mudhorn/node_exporter
1
0
Fork 0
mirror of https://github.com/prometheus/node_exporter.git synced 2024-12-28 06:59:44 -08:00
Code Issues Actions Packages Projects Releases Wiki Activity

NRestarts or NRefused aren't available on older systemd versions (#1039)

Browse source 
* If NRestarts or NRefused are not available, don't ignore the unit itself
* Don't report systemd metrics (NRestarts/NRefused) that are not available

Signed-off-by: James Hartig <james@getadmiral.com>
This commit is contained in:
James Hartig 2018-08-14 08:28:26 -04:00 committed by Ben Kochie
parent fe5a117831
commit 60c827231a
2 changed files with 28 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
CHANGELOG.md
View file

@ -9,6 +9,7 @@ The wifi collector is disabled by default due to suspected caching issues and go
* https://github.com/prometheus/node_exporter/issues/1008 * https://github.com/prometheus/node_exporter/issues/1008
* [CHANGE] Filter out non-installed units when collecting all systemd units #1011 * [CHANGE] Filter out non-installed units when collecting all systemd units #1011
* [CHANGE] `service_restart_total` and `socket_refused_connections_total` will not be reported if you're running an older version of systemd
* [FEATURE] Collect NRefused property for systemd socket units (available as of systemd v239) * [FEATURE] Collect NRefused property for systemd socket units (available as of systemd v239)
* [FEATURE] Collect NRestarts property for systemd service units * [FEATURE] Collect NRestarts property for systemd service units
* [FEATURE] Add socket unit stats to systemd collector #968 * [FEATURE] Add socket unit stats to systemd collector #968
@ -16,6 +17,7 @@ The wifi collector is disabled by default due to suspected caching issues and go
* [ENHANCEMENT] * [ENHANCEMENT]
* [BUGFIX] Fix goroutine leak in supervisord collector * [BUGFIX] Fix goroutine leak in supervisord collector
* [BUGFIX] Systemd units will not be ignored if you're running older versions of systemd #1039
* [BUGFIX] Handle vanishing PIDs #1043 * [BUGFIX] Handle vanishing PIDs #1043
## 0.16.0 / 2018-05-15 ## 0.16.0 / 2018-05-15

