Add a "kernel" plugin for /proc/stat statistics

CHANGELOG.md

@@ -1,4 +1,4 @@
-## v0.10.5 [unreleased]
+## v0.11.0 [unreleased]
 
 ### Release Notes
 
@@ -16,6 +16,7 @@
 - [#776](https://github.com/influxdata/telegraf/pull/776): Add Zookeeper chroot option to kafka_consumer. Thanks @prune998!
 - [#811](https://github.com/influxdata/telegraf/pull/811): Add processes plugin for classifying total procs on system. Thanks @titilambert!
 - [#235](https://github.com/influxdata/telegraf/issues/235): Add number of users to the `system` input plugin.
+- [#826](https://github.com/influxdata/telegraf/pull/826): "kernel" linux plugin for /proc/stat metrics (context switches, interrupts, etc.)
 
 ### Bugfixes
 - [#748](https://github.com/influxdata/telegraf/issues/748): Fix sensor plugin split on ":"

README.md

@@ -215,6 +215,7 @@ Currently implemented sources:
     * diskio
     * swap
     * processes
+    * kernel (/proc/stat)
 
 Telegraf can also collect metrics via the following service plugins:
 

etc/telegraf.conf

@@ -119,6 +119,10 @@
   # Uncomment the following line if you do not need disk serial numbers.
   # skip_serial_number = true
 
+# Get kernel statistics from /proc/stat
+[[inputs.kernel]]
+  # no configuration
+
 # Read metrics about memory usage
 [[inputs.mem]]
   # no configuration

plugins/inputs/system/KERNEL_README.md

@@ -0,0 +1,64 @@
+# Kernel Input Plugin
+
+This plugin is only available on Linux.
+
+The kernel plugin gathers info about the kernel that doesn't fit into other
+plugins. In general, it is the statistics available in `/proc/stat` that are
+not covered by other plugins.
+
+The metrics are documented in `man proc` under the `/proc/stat` section.
+
+```
+/proc/stat
+kernel/system statistics. Varies with architecture. Common entries include:
+
+page 5741 1808
+The number of pages the system paged in and the number that were paged out (from disk).
+
+swap 1 0
+The number of swap pages that have been brought in and out.
+
+intr 1462898
+This line shows counts of interrupts serviced since boot time, for each of
+the possible system interrupts. The first column is the total of all
+interrupts serviced; each subsequent column is the total for a particular interrupt.
+
+ctxt 115315
+The number of context switches that the system underwent.
+
+btime 769041601
+boot time, in seconds since the Epoch, 1970-01-01 00:00:00 +0000 (UTC).
+
+processes 86031
+Number of forks since boot.
+```
+
+### Configuration:
+
+```toml
+# Get kernel statistics from /proc/stat
+[[inputs.kernel]]
+  # no configuration
+```
+
+### Measurements & Fields:
+
+- kernel
+    - boot_time (integer, seconds since epoch, `btime`)
+    - context_switches (integer, `ctxt`)
+    - disk_pages_in (integer, `page (0)`)
+    - disk_pages_out (integer, `page (1)`)
+    - interrupts (integer, `intr`)
+    - processes_forked (integer, `processes`)
+
+### Tags:
+
+None
+
+### Example Output:
+
+```
+$ telegraf -config ~/ws/telegraf.conf -input-filter kernel -test
+* Plugin: kernel, Collection 1
+> kernel boot_time=1457505775i,context_switches=2626618i,disk_pages_in=5741i,disk_pages_out=1808i,interrupts=1472736i,processes_forked=10673i 1457613402960879816
+```

plugins/inputs/system/kernel.go

@@ -0,0 +1,110 @@
+// +build linux
+
+package system
+
+import (
+	"bytes"
+	"fmt"
+	"io/ioutil"
+	"os"
+	"strconv"
+
+	"github.com/influxdata/telegraf"
+	"github.com/influxdata/telegraf/plugins/inputs"
+)
+
+// /proc/stat file line prefixes to gather stats on:
+var (
+	interrupts       = []byte("intr")
+	context_switches = []byte("ctxt")
+	processes_forked = []byte("processes")
+	disk_pages       = []byte("page")
+	boot_time        = []byte("btime")
+)
+
+type Kernel struct {
+	statFile string
+}
+
+func (k *Kernel) Description() string {
+	return "Get kernel statistics from /proc/stat"
+}
+
+func (k *Kernel) SampleConfig() string { return "" }
+
+func (k *Kernel) Gather(acc telegraf.Accumulator) error {
+	data, err := k.getProcStat()
+	if err != nil {
+		return err
+	}
+
+	fields := make(map[string]interface{})
+
+	dataFields := bytes.Fields(data)
+	for i, field := range dataFields {
+		switch {
+		case bytes.Equal(field, interrupts):
+			m, err := strconv.Atoi(string(dataFields[i+1]))
+			if err != nil {
+				return err
+			}
+			fields["interrupts"] = int64(m)
+		case bytes.Equal(field, context_switches):
+			m, err := strconv.Atoi(string(dataFields[i+1]))
+			if err != nil {
+				return err
+			}
+			fields["context_switches"] = int64(m)
+		case bytes.Equal(field, processes_forked):
+			m, err := strconv.Atoi(string(dataFields[i+1]))
+			if err != nil {
+				return err
+			}
+			fields["processes_forked"] = int64(m)
+		case bytes.Equal(field, boot_time):
+			m, err := strconv.Atoi(string(dataFields[i+1]))
+			if err != nil {
+				return err
+			}
+			fields["boot_time"] = int64(m)
+		case bytes.Equal(field, disk_pages):
+			in, err := strconv.Atoi(string(dataFields[i+1]))
+			if err != nil {
+				return err
+			}
+			out, err := strconv.Atoi(string(dataFields[i+2]))
+			if err != nil {
+				return err
+			}
+			fields["disk_pages_in"] = int64(in)
+			fields["disk_pages_out"] = int64(out)
+		}
+	}
+
+	acc.AddFields("kernel", fields, map[string]string{})
+
+	return nil
+}
+
+func (k *Kernel) getProcStat() ([]byte, error) {
+	if _, err := os.Stat(k.statFile); os.IsNotExist(err) {
+		return nil, fmt.Errorf("kernel: %s does not exist!", k.statFile)
+	} else if err != nil {
+		return nil, err
+	}
+
+	data, err := ioutil.ReadFile(k.statFile)
+	if err != nil {
+		return nil, err
+	}
+
+	return data, nil
+}
+
+func init() {
+	inputs.Add("kernel", func() telegraf.Input {
+		return &Kernel{
+			statFile: "/proc/stat",
+		}
+	})
+}

