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>
pull/968/head
xginn8 2018-07-05 07:31:45 -04:00 committed by Ben Kochie
parent ee1e1997bc
commit 8af84a215d
2 changed files with 20 additions and 1 deletions

View File

@ -3,7 +3,7 @@
**Breaking changes** **Breaking changes**
* [CHANGE] * [CHANGE]
* [FEATURE] * [FEATURE] Collect NRestarts property for systemd service units
* [ENHANCEMENT] * [ENHANCEMENT]
* [BUGFIX] * [BUGFIX]

View File

@ -36,6 +36,7 @@ type systemdCollector struct {
unitDesc *prometheus.Desc unitDesc *prometheus.Desc
systemRunningDesc *prometheus.Desc systemRunningDesc *prometheus.Desc
summaryDesc *prometheus.Desc summaryDesc *prometheus.Desc
nRestartsDesc *prometheus.Desc
timerLastTriggerDesc *prometheus.Desc timerLastTriggerDesc *prometheus.Desc
unitWhitelistPattern *regexp.Regexp unitWhitelistPattern *regexp.Regexp
unitBlacklistPattern *regexp.Regexp unitBlacklistPattern *regexp.Regexp
@ -63,6 +64,9 @@ func NewSystemdCollector() (Collector, error) {
summaryDesc := prometheus.NewDesc( summaryDesc := prometheus.NewDesc(
prometheus.BuildFQName(namespace, subsystem, "units"), prometheus.BuildFQName(namespace, subsystem, "units"),
"Summary of systemd unit states", []string{"state"}, nil) "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( timerLastTriggerDesc := prometheus.NewDesc(
prometheus.BuildFQName(namespace, subsystem, "timer_last_trigger_seconds"), prometheus.BuildFQName(namespace, subsystem, "timer_last_trigger_seconds"),
"Seconds since epoch of last trigger.", []string{"name"}, nil) "Seconds since epoch of last trigger.", []string{"name"}, nil)
@ -73,6 +77,7 @@ func NewSystemdCollector() (Collector, error) {
unitDesc: unitDesc, unitDesc: unitDesc,
systemRunningDesc: systemRunningDesc, systemRunningDesc: systemRunningDesc,
summaryDesc: summaryDesc, summaryDesc: summaryDesc,
nRestartsDesc: nRestartsDesc,
timerLastTriggerDesc: timerLastTriggerDesc, timerLastTriggerDesc: timerLastTriggerDesc,
unitWhitelistPattern: unitWhitelistPattern, unitWhitelistPattern: unitWhitelistPattern,
unitBlacklistPattern: unitBlacklistPattern, unitBlacklistPattern: unitBlacklistPattern,
@ -112,6 +117,11 @@ func (c *systemdCollector) collectUnitStatusMetrics(ch chan<- prometheus.Metric,
c.unitDesc, prometheus.GaugeValue, isActive, c.unitDesc, prometheus.GaugeValue, isActive,
unit.Name, stateName) 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 { type unit struct {
dbus.UnitStatus dbus.UnitStatus
lastTriggerUsec uint64 lastTriggerUsec uint64
nRestarts uint32
} }
func (c *systemdCollector) getAllUnits() ([]unit, error) { func (c *systemdCollector) getAllUnits() ([]unit, error) {
@ -181,6 +192,14 @@ func (c *systemdCollector) getAllUnits() ([]unit, error) {
unit.lastTriggerUsec = lastTriggerValue.Value.Value().(uint64) 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) result = append(result, unit)
} }