etcdserver: add OS level FD metrics

Similar counts are exposed via Prometheus.
This adds the one that are perceived by etcd server.

e.g.

os_fd_limit 120000
os_fd_used 14
process_cpu_seconds_total 0.31
process_max_fds 120000
process_open_fds 17

Signed-off-by: Gyuho Lee <leegyuho@amazon.com>

etcdserver/metrics.go

@@ -151,6 +151,19 @@ var (
 		Help:      "Server or member ID in hexadecimal format. 1 for 'server_id' label with current ID.",
 	},
 		[]string{"server_id"})
+
+	fdUsed = prometheus.NewGauge(prometheus.GaugeOpts{
+		Namespace: "os",
+		Subsystem: "fd",
+		Name:      "used",
+		Help:      "The number of used file descriptors.",
+	})
+	fdLimit = prometheus.NewGauge(prometheus.GaugeOpts{
+		Namespace: "os",
+		Subsystem: "fd",
+		Name:      "limit",
+		Help:      "The file descriptor limit.",
+	})
 )
 
 func init() {
@@ -174,6 +187,8 @@ func init() {
 	prometheus.MustRegister(isLearner)
 	prometheus.MustRegister(learnerPromoteSucceed)
 	prometheus.MustRegister(learnerPromoteFailed)
+	prometheus.MustRegister(fdUsed)
+	prometheus.MustRegister(fdLimit)
 
 	currentVersion.With(prometheus.Labels{
 		"server_version": version.Version,
@@ -184,7 +199,6 @@ func init() {
 }
 
 func monitorFileDescriptor(lg *zap.Logger, done <-chan struct{}) {
-
 	// This ticker will check File Descriptor Requirements ,and count all fds in used.
 	// And recorded some logs when in used >= limit/5*4. Just recorded message.
 	// If fds was more than 10K,It's low performance due to FDUsage() works.
@@ -198,11 +212,13 @@ func monitorFileDescriptor(lg *zap.Logger, done <-chan struct{}) {
 			lg.Warn("failed to get file descriptor usage", zap.Error(err))
 			return
 		}
+		fdUsed.Set(float64(used))
 		limit, err := runtime.FDLimit()
 		if err != nil {
 			lg.Warn("failed to get file descriptor limit", zap.Error(err))
 			return
 		}
+		fdLimit.Set(float64(limit))
 		if used >= limit/5*4 {
 			lg.Warn("80% of file descriptors are used", zap.Uint64("used", used), zap.Uint64("limit", limit))
 		}