", prometheus.NewCounter(), "This is the total number of milliseconds spent by all writes."},
{"io_now", prometheus.NewGauge(), "The number of I/Os currently in progress."},
{"io_time_ms", prometheus.NewCounter(), "Milliseconds spent doing I/Os."},
{"io_time_weighted", prometheus.NewCounter(), "The weighted # of milliseconds spent doing I/Os. See https://www.kernel.org/doc/Documentation/iostats.txt"},
diskStatsMetrics = []prometheus.Collector{
prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "reads_completed",
Help: "The total number of reads completed successfully.",
},
diskLabelNames,
),
prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "reads_merged",
Help: "The number of reads merged. See https://www.kernel.org/doc/Documentation/iostats.txt.",
},
diskLabelNames,
),
prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "sectors_read",
Help: "The total number of sectors read successfully.",
},
diskLabelNames,
),
prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "read_time_ms",
Help: "The total number of milliseconds spent by all reads.",
},
diskLabelNames,
),
prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "writes_completed",
Help: "The total number of writes completed successfully.",
},
diskLabelNames,
),
prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "writes_merged",
Help: "The number of writes merged. See https://www.kernel.org/doc/Documentation/iostats.txt.",
},
diskLabelNames,
),
prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "sectors_written",
Help: "The total number of sectors written successfully.",
},
diskLabelNames,
),
prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "write_time_ms",
Help: "This is the total number of milliseconds spent by all writes.",
},
diskLabelNames,
),
prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "io_now",
Help: "The number of I/Os currently in progress.",
},
diskLabelNames,
),
prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "io_time_ms",
Help: "Milliseconds spent doing I/Os.",
},
diskLabelNames,
),
prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Subsystem: diskSubsystem,
Name: "io_time_weighted",
Help: "The weighted # of milliseconds spent doing I/Os. See https://www.kernel.org/doc/Documentation/iostats.txt.",
},
diskLabelNames,
),
}
)
type diskstatsCollector struct {
registry prometheus.Registry
config Config
ignoredDevicesPattern *regexp.Regexp
}
@ -57,20 +141,16 @@ func init() {
// Takes a config struct and prometheus registry and returns a new Collector exposing
// disk device stats.
func NewDiskstatsCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewDiskstatsCollector(config Config) (Collector, error) {
c := diskstatsCollector{
config: config,
registry: registry,
ignoredDevicesPattern: regexp.MustCompile(*ignoredDevices),
}
for _, v := range diskStatsMetrics {
registry.Register(
"node_disk_"+v.name,
v.documentation,
prometheus.NilLabels,
v.metric,
)
for _, c := range diskStatsMetrics {
if _, err := prometheus.RegisterOrGet(c); err != nil {
return nil, err
}
}
return &c, nil
}
@ -91,13 +171,12 @@ func (c *diskstatsCollector) Update() (updates int, err error) {
if err != nil {
return updates, fmt.Errorf("Invalid value %s in diskstats: %s", value, err)
}
labels := map[string]string{"device": dev}
counter, ok := diskStatsMetrics[k].metric.(prometheus.Counter)
counter, ok := diskStatsMetrics[k].(*prometheus.CounterVec)
if ok {
counter.Set(labels, v)
counter.WithLabelValues(dev).Set(v)
} else {
var gauge = diskStatsMetrics[k].metric.(prometheus.Gauge)
gauge.Set(labels, v)
var gauge = diskStatsMetrics[k].(*prometheus.GaugeVec)
gauge.WithLabelValues(dev).Set(v)
}
}
}

