From 8af84a215dd21a7d78be0a82cf88b3b5ce52250c Mon Sep 17 00:00:00 2001
From: xginn8 <xginn8@users.noreply.github.com>
Date: Thu, 5 Jul 2018 07:31:45 -0400
Subject: [PATCH] Add support for NRestarts counter introduced in systemd 235
 (#992)

* Add support for NRestarts counter introduced in systemd 235

`.service` units increment this counter any time the Restart= condition is
triggered.

Signed-off-by: Matthew McGinn <mamcgi@gmail.com>
---
 CHANGELOG.md               |  2 +-
 collector/systemd_linux.go | 19 +++++++++++++++++++
 2 files changed, 20 insertions(+), 1 deletion(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index b9f4b153cc..35e8899b2c 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -3,7 +3,7 @@
 **Breaking changes**
 
 * [CHANGE]
-* [FEATURE]
+* [FEATURE] Collect NRestarts property for systemd service units
 * [ENHANCEMENT]
 * [BUGFIX]
 
diff --git a/collector/systemd_linux.go b/collector/systemd_linux.go
index f90b49a502..e05c289aad 100644
--- a/collector/systemd_linux.go
+++ b/collector/systemd_linux.go
@@ -36,6 +36,7 @@ type systemdCollector struct {
 	unitDesc             *prometheus.Desc
 	systemRunningDesc    *prometheus.Desc
 	summaryDesc          *prometheus.Desc
+	nRestartsDesc        *prometheus.Desc
 	timerLastTriggerDesc *prometheus.Desc
 	unitWhitelistPattern *regexp.Regexp
 	unitBlacklistPattern *regexp.Regexp
@@ -63,6 +64,9 @@ func NewSystemdCollector() (Collector, error) {
 	summaryDesc := prometheus.NewDesc(
 		prometheus.BuildFQName(namespace, subsystem, "units"),
 		"Summary of systemd unit states", []string{"state"}, nil)
+	nRestartsDesc := prometheus.NewDesc(
+		prometheus.BuildFQName(namespace, subsystem, "service_restart_total"),
+		"Service unit count of Restart triggers", []string{"state"}, nil)
 	timerLastTriggerDesc := prometheus.NewDesc(
 		prometheus.BuildFQName(namespace, subsystem, "timer_last_trigger_seconds"),
 		"Seconds since epoch of last trigger.", []string{"name"}, nil)
@@ -73,6 +77,7 @@ func NewSystemdCollector() (Collector, error) {
 		unitDesc:             unitDesc,
 		systemRunningDesc:    systemRunningDesc,
 		summaryDesc:          summaryDesc,
+		nRestartsDesc:        nRestartsDesc,
 		timerLastTriggerDesc: timerLastTriggerDesc,
 		unitWhitelistPattern: unitWhitelistPattern,
 		unitBlacklistPattern: unitBlacklistPattern,
@@ -112,6 +117,11 @@ func (c *systemdCollector) collectUnitStatusMetrics(ch chan<- prometheus.Metric,
 				c.unitDesc, prometheus.GaugeValue, isActive,
 				unit.Name, stateName)
 		}
+		if strings.HasSuffix(unit.Name, ".service") {
+			ch <- prometheus.MustNewConstMetric(
+				c.nRestartsDesc, prometheus.CounterValue,
+				float64(unit.nRestarts), unit.Name)
+		}
 	}
 }
 
@@ -153,6 +163,7 @@ func (c *systemdCollector) newDbus() (*dbus.Conn, error) {
 type unit struct {
 	dbus.UnitStatus
 	lastTriggerUsec uint64
+	nRestarts       uint32
 }
 
 func (c *systemdCollector) getAllUnits() ([]unit, error) {
@@ -181,6 +192,14 @@ func (c *systemdCollector) getAllUnits() ([]unit, error) {
 
 			unit.lastTriggerUsec = lastTriggerValue.Value.Value().(uint64)
 		}
+		if strings.HasSuffix(unit.Name, ".service") {
+			nRestarts, err := conn.GetUnitTypeProperty(unit.Name, "Service", "NRestarts")
+			if err != nil {
+				log.Debugf("couldn't get unit '%s' NRestarts: %s\n", unit.Name, err)
+				continue
+			}
+			unit.nRestarts = nRestarts.Value.Value().(uint32)
+		}
 
 		result = append(result, unit)
 	}