Expose cpu bugs and flags as info metrics. (#1788)

* Expose cpu bugs and flags as info metrics with a regexp filter.
* Automatically enable CPU info metrics when using flags or bugs feature.

Signed-off-by: domgoer <[email protected]>

Author: domechn

collector/cpu_linux.go

@@ -18,6 +18,7 @@ package collector
 import (
 	"fmt"
 	"path/filepath"
+	"regexp"
 	"strconv"
 	"sync"
 
@@ -32,16 +33,23 @@ type cpuCollector struct {
 	fs                 procfs.FS
 	cpu                *prometheus.Desc
 	cpuInfo            *prometheus.Desc
+	cpuFlagsInfo       *prometheus.Desc
+	cpuBugsInfo        *prometheus.Desc
 	cpuGuest           *prometheus.Desc
 	cpuCoreThrottle    *prometheus.Desc
 	cpuPackageThrottle *prometheus.Desc
 	logger             log.Logger
 	cpuStats           []procfs.CPUStat
 	cpuStatsMutex      sync.Mutex
+
+	cpuFlagsIncludeRegexp *regexp.Regexp
+	cpuBugsIncludeRegexp  *regexp.Regexp
 }
 
 var (
 	enableCPUInfo = kingpin.Flag("collector.cpu.info", "Enables metric cpu_info").Bool()
+	flagsInclude  = kingpin.Flag("collector.cpu.info.flags-include", "Filter the `flags` field in cpuInfo with a value that must be a regular expression").String()
+	bugsInclude   = kingpin.Flag("collector.cpu.info.bugs-include", "Filter the `bugs` field in cpuInfo with a value that must be a regular expression").String()
 )
 
 func init() {
@@ -54,14 +62,24 @@ func NewCPUCollector(logger log.Logger) (Collector, error) {
 	if err != nil {
 		return nil, fmt.Errorf("failed to open procfs: %w", err)
 	}
-	return &cpuCollector{
+	c := &cpuCollector{
 		fs:  fs,
 		cpu: nodeCPUSecondsDesc,
 		cpuInfo: prometheus.NewDesc(
 			prometheus.BuildFQName(namespace, cpuCollectorSubsystem, "info"),
 			"CPU information from /proc/cpuinfo.",
 			[]string{"package", "core", "cpu", "vendor", "family", "model", "model_name", "microcode", "stepping", "cachesize"}, nil,
 		),
+		cpuFlagsInfo: prometheus.NewDesc(
+			prometheus.BuildFQName(namespace, cpuCollectorSubsystem, "flag_info"),
+			"The `flags` field of CPU information from /proc/cpuinfo.",
+			[]string{"flag"}, nil,
+		),
+		cpuBugsInfo: prometheus.NewDesc(
+			prometheus.BuildFQName(namespace, cpuCollectorSubsystem, "bug_info"),
+			"The `bugs` field of CPU information from /proc/cpuinfo.",
+			[]string{"bug"}, nil,
+		),
 		cpuGuest: prometheus.NewDesc(
 			prometheus.BuildFQName(namespace, cpuCollectorSubsystem, "guest_seconds_total"),
 			"Seconds the cpus spent in guests (VMs) for each mode.",
@@ -78,7 +96,34 @@ func NewCPUCollector(logger log.Logger) (Collector, error) {
 			[]string{"package"}, nil,
 		),
 		logger: logger,
-	}, nil
+	}
+	err = c.compileIncludeFlags(flagsInclude, bugsInclude)
+	if err != nil {
+		return nil, fmt.Errorf("fail to compile --collector.cpu.info.flags-include and --collector.cpu.info.bugs-include, the values of them must be regular expressions: %w", err)
+	}
+	return c, nil
+}
+
+func (c *cpuCollector) compileIncludeFlags(flagsIncludeFlag, bugsIncludeFlag *string) error {
+	if (*flagsIncludeFlag != "" || *bugsIncludeFlag != "") && !*enableCPUInfo {
+		*enableCPUInfo = true
+		level.Info(c.logger).Log("msg", "--collector.cpu.info has been set to `true` because you set the following flags, like --collector.cpu.info.flags-include and --collector.cpu.info.bugs-include")
+	}
+
+	var err error
+	if *flagsIncludeFlag != "" {
+		c.cpuFlagsIncludeRegexp, err = regexp.Compile(*flagsIncludeFlag)
+		if err != nil {
+			return err
+		}
+	}
+	if *bugsIncludeFlag != "" {
+		c.cpuBugsIncludeRegexp, err = regexp.Compile(*bugsIncludeFlag)
+		if err != nil {
+			return err
+		}
+	}
+	return nil
 }
 