View file

@ -11,26 +11,68 @@ import (
"strings"
"syscall"
"github.com/golang/glog"
"github.com/golang/glog"
"github.com/prometheus/client_golang/prometheus"
)
const (
procMounts = "/proc/mounts"
procMounts = "/proc/mounts"
filesystemSubsystem = "filesystem"
)
var (
fsSizeMetric = prometheus.NewGauge()
fsFreeMetric = prometheus.NewGauge()
fsAvailMetric = prometheus.NewGauge()
fsFilesMetric = prometheus.NewGauge()
fsFilesFreeMetric = prometheus.NewGauge()
filesystemLabelNames = []string{"filesystem"}
fsSizeMetric = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Subsystem: filesystemSubsystem,
Name: "size",
Help: "Filesystem size in bytes.",
},
filesystemLabelNames,
)
fsFreeMetric = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Subsystem: filesystemSubsystem,
Name: "free",
Help: "Filesystem free space in bytes.",
},
filesystemLabelNames,
)
fsAvailMetric = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Subsystem: filesystemSubsystem,
Name: "avail",
Help: "Filesystem space available to non-root users in bytes.",
},
filesystemLabelNames,
)
fsFilesMetric = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Subsystem: filesystemSubsystem,
Name: "files",
Help: "Filesystem total file nodes.",
},
filesystemLabelNames,
)
fsFilesFreeMetric = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Subsystem: filesystemSubsystem,
Name: "files_free",
Help: "Filesystem total free file nodes.",
},
filesystemLabelNames,
)
ignoredMountPoints = flag.String("filesystemIgnoredMountPoints", "^/(sys|proc|dev)($|/)", "Regexp of mount points to ignore for filesystem collector.")
)
type filesystemCollector struct {
registry prometheus.Registry
config Config
ignoredMountPointsPattern *regexp.Regexp
}
@ -41,42 +83,26 @@ func init() {
// Takes a config struct and prometheus registry and returns a new Collector exposing
// network device filesystems.
func NewFilesystemCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewFilesystemCollector(config Config) (Collector, error) {
c := filesystemCollector{
config: config,
registry: registry,
config: config,
ignoredMountPointsPattern: regexp.MustCompile(*ignoredMountPoints),
}
registry.Register(
"node_filesystem_size",
"Filesystem size in bytes.",
prometheus.NilLabels,
fsSizeMetric,
)
registry.Register(
"node_filesystem_free",
"Filesystem free space in bytes.",
prometheus.NilLabels,
fsFreeMetric,
)
registry.Register(
"node_filesystem_avail",
"Filesystem space available to non-root users in bytes.",
prometheus.NilLabels,
fsAvailMetric,
)
registry.Register(
"node_filesystem_files",
"Filesystem total file nodes.",
prometheus.NilLabels,
fsFilesMetric,
)
registry.Register(
"node_filesystem_files_free",
"Filesystem total free file nodes.",
prometheus.NilLabels,
fsFilesFreeMetric,
)
if _, err := prometheus.RegisterOrGet(fsSizeMetric); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(fsFreeMetric); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(fsAvailMetric); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(fsFilesMetric); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(fsFilesFreeMetric); err != nil {
return nil, err
}
return &c, nil
}
@ -96,11 +122,11 @@ func (c *filesystemCollector) Update() (updates int, err error) {
if err != nil {
return updates, fmt.Errorf("Statfs on %s returned %s", mp, err)
}
fsSizeMetric.Set(map[string]string{"filesystem": mp}, float64(buf.Blocks)*float64(buf.Bsize))
fsFreeMetric.Set(map[string]string{"filesystem": mp}, float64(buf.Bfree)*float64(buf.Bsize))
fsAvailMetric.Set(map[string]string{"filesystem": mp}, float64(buf.Bavail)*float64(buf.Bsize))
fsFilesMetric.Set(map[string]string{"filesystem": mp}, float64(buf.Files))
fsFilesFreeMetric.Set(map[string]string{"filesystem": mp}, float64(buf.Ffree))
fsSizeMetric.WithLabelValues(mp).Set(float64(buf.Blocks) * float64(buf.Bsize))
fsFreeMetric.WithLabelValues(mp).Set(float64(buf.Bfree) * float64(buf.Bsize))
fsAvailMetric.WithLabelValues(mp).Set(float64(buf.Bavail) * float64(buf.Bsize))
fsFilesMetric.WithLabelValues(mp).Set(float64(buf.Files))
fsFilesFreeMetric.WithLabelValues(mp).Set(float64(buf.Ffree))
updates++
}
return updates, err