42
collector/systemd_linux.go
View file

@ -140,10 +140,10 @@ 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") { if strings.HasSuffix(unit.Name, ".service") && unit.nRestarts != nil {
ch <- prometheus.MustNewConstMetric( ch <- prometheus.MustNewConstMetric(
c.nRestartsDesc, prometheus.CounterValue, c.nRestartsDesc, prometheus.CounterValue,
float64(unit.nRestarts), unit.Name) float64(*unit.nRestarts), unit.Name)
} }
} }
} }
@ -160,9 +160,11 @@ func (c *systemdCollector) collectSockets(ch chan<- prometheus.Metric, units []u
ch <- prometheus.MustNewConstMetric( ch <- prometheus.MustNewConstMetric(
c.socketCurrentConnectionsDesc, prometheus.GaugeValue, c.socketCurrentConnectionsDesc, prometheus.GaugeValue,
float64(unit.currentConnections), unit.Name) float64(unit.currentConnections), unit.Name)
ch <- prometheus.MustNewConstMetric( if unit.refusedConnections != nil {
c.socketRefusedConnectionsDesc, prometheus.GaugeValue, ch <- prometheus.MustNewConstMetric(
float64(unit.refusedConnections), unit.Name) c.socketRefusedConnectionsDesc, prometheus.GaugeValue,
float64(*unit.refusedConnections), unit.Name)
}
} }
} }
@ -212,10 +214,10 @@ type unit struct {
dbus.UnitStatus dbus.UnitStatus
lastTriggerUsec uint64 lastTriggerUsec uint64
startTimeUsec uint64 startTimeUsec uint64
nRestarts uint32 nRestarts *uint32
acceptedConnections uint32 acceptedConnections uint32
currentConnections uint32 currentConnections uint32
refusedConnections uint32 refusedConnections *uint32
} }
func (c *systemdCollector) getAllUnits() ([]unit, error) { func (c *systemdCollector) getAllUnits() ([]unit, error) {
@ -241,40 +243,47 @@ func (c *systemdCollector) getAllUnits() ([]unit, error) {
if strings.HasSuffix(unit.Name, ".timer") { if strings.HasSuffix(unit.Name, ".timer") {
lastTriggerValue, err := conn.GetUnitTypeProperty(unit.Name, "Timer", "LastTriggerUSec") lastTriggerValue, err := conn.GetUnitTypeProperty(unit.Name, "Timer", "LastTriggerUSec")
if err != nil { if err != nil {
return nil, fmt.Errorf("couldn't get unit '%s' LastTriggerUSec: %s", unit.Name, err) log.Debugf("couldn't get unit '%s' LastTriggerUSec: %s\n", unit.Name, err)
continue
} }
unit.lastTriggerUsec = lastTriggerValue.Value.Value().(uint64) unit.lastTriggerUsec = lastTriggerValue.Value.Value().(uint64)
} }
if strings.HasSuffix(unit.Name, ".service") { if strings.HasSuffix(unit.Name, ".service") {
nRestarts, err := conn.GetUnitTypeProperty(unit.Name, "Service", "NRestarts") // NRestarts wasn't added until systemd 235.
restartsCount, err := conn.GetUnitTypeProperty(unit.Name, "Service", "NRestarts")
if err != nil { if err != nil {
log.Debugf("couldn't get unit '%s' NRestarts: %s\n", unit.Name, err) log.Debugf("couldn't get unit '%s' NRestarts: %s\n", unit.Name, err)
continue } else {
nRestarts := restartsCount.Value.Value().(uint32)
unit.nRestarts = &nRestarts
} }
unit.nRestarts = nRestarts.Value.Value().(uint32)
} }
if strings.HasSuffix(unit.Name, ".socket") { if strings.HasSuffix(unit.Name, ".socket") {
acceptedConnectionCount, err := conn.GetUnitTypeProperty(unit.Name, "Socket", "NAccepted") acceptedConnectionCount, err := conn.GetUnitTypeProperty(unit.Name, "Socket", "NAccepted")
if err != nil { if err != nil {
return nil, fmt.Errorf("couldn't get unit '%s' NAccepted: %s", unit.Name, err) log.Debugf("couldn't get unit '%s' NAccepted: %s\n", unit.Name, err)
continue
} }
unit.acceptedConnections = acceptedConnectionCount.Value.Value().(uint32) unit.acceptedConnections = acceptedConnectionCount.Value.Value().(uint32)
currentConnectionCount, err := conn.GetUnitTypeProperty(unit.Name, "Socket", "NConnections") currentConnectionCount, err := conn.GetUnitTypeProperty(unit.Name, "Socket", "NConnections")
if err != nil { if err != nil {
return nil, fmt.Errorf("couldn't get unit '%s' NConnections: %s", unit.Name, err) log.Debugf("couldn't get unit '%s' NConnections: %s\n", unit.Name, err)
continue
} }
unit.currentConnections = currentConnectionCount.Value.Value().(uint32) unit.currentConnections = currentConnectionCount.Value.Value().(uint32)
// NRefused wasn't added until systemd 239.
refusedConnectionCount, err := conn.GetUnitTypeProperty(unit.Name, "Socket", "NRefused") refusedConnectionCount, err := conn.GetUnitTypeProperty(unit.Name, "Socket", "NRefused")
if err != nil { if err != nil {
log.Debugf("couldn't get unit '%s' NRefused: %s\n", unit.Name, err) log.Debugf("couldn't get unit '%s' NRefused: %s\n", unit.Name, err)
continue } else {
nRefused := refusedConnectionCount.Value.Value().(uint32)
unit.refusedConnections = &nRefused
} }
unit.refusedConnections = refusedConnectionCount.Value.Value().(uint32)
} }
if unit.ActiveState != "active" { if unit.ActiveState != "active" {
@ -282,7 +291,8 @@ func (c *systemdCollector) getAllUnits() ([]unit, error) {
} else { } else {
timestampValue, err := conn.GetUnitProperty(unit.Name, "ActiveEnterTimestamp") timestampValue, err := conn.GetUnitProperty(unit.Name, "ActiveEnterTimestamp")
if err != nil { if err != nil {
return nil, fmt.Errorf("couldn't get unit '%s' StartTimeUsec: %s", unit.Name, err) log.Debugf("couldn't get unit '%s' StartTimeUsec: %s\n", unit.Name, err)
continue
} }
unit.startTimeUsec = timestampValue.Value.Value().(uint64) unit.startTimeUsec = timestampValue.Value.Value().(uint64)