 // Update implements Collector and exposes cpu related metrics from /proc/stat and /sys/.../cpu/.
@@ -117,6 +162,31 @@ func (c *cpuCollector) updateInfo(ch chan<- prometheus.Metric) error {
 			cpu.Microcode,
 			cpu.Stepping,
 			cpu.CacheSize)
+
+		if err := updateFieldInfo(cpu.Flags, c.cpuFlagsIncludeRegexp, c.cpuFlagsInfo, ch); err != nil {
+			return err
+		}
+		if err := updateFieldInfo(cpu.Bugs, c.cpuBugsIncludeRegexp, c.cpuBugsInfo, ch); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func updateFieldInfo(valueList []string, filter *regexp.Regexp, desc *prometheus.Desc, ch chan<- prometheus.Metric) error {
+	if filter == nil {
+		return nil
+	}
+
+	for _, val := range valueList {
+		if !filter.MatchString(val) {
+			continue
+		}
+		ch <- prometheus.MustNewConstMetric(desc,
+			prometheus.GaugeValue,
+			1,
+			val,
+		)
 	}
 	return nil
 }

collector/fixtures/e2e-64k-page-output.txt

@@ -184,12 +184,24 @@ node_cooling_device_cur_state{name="0",type="Processor"} 0
 # HELP node_cooling_device_max_state Maximum throttle state of the cooling device
 # TYPE node_cooling_device_max_state gauge
 node_cooling_device_max_state{name="0",type="Processor"} 3
+# HELP node_cpu_bug_info The `bugs` field of CPU information from /proc/cpuinfo.
+# TYPE node_cpu_bug_info gauge
+node_cpu_bug_info{bug="cpu_meltdown"} 1
+node_cpu_bug_info{bug="mds"} 1
+node_cpu_bug_info{bug="spectre_v1"} 1
+node_cpu_bug_info{bug="spectre_v2"} 1
 # HELP node_cpu_core_throttles_total Number of times this cpu core has been throttled.
 # TYPE node_cpu_core_throttles_total counter
 node_cpu_core_throttles_total{core="0",package="0"} 5
 node_cpu_core_throttles_total{core="0",package="1"} 0
 node_cpu_core_throttles_total{core="1",package="0"} 0
 node_cpu_core_throttles_total{core="1",package="1"} 9
+# HELP node_cpu_flag_info The `flags` field of CPU information from /proc/cpuinfo.
+# TYPE node_cpu_flag_info gauge
+node_cpu_flag_info{flag="aes"} 1
+node_cpu_flag_info{flag="avx"} 1
+node_cpu_flag_info{flag="avx2"} 1
+node_cpu_flag_info{flag="constant_tsc"} 1
 # HELP node_cpu_guest_seconds_total Seconds the cpus spent in guests (VMs) for each mode.
 # TYPE node_cpu_guest_seconds_total counter
 node_cpu_guest_seconds_total{cpu="0",mode="nice"} 0.01

collector/fixtures/e2e-output.txt

@@ -232,12 +232,24 @@ node_cooling_device_cur_state{name="0",type="Processor"} 0
 # HELP node_cooling_device_max_state Maximum throttle state of the cooling device
 # TYPE node_cooling_device_max_state gauge
 node_cooling_device_max_state{name="0",type="Processor"} 3
+# HELP node_cpu_bug_info The `bugs` field of CPU information from /proc/cpuinfo.
+# TYPE node_cpu_bug_info gauge
+node_cpu_bug_info{bug="cpu_meltdown"} 1
+node_cpu_bug_info{bug="mds"} 1
+node_cpu_bug_info{bug="spectre_v1"} 1
+node_cpu_bug_info{bug="spectre_v2"} 1
 # HELP node_cpu_core_throttles_total Number of times this cpu core has been throttled.
 # TYPE node_cpu_core_throttles_total counter
 node_cpu_core_throttles_total{core="0",package="0"} 5
 node_cpu_core_throttles_total{core="0",package="1"} 0
 node_cpu_core_throttles_total{core="1",package="0"} 0
 node_cpu_core_throttles_total{core="1",package="1"} 9
+# HELP node_cpu_flag_info The `flags` field of CPU information from /proc/cpuinfo.
+# TYPE node_cpu_flag_info gauge
+node_cpu_flag_info{flag="aes"} 1
+node_cpu_flag_info{flag="avx"} 1
+node_cpu_flag_info{flag="avx2"} 1
+node_cpu_flag_info{flag="constant_tsc"} 1
 # HELP node_cpu_guest_seconds_total Seconds the cpus spent in guests (VMs) for each mode.
 # TYPE node_cpu_guest_seconds_total counter
 node_cpu_guest_seconds_total{cpu="0",mode="nice"} 0.01

end-to-end-test.sh

@@ -107,6 +107,8 @@ fi
   --collector.qdisc.fixtures="collector/fixtures/qdisc/" \
   --collector.netclass.ignored-devices="(bond0|dmz|int)" \
   --collector.cpu.info \
+  --collector.cpu.info.flags-include="^(aes|avx.?|constant_tsc)$" \
+  --collector.cpu.info.bugs-include="^(cpu_meltdown|spectre_.*|mds)$" \
   --web.listen-address "127.0.0.1:${port}" \
   --log.level="debug" > "${tmpdir}/node_exporter.log" 2>&1 &