View file

@ -24,9 +24,8 @@ const (
)
type gmondCollector struct {
Metrics map[string]prometheus.Gauge
config Config
registry prometheus.Registry
Metrics map[string]*prometheus.GaugeVec
config Config
}
func init() {
@ -36,17 +35,16 @@ func init() {
var illegalCharsRE = regexp.MustCompile(`[^a-zA-Z0-9_]`)
// Takes a config struct and prometheus registry and returns a new Collector scraping ganglia.
func NewGmondCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewGmondCollector(config Config) (Collector, error) {
c := gmondCollector{
config: config,
Metrics: make(map[string]prometheus.Gauge),
registry: registry,
config: config,
Metrics: map[string]*prometheus.GaugeVec{},
}
return &c, nil
}
func (c *gmondCollector) setMetric(name string, labels map[string]string, metric ganglia.Metric) {
func (c *gmondCollector) setMetric(name, cluster string, metric ganglia.Metric) {
if _, ok := c.Metrics[name]; !ok {
var desc string
var title string
@ -62,12 +60,18 @@ func (c *gmondCollector) setMetric(name string, labels map[string]string, metric
}
}
glog.V(1).Infof("Register %s: %s", name, desc)
gauge := prometheus.NewGauge()
c.Metrics[name] = gauge
c.registry.Register(name, desc, prometheus.NilLabels, gauge) // one gauge per metric!
gv := prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: gangliaMetricsPrefix,
Name: name,
Help: desc,
},
[]string{"cluster"},
)
c.Metrics[name] = prometheus.MustRegisterOrGet(gv).(*prometheus.GaugeVec)
}
glog.V(1).Infof("Set %s{%s}: %f", name, labels, metric.Value)
c.Metrics[name].Set(labels, metric.Value)
glog.V(1).Infof("Set %s{cluster=%q}: %f", name, cluster, metric.Value)
c.Metrics[name].WithLabelValues(cluster).Set(metric.Value)
}
func (c *gmondCollector) Update() (updates int, err error) {
@ -91,12 +95,9 @@ func (c *gmondCollector) Update() (updates int, err error) {
for _, host := range cluster.Hosts {
for _, metric := range host.Metrics {
name := gangliaMetricsPrefix + illegalCharsRE.ReplaceAllString(metric.Name, "_")
name := illegalCharsRE.ReplaceAllString(metric.Name, "_")
var labels = map[string]string{
"cluster": cluster.Name,
}
c.setMetric(name, labels, metric)
c.setMetric(name, cluster.Name, metric)
updates++
}
}

View file

@ -18,12 +18,18 @@ const (
)
var (
interruptsMetric = prometheus.NewCounter()
interruptsMetric = prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Name: "interrupts",
Help: "Interrupt details from /proc/interrupts.",
},
[]string{"CPU", "type", "info", "devices"},
)
)
type interruptsCollector struct {
registry prometheus.Registry
config Config
config Config
}
func init() {
@ -32,17 +38,13 @@ func init() {
// Takes a config struct and prometheus registry and returns a new Collector exposing
// interrupts stats
func NewInterruptsCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewInterruptsCollector(config Config) (Collector, error) {
c := interruptsCollector{
config: config,
registry: registry,
config: config,
}
if _, err := prometheus.RegisterOrGet(interruptsMetric); err != nil {
return nil, err
}
registry.Register(
"node_interrupts",
"Interrupt details from /proc/interrupts",
prometheus.NilLabels,
interruptsMetric,
)
return &c, nil
}
@ -58,13 +60,13 @@ func (c *interruptsCollector) Update() (updates int, err error) {
if err != nil {
return updates, fmt.Errorf("Invalid value %s in interrupts: %s", value, err)
}
labels := map[string]string{
labels := prometheus.Labels{
"CPU": strconv.Itoa(cpuNo),
"type": name,
"info": interrupt.info,
"devices": interrupt.devices,
}
interruptsMetric.Set(labels, fv)
interruptsMetric.With(labels).Set(fv)
}
}
return updates, err

View file

@ -14,13 +14,19 @@ import (
"github.com/prometheus/client_golang/prometheus"
)
const lastLoginSubsystem = "last_login"
var (
lastSeen = prometheus.NewGauge()
lastSeen = prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: Namespace,
Subsystem: lastLoginSubsystem,
Name: "time",
Help: "The time of the last login.",
})
)
type lastLoginCollector struct {
registry prometheus.Registry
config Config
config Config
}
func init() {
@ -29,17 +35,13 @@ func init() {
// Takes a config struct and prometheus registry and returns a new Collector exposing
// load, seconds since last login and a list of tags as specified by config.
func NewLastLoginCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewLastLoginCollector(config Config) (Collector, error) {
c := lastLoginCollector{
config: config,
registry: registry,
config: config,
}
if _, err := prometheus.RegisterOrGet(lastSeen); err != nil {
return nil, err
}
registry.Register(
"node_last_login_time",
"The time of the last login.",
prometheus.NilLabels,
lastSeen,
)
return &c, nil
}
@ -50,7 +52,7 @@ func (c *lastLoginCollector) Update() (updates int, err error) {
}
updates++
glog.V(1).Infof("Set node_last_login_time: %f", last)
lastSeen.Set(nil, last)
lastSeen.Set(last)
return updates, err
}

