Add interrupts collector filtering (#3028)

In order to reduce cardinality of the interrupts collector add
filtering options
* Add include/exclude regexp filter flags.
* Add boolean flag to include zero values, enabled by default.

Signed-off-by: Ben Kochie <superq@gmail.com>

README.md

@@ -103,6 +103,7 @@ filesystem | fs-types | N/A | --collector.filesystem.fs-types-exclude
 filesystem | mount-points | N/A | --collector.filesystem.mount-points-exclude
 hwmon | chip | --collector.hwmon.chip-include | --collector.hwmon.chip-exclude
 hwmon | sensor | --collector.hwmon.sensor-include | --collector.hwmon.sensor-exclude
+interrupts | name | --collector.interrupts.name-include | --collector.interrupts.name-exclude
 netdev | device | --collector.netdev.device-include | --collector.netdev.device-exclude
 qdisk | device | --collector.qdisk.device-include | --collector.qdisk.device-exclude
 slabinfo | slab-names | --collector.slabinfo.slabs-include | --collector.slabinfo.slabs-exclude

collector/interrupts_common.go

@@ -18,19 +18,28 @@
 package collector
 
 import (
+	"github.com/alecthomas/kingpin/v2"
 	"github.com/go-kit/log"
 	"github.com/prometheus/client_golang/prometheus"
 )
 
 type interruptsCollector struct {
-	desc   typedDesc
-	logger log.Logger
+	desc         typedDesc
+	logger       log.Logger
+	nameFilter   deviceFilter
+	includeZeros bool
 }
 
 func init() {
 	registerCollector("interrupts", defaultDisabled, NewInterruptsCollector)
 }
 
+var (
+	interruptsInclude      = kingpin.Flag("collector.interrupts.name-include", "Regexp of interrupts name to include (mutually exclusive to --collector.interrupts.name-exclude).").String()
+	interruptsExclude      = kingpin.Flag("collector.interrupts.name-exclude", "Regexp of interrupts name to exclude (mutually exclusive to --collector.interrupts.name-include).").String()
+	interruptsIncludeZeros = kingpin.Flag("collector.interrupts.include-zeros", "Include interrupts that have a zero value").Default("true").Bool()
+)
+
 // NewInterruptsCollector returns a new Collector exposing interrupts stats.
 func NewInterruptsCollector(logger log.Logger) (Collector, error) {
 	return &interruptsCollector{
@@ -39,6 +48,8 @@ func NewInterruptsCollector(logger log.Logger) (Collector, error) {
 			"Interrupt details.",
 			interruptLabelNames, nil,
 		), prometheus.CounterValue},
-		logger: logger,
+		logger:       logger,
+		nameFilter:   newDeviceFilter(*interruptsExclude, *interruptsInclude),
+		includeZeros: *interruptsIncludeZeros,
 	}, nil
 }

collector/interrupts_linux.go

@@ -25,6 +25,7 @@ import (
 	"strconv"
 	"strings"
 
+	"github.com/go-kit/log/level"
 	"github.com/prometheus/client_golang/prometheus"
 )
 
@@ -39,10 +40,19 @@ func (c *interruptsCollector) Update(ch chan<- prometheus.Metric) (err error) {
 	}
 	for name, interrupt := range interrupts {
 		for cpuNo, value := range interrupt.values {
+			filterName := name + ";" + interrupt.info + ";" + interrupt.devices
+			if c.nameFilter.ignored(filterName) {
+				level.Debug(c.logger).Log("msg", "ignoring interrupt name", "filter_name", filterName)
+				continue
+			}
 			fv, err := strconv.ParseFloat(value, 64)
 			if err != nil {
 				return fmt.Errorf("invalid value %s in interrupts: %w", value, err)
 			}
+			if !c.includeZeros && fv == 0.0 {
+				level.Debug(c.logger).Log("msg", "ignoring interrupt with zero value", "filter_name", filterName, "cpu", cpuNo)
+				continue
+			}
 			ch <- c.desc.mustNewConstMetric(fv, strconv.Itoa(cpuNo), name, interrupt.info, interrupt.devices)
 		}
 	}

collector/interrupts_openbsd.go

@@ -20,6 +20,7 @@ import (
 	"fmt"
 	"strconv"
 
+	"github.com/go-kit/log/level"
 	"github.com/prometheus/client_golang/prometheus"
 )
 
@@ -106,10 +107,20 @@ func (c *interruptsCollector) Update(ch chan<- prometheus.Metric) error {
 	}
 	for dev, interrupt := range interrupts {
 		for cpuNo, value := range interrupt.values {
+			interruptType := fmt.Sprintf("%d", interrupt.vector)
+			filterName := interruptType + ";" + dev
+			if c.nameFilter.ignored(filterName) {
+				level.Debug(c.logger).Log("msg", "ignoring interrupt name", "filter_name", filterName)
+				continue
+			}
+			if !c.includeZeros && value == 0.0 {
+				level.Debug(c.logger).Log("msg", "ignoring interrupt with zero value", "filter_name", filterName, "cpu", cpuNo)
+				continue
+			}
 			ch <- c.desc.mustNewConstMetric(
 				value,
 				strconv.Itoa(cpuNo),
-				strconv.Itoa(interrupt.vector),
+				interruptType,
 				dev,
 			)
 		}

collector/interrupts_openbsd_amd64.go

@@ -21,6 +21,7 @@ import (
 	"strconv"
 	"unsafe"
 
+	"github.com/go-kit/log/level"
 	"github.com/prometheus/client_golang/prometheus"
 	"golang.org/x/sys/unix"
 )
@@ -77,10 +78,20 @@ func (c *interruptsCollector) Update(ch chan<- prometheus.Metric) error {
 	}
 	for dev, interrupt := range interrupts {
 		for cpuNo, value := range interrupt.values {
+			interruptType := fmt.Sprintf("%d", interrupt.vector)
+			filterName := interruptType + ";" + dev
+			if c.nameFilter.ignored(filterName) {
+				level.Debug(c.logger).Log("msg", "ignoring interrupt name", "filter_name", filterName)
+				continue
+			}
+			if !c.includeZeros && value == 0.0 {
+				level.Debug(c.logger).Log("msg", "ignoring interrupt with zero value", "filter_name", filterName, "cpu", cpuNo)
+				continue
+			}
 			ch <- c.desc.mustNewConstMetric(
 				value,
 				strconv.Itoa(cpuNo),
-				fmt.Sprintf("%d", interrupt.vector),
+				interruptType,
 				dev,
 			)
 		}