moby/daemon/metrics.go
Brian Goff 426e610e43 Use runtime spec modifier for metrics plugin hook
Currently the metrics plugin uses a really hackish host mount with
propagated mounts to get the metrics socket into a plugin after the
plugin is alreay running.
This approach ends up leaking mounts which requires setting the plugin
manager root to private, which causes some other issues.

With this change, plugin subsystems can register a set of modifiers to
apply to the plugin's runtime spec before the plugin is ever started.
This will help to generalize some of the customization work that needs
to happen for various plugin subsystems (and future ones).

Specifically it lets the metrics plugin subsystem append a mount to the
runtime spec to mount the metrics socket in the plugin's mount namespace
rather than the host's and prevetns any leaking due to this mount.

Signed-off-by: Brian Goff <cpuguy83@gmail.com>
2018-02-07 15:48:26 -05:00

151 lines
4.3 KiB
Go

package daemon
import (
"sync"
"github.com/docker/docker/pkg/plugingetter"
metrics "github.com/docker/go-metrics"
"github.com/pkg/errors"
"github.com/prometheus/client_golang/prometheus"
"github.com/sirupsen/logrus"
)
const metricsPluginType = "MetricsCollector"
var (
containerActions metrics.LabeledTimer
imageActions metrics.LabeledTimer
networkActions metrics.LabeledTimer
engineInfo metrics.LabeledGauge
engineCpus metrics.Gauge
engineMemory metrics.Gauge
healthChecksCounter metrics.Counter
healthChecksFailedCounter metrics.Counter
stateCtr *stateCounter
)
func init() {
ns := metrics.NewNamespace("engine", "daemon", nil)
containerActions = ns.NewLabeledTimer("container_actions", "The number of seconds it takes to process each container action", "action")
for _, a := range []string{
"start",
"changes",
"commit",
"create",
"delete",
} {
containerActions.WithValues(a).Update(0)
}
networkActions = ns.NewLabeledTimer("network_actions", "The number of seconds it takes to process each network action", "action")
engineInfo = ns.NewLabeledGauge("engine", "The information related to the engine and the OS it is running on", metrics.Unit("info"),
"version",
"commit",
"architecture",
"graphdriver",
"kernel", "os",
"os_type",
"daemon_id", // ID is a randomly generated unique identifier (e.g. UUID4)
)
engineCpus = ns.NewGauge("engine_cpus", "The number of cpus that the host system of the engine has", metrics.Unit("cpus"))
engineMemory = ns.NewGauge("engine_memory", "The number of bytes of memory that the host system of the engine has", metrics.Bytes)
healthChecksCounter = ns.NewCounter("health_checks", "The total number of health checks")
healthChecksFailedCounter = ns.NewCounter("health_checks_failed", "The total number of failed health checks")
imageActions = ns.NewLabeledTimer("image_actions", "The number of seconds it takes to process each image action", "action")
stateCtr = newStateCounter(ns.NewDesc("container_states", "The count of containers in various states", metrics.Unit("containers"), "state"))
ns.Add(stateCtr)
metrics.Register(ns)
}
type stateCounter struct {
mu sync.Mutex
states map[string]string
desc *prometheus.Desc
}
func newStateCounter(desc *prometheus.Desc) *stateCounter {
return &stateCounter{
states: make(map[string]string),
desc: desc,
}
}
func (ctr *stateCounter) get() (running int, paused int, stopped int) {
ctr.mu.Lock()
defer ctr.mu.Unlock()
states := map[string]int{
"running": 0,
"paused": 0,
"stopped": 0,
}
for _, state := range ctr.states {
states[state]++
}
return states["running"], states["paused"], states["stopped"]
}
func (ctr *stateCounter) set(id, label string) {
ctr.mu.Lock()
ctr.states[id] = label
ctr.mu.Unlock()
}
func (ctr *stateCounter) del(id string) {
ctr.mu.Lock()
delete(ctr.states, id)
ctr.mu.Unlock()
}
func (ctr *stateCounter) Describe(ch chan<- *prometheus.Desc) {
ch <- ctr.desc
}
func (ctr *stateCounter) Collect(ch chan<- prometheus.Metric) {
running, paused, stopped := ctr.get()
ch <- prometheus.MustNewConstMetric(ctr.desc, prometheus.GaugeValue, float64(running), "running")
ch <- prometheus.MustNewConstMetric(ctr.desc, prometheus.GaugeValue, float64(paused), "paused")
ch <- prometheus.MustNewConstMetric(ctr.desc, prometheus.GaugeValue, float64(stopped), "stopped")
}
func (d *Daemon) cleanupMetricsPlugins() {
ls := d.PluginStore.GetAllManagedPluginsByCap(metricsPluginType)
var wg sync.WaitGroup
wg.Add(len(ls))
for _, plugin := range ls {
p := plugin
go func() {
defer wg.Done()
pluginStopMetricsCollection(p)
}()
}
wg.Wait()
if d.metricsPluginListener != nil {
d.metricsPluginListener.Close()
}
}
func pluginStartMetricsCollection(p plugingetter.CompatPlugin) error {
type metricsPluginResponse struct {
Err string
}
var res metricsPluginResponse
if err := p.Client().Call(metricsPluginType+".StartMetrics", nil, &res); err != nil {
return errors.Wrap(err, "could not start metrics plugin")
}
if res.Err != "" {
return errors.New(res.Err)
}
return nil
}
func pluginStopMetricsCollection(p plugingetter.CompatPlugin) {
if err := p.Client().Call(metricsPluginType+".StopMetrics", nil, nil); err != nil {
logrus.WithError(err).WithField("name", p.Name()).Error("error stopping metrics collector")
}
}