View file

@ -17,12 +17,15 @@ const (
)
var (
load1 = prometheus.NewGauge()
load1 = prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: Namespace,
Name: "load1",
Help: "1m load average.",
})
)
type loadavgCollector struct {
registry prometheus.Registry
config Config
config Config
}
func init() {
@ -31,18 +34,14 @@ func init() {
// Takes a config struct and prometheus registry and returns a new Collector exposing
// load, seconds since last login and a list of tags as specified by config.
func NewLoadavgCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewLoadavgCollector(config Config) (Collector, error) {
c := loadavgCollector{
config: config,
registry: registry,
config: config,
}
registry.Register(
"node_load1",
"1m load average",
prometheus.NilLabels,
load1,
)
if _, err := prometheus.RegisterOrGet(load1); err != nil {
return nil, err
}
return &c, nil
}
@ -53,7 +52,7 @@ func (c *loadavgCollector) Update() (updates int, err error) {
}
updates++
glog.V(1).Infof("Set node_load: %f", load)
load1.Set(nil, load)
load1.Set(load)
return updates, err
}

View file

@ -16,7 +16,8 @@ import (
)
const (
procMemInfo = "/proc/meminfo"
procMemInfo = "/proc/meminfo"
memInfoSubsystem = "memory"
)
var (
@ -24,8 +25,7 @@ var (
)
type meminfoCollector struct {
registry prometheus.Registry
config Config
config Config
}
func init() {
@ -34,10 +34,9 @@ func init() {
// Takes a config struct and prometheus registry and returns a new Collector exposing
// memory stats.
func NewMeminfoCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewMeminfoCollector(config Config) (Collector, error) {
c := meminfoCollector{
config: config,
registry: registry,
config: config,
}
return &c, nil
}
@ -50,16 +49,16 @@ func (c *meminfoCollector) Update() (updates int, err error) {
glog.V(1).Infof("Set node_mem: %#v", memInfo)
for k, v := range memInfo {
if _, ok := memInfoMetrics[k]; !ok {
memInfoMetrics[k] = prometheus.NewGauge()
c.registry.Register(
"node_memory_"+k,
k+" from /proc/meminfo",
prometheus.NilLabels,
memInfoMetrics[k],
)
gauge := prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: Namespace,
Subsystem: memInfoSubsystem,
Name: k,
Help: k + " from /proc/meminfo.",
})
memInfoMetrics[k] = prometheus.MustRegisterOrGet(gauge).(prometheus.Gauge)
}
updates++
memInfoMetrics[k].Set(nil, v)
memInfoMetrics[k].Set(v)
}
return updates, err
}

View file

@ -14,16 +14,16 @@ import (
)
const (
procNetDev = "/proc/net/dev"
procNetDev = "/proc/net/dev"
netStatsSubsystem = "network"
)
var (
netStatsMetrics = map[string]prometheus.Gauge{}
netStatsMetrics = map[string]*prometheus.GaugeVec{}
)
type netDevCollector struct {
registry prometheus.Registry
config Config
config Config
}
func init() {
@ -32,10 +32,9 @@ func init() {
// Takes a config struct and prometheus registry and returns a new Collector exposing
// network device stats.
func NewNetDevCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewNetDevCollector(config Config) (Collector, error) {
c := netDevCollector{
config: config,
registry: registry,
config: config,
}
return &c, nil
}
@ -50,20 +49,23 @@ func (c *netDevCollector) Update() (updates int, err error) {
for t, value := range stats {
key := direction + "_" + t
if _, ok := netStatsMetrics[key]; !ok {
netStatsMetrics[key] = prometheus.NewGauge()
c.registry.Register(
"node_network_"+key,
t+" "+direction+" from /proc/net/dev",
prometheus.NilLabels,
netStatsMetrics[key],
gv := prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Subsystem: netStatsSubsystem,
Name: key,
Help: fmt.Sprintf("%s %s from /proc/net/dev.", t, direction),
},
[]string{"device"},
)
netStatsMetrics[key] = prometheus.MustRegisterOrGet(gv).(*prometheus.GaugeVec)
}
updates++
v, err := strconv.ParseFloat(value, 64)
if err != nil {
return updates, fmt.Errorf("Invalid value %s in netstats: %s", value, err)
}
netStatsMetrics[key].Set(map[string]string{"device": dev}, v)
netStatsMetrics[key].WithLabelValues(dev).Set(v)
}
}
}

View file

@ -8,45 +8,57 @@ import (
"github.com/soundcloud/go-runit/runit"
)
var (
runitLabelNames = []string{"service"}
runitState = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Name: "service_state",
Help: "node_exporter: state of runit service.",
},
runitLabelNames,
)
runitStateDesired = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Name: "service_desired_state",
Help: "node_exporter: desired state of runit service.",
},
runitLabelNames,
)
runitStateNormal = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: Namespace,
Name: "service_normal_state",
Help: "node_exporter: normal state of runit service.",
},
runitLabelNames,
)
)
type runitCollector struct {
config Config
state prometheus.Gauge
stateDesired prometheus.Gauge
stateNormal prometheus.Gauge
config Config
}
func init() {
Factories["runit"] = NewRunitCollector
}
func NewRunitCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewRunitCollector(config Config) (Collector, error) {
c := runitCollector{
config: config,
state: prometheus.NewGauge(),
stateDesired: prometheus.NewGauge(),
stateNormal: prometheus.NewGauge(),
config: config,
}
registry.Register(
"node_service_state",
"node_exporter: state of runit service.",
prometheus.NilLabels,
c.state,
)
registry.Register(
"node_service_desired_state",
"node_exporter: desired state of runit service.",
prometheus.NilLabels,
c.stateDesired,
)
registry.Register(
"node_service_normal_state",
"node_exporter: normal state of runit service.",
prometheus.NilLabels,
c.stateNormal,
)
if _, err := prometheus.RegisterOrGet(runitState); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(runitStateDesired); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(runitStateNormal); err != nil {
return nil, err
}
return &c, nil
}
@ -65,16 +77,12 @@ func (c *runitCollector) Update() (updates int, err error) {
}
glog.V(1).Infof("%s is %d on pid %d for %d seconds", service.Name, status.State, status.Pid, status.Duration)
labels := map[string]string{
"service": service.Name,
}
c.state.Set(labels, float64(status.State))
c.stateDesired.Set(labels, float64(status.Want))
runitState.WithLabelValues(service.Name).Set(float64(status.State))
runitStateDesired.WithLabelValues(service.Name).Set(float64(status.Want))
if status.NormallyUp {
c.stateNormal.Set(labels, 1)
runitStateNormal.WithLabelValues(service.Name).Set(1)
} else {
c.stateNormal.Set(labels, 1)
runitStateNormal.WithLabelValues(service.Name).Set(1)
}
updates += 3
}

View file

@ -19,18 +19,48 @@ const (
)
var (
cpuMetrics = prometheus.NewCounter()
intrMetric = prometheus.NewCounter()
ctxtMetric = prometheus.NewCounter()
btimeMetric = prometheus.NewGauge()
forksMetric = prometheus.NewCounter()
procsRunningMetric = prometheus.NewGauge()
procsBlockedMetric = prometheus.NewGauge()
cpuMetrics = prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: Namespace,
Name: "cpu",
Help: "Seconds the cpus spent in each mode.",
},
[]string{"cpu", "mode"},
)
intrMetric = prometheus.NewCounter(prometheus.CounterOpts{
Namespace: Namespace,
Name: "intr",
Help: "Total number of interrupts serviced.",
})
ctxtMetric = prometheus.NewCounter(prometheus.CounterOpts{
Namespace: Namespace,
Name: "context_switches",
Help: "Total number of context switches.",
})
forksMetric = prometheus.NewCounter(prometheus.CounterOpts{
Namespace: Namespace,
Name: "forks",
Help: "Total number of forks.",
})
btimeMetric = prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: Namespace,
Name: "boot_time",
Help: "Node boot time, in unixtime.",
})
procsRunningMetric = prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: Namespace,
Name: "procs_running",
Help: "Number of processes in runnable state.",
})
procsBlockedMetric = prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: Namespace,
Name: "procs_blocked",
Help: "Number of processes blocked waiting for I/O to complete.",
})
)
type statCollector struct {
registry prometheus.Registry
config Config
config Config
}
func init() {
@ -39,53 +69,31 @@ func init() {
// Takes a config struct and prometheus registry and returns a new Collector exposing
// network device stats.
func NewStatCollector(config Config, registry prometheus.Registry) (Collector, error) {
func NewStatCollector(config Config) (Collector, error) {
c := statCollector{
config: config,
registry: registry,
config: config,
}
if _, err := prometheus.RegisterOrGet(cpuMetrics); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(intrMetric); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(ctxtMetric); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(forksMetric); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(btimeMetric); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(procsRunningMetric); err != nil {
return nil, err
}
if _, err := prometheus.RegisterOrGet(procsBlockedMetric); err != nil {
return nil, err
}
registry.Register(
"node_cpu",
"Seconds the cpus spent in each mode.",
prometheus.NilLabels,
cpuMetrics,
)
registry.Register(
"node_intr",
"Total number of interrupts serviced",
prometheus.NilLabels,
intrMetric,
)
registry.Register(
"node_context_switches",
"Total number of context switches.",
prometheus.NilLabels,
ctxtMetric,
)
registry.Register(
"node_forks",
"Total number of forks.",
prometheus.NilLabels,
forksMetric,
)
registry.Register(
"node_boot_time",
"Node boot time, in unixtime.",
prometheus.NilLabels,
btimeMetric,
)
registry.Register(
"node_procs_running",
"Number of processes in runnable state.",
prometheus.NilLabels,
procsRunningMetric,
)
registry.Register(
"node_procs_blocked",
"Number of processes blocked waiting for I/O to complete.",
prometheus.NilLabels,
procsBlockedMetric,
)
return &c, nil
}
@ -113,9 +121,9 @@ func (c *statCollector) Update() (updates int, err error) {
if err != nil {
return updates, err
}
// Convert from ticks to seconds
// Convert from ticks to seconds
value /= float64(C.sysconf(C._SC_CLK_TCK))
cpuMetrics.Set(map[string]string{"cpu": parts[0], "mode": cpuFields[i]}, value)
cpuMetrics.With(prometheus.Labels{"cpu": parts[0], "mode": cpuFields[i]}).Set(value)
}
case parts[0] == "intr":
// Only expose the overall number, use the 'interrupts' collector for more detail.
@ -123,37 +131,37 @@ func (c *statCollector) Update() (updates int, err error) {
if err != nil {
return updates, err
}
intrMetric.Set(prometheus.NilLabels, value)
intrMetric.Set(value)
case parts[0] == "ctxt":
value, err := strconv.ParseFloat(parts[1], 64)
if err != nil {
return updates, err
}
ctxtMetric.Set(prometheus.NilLabels, value)
ctxtMetric.Set(value)
case parts[0] == "processes":
value, err := strconv.ParseFloat(parts[1], 64)
if err != nil {
return updates, err
}
forksMetric.Set(prometheus.NilLabels, value)
forksMetric.Set(value)
case parts[0] == "btime":
value, err := strconv.ParseFloat(parts[1], 64)
if err != nil {
return updates, err
}
btimeMetric.Set(prometheus.NilLabels, value)
btimeMetric.Set(value)
case parts[0] == "procs_running":
value, err := strconv.ParseFloat(parts[1], 64)
if err != nil {
return updates, err
}
procsRunningMetric.Set(prometheus.NilLabels, value)
procsRunningMetric.Set(value)
case parts[0] == "procs_blocked":
value, err := strconv.ParseFloat(parts[1], 64)
if err != nil {
return updates, err
}
procsBlockedMetric.Set(prometheus.NilLabels, value)
procsBlockedMetric.Set(value)
}
}
return updates, err

View file

@ -1,6 +1,6 @@
{
"attributes" : {
"web-server" : "1",
"web_server" : "1",
"zone" : "a",
"default" : "1"
}

View file

@ -17,10 +17,11 @@ import (
"github.com/golang/glog"
"github.com/prometheus/client_golang/prometheus"
"github.com/prometheus/client_golang/prometheus/exp"
"github.com/prometheus/node_exporter/collector"
)
const subsystem = "exporter"
var (
configFile = flag.String("config", "node_exporter.conf", "config file.")
memProfile = flag.String("memprofile", "", "write memory profile to this file")
@ -28,8 +29,27 @@ var (
enabledCollectors = flag.String("enabledCollectors", "attributes,diskstats,filesystem,loadavg,meminfo,stat,netdev", "comma-seperated list of collectors to use")
printCollectors = flag.Bool("printCollectors", false, "If true, print available collectors and exit")
interval = flag.Duration("interval", 60*time.Second, "refresh interval")
scrapeDurations = prometheus.NewDefaultHistogram()
metricsUpdated = prometheus.NewGauge()
collectorLabelNames = []string{"collector", "result"}
scrapeDurations = prometheus.NewSummaryVec(
prometheus.SummaryOpts{
Namespace: collector.Namespace,
Subsystem: subsystem,
Name: "scrape_duration_seconds",
Help: "node_exporter: Duration of a scrape job.",
},
collectorLabelNames,
)
metricsUpdated = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Namespace: collector.Namespace,
Subsystem: subsystem,
Name: "metrics_updated",
Help: "node_exporter: Number of metrics updated.",
},
collectorLabelNames,
)
)
func main() {
@ -41,14 +61,13 @@ func main() {
}
return
}
registry := prometheus.NewRegistry()
collectors, err := loadCollectors(*configFile, registry)
collectors, err := loadCollectors(*configFile)
if err != nil {
log.Fatalf("Couldn't load config and collectors: %s", err)
}
registry.Register("node_exporter_scrape_duration_seconds", "node_exporter: Duration of a scrape job.", prometheus.NilLabels, scrapeDurations)
registry.Register("node_exporter_metrics_updated", "node_exporter: Number of metrics updated.", prometheus.NilLabels, metricsUpdated)
prometheus.MustRegister(scrapeDurations)
prometheus.MustRegister(metricsUpdated)
glog.Infof("Enabled collectors:")
for n, _ := range collectors {
@ -60,7 +79,7 @@ func main() {
signal.Notify(sigHup, syscall.SIGHUP)
signal.Notify(sigUsr1, syscall.SIGUSR1)
go serveStatus(registry)
go serveStatus()
glog.Infof("Starting initial collection")
collect(collectors)
@ -69,7 +88,7 @@ func main() {
for {
select {
case <-sigHup:
collectors, err = loadCollectors(*configFile, registry)
collectors, err = loadCollectors(*configFile)
if err != nil {
log.Fatalf("Couldn't load config and collectors: %s", err)
}
@ -96,7 +115,7 @@ func main() {
}
func loadCollectors(file string, registry prometheus.Registry) (map[string]collector.Collector, error) {
func loadCollectors(file string) (map[string]collector.Collector, error) {
collectors := map[string]collector.Collector{}
config, err := getConfig(file)
if err != nil {
@ -107,7 +126,7 @@ func loadCollectors(file string, registry prometheus.Registry) (map[string]colle
if !ok {
log.Fatalf("Collector '%s' not available", name)
}
c, err := fn(*config, registry)
c, err := fn(*config)
if err != nil {
return nil, err
}
@ -126,9 +145,9 @@ func getConfig(file string) (*collector.Config, error) {
return config, json.Unmarshal(bytes, &config)
}
func serveStatus(registry prometheus.Registry) {
exp.Handle(prometheus.ExpositionResource, registry.Handler())
http.ListenAndServe(*listeningAddress, exp.DefaultCoarseMux)
func serveStatus() {
http.Handle("/metrics", prometheus.Handler())
http.ListenAndServe(*listeningAddress, nil)
}
func collect(collectors map[string]collector.Collector) {
@ -147,17 +166,15 @@ func Execute(name string, c collector.Collector) {
begin := time.Now()
updates, err := c.Update()
duration := time.Since(begin)
var result string
label := map[string]string{
"collector": name,
}
if err != nil {
glog.Infof("ERROR: %s failed after %fs: %s", name, duration.Seconds(), err)
label["result"] = "error"
result = "error"
} else {
glog.Infof("OK: %s success after %fs.", name, duration.Seconds())
label["result"] = "success"
result = "success"
}
scrapeDurations.Add(label, duration.Seconds())
metricsUpdated.Set(label, float64(updates))
scrapeDurations.WithLabelValues(name, result).Observe(duration.Seconds())
metricsUpdated.WithLabelValues(name, result).Set(float64(updates))
}