plugins/inputs/system/kernel_test.go

@@ -0,0 +1,164 @@
+// +build linux
+
+package system
+
+import (
+	"io/ioutil"
+	"os"
+	"testing"
+
+	"github.com/influxdata/telegraf/testutil"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestFullProcFile(t *testing.T) {
+	tmpfile := makeFakeStatFile([]byte(statFile_Full))
+	defer os.Remove(tmpfile)
+
+	k := Kernel{
+		statFile: tmpfile,
+	}
+
+	acc := testutil.Accumulator{}
+	err := k.Gather(&acc)
+	assert.NoError(t, err)
+
+	fields := map[string]interface{}{
+		"boot_time":        int64(1457505775),
+		"context_switches": int64(2626618),
+		"disk_pages_in":    int64(5741),
+		"disk_pages_out":   int64(1808),
+		"interrupts":       int64(1472736),
+		"processes_forked": int64(10673),
+	}
+	acc.AssertContainsFields(t, "kernel", fields)
+}
+
+func TestPartialProcFile(t *testing.T) {
+	tmpfile := makeFakeStatFile([]byte(statFile_Partial))
+	defer os.Remove(tmpfile)
+
+	k := Kernel{
+		statFile: tmpfile,
+	}
+
+	acc := testutil.Accumulator{}
+	err := k.Gather(&acc)
+	assert.NoError(t, err)
+
+	fields := map[string]interface{}{
+		"boot_time":        int64(1457505775),
+		"context_switches": int64(2626618),
+		"disk_pages_in":    int64(5741),
+		"disk_pages_out":   int64(1808),
+		"interrupts":       int64(1472736),
+	}
+	acc.AssertContainsFields(t, "kernel", fields)
+}
+
+func TestInvalidProcFile1(t *testing.T) {
+	tmpfile := makeFakeStatFile([]byte(statFile_Invalid))
+	defer os.Remove(tmpfile)
+
+	k := Kernel{
+		statFile: tmpfile,
+	}
+
+	acc := testutil.Accumulator{}
+	err := k.Gather(&acc)
+	assert.Error(t, err)
+}
+
+func TestInvalidProcFile2(t *testing.T) {
+	tmpfile := makeFakeStatFile([]byte(statFile_Invalid2))
+	defer os.Remove(tmpfile)
+
+	k := Kernel{
+		statFile: tmpfile,
+	}
+
+	acc := testutil.Accumulator{}
+	err := k.Gather(&acc)
+	assert.Error(t, err)
+}
+
+func TestNoProcFile(t *testing.T) {
+	tmpfile := makeFakeStatFile([]byte(statFile_Invalid2))
+	os.Remove(tmpfile)
+
+	k := Kernel{
+		statFile: tmpfile,
+	}
+
+	acc := testutil.Accumulator{}
+	err := k.Gather(&acc)
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "does not exist")
+}
+
+const statFile_Full = `cpu  6796 252 5655 10444977 175 0 101 0 0 0
+cpu0 6796 252 5655 10444977 175 0 101 0 0 0
+intr 1472736 57 10 0 0 0 0 0 0 0 0 0 0 156 0 0 0 0 0 0 111551 42541 12356 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
+ctxt 2626618
+btime 1457505775
+processes 10673
+procs_running 2
+procs_blocked 0
+softirq 1031662 0 649485 20946 111071 11620 0 1 0 994 237545
+page 5741 1808
+swap 1 0
+`
+
+const statFile_Partial = `cpu  6796 252 5655 10444977 175 0 101 0 0 0
+cpu0 6796 252 5655 10444977 175 0 101 0 0 0
+intr 1472736 57 10 0 0 0 0 0 0 0 0 0 0 156 0 0 0 0 0 0 111551 42541 12356 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
+ctxt 2626618
+btime 1457505775
+procs_running 2
+procs_blocked 0
+softirq 1031662 0 649485 20946 111071 11620 0 1 0 994 237545
+page 5741 1808
+`
+
+// missing btime measurement
+const statFile_Invalid = `cpu  6796 252 5655 10444977 175 0 101 0 0 0
+cpu0 6796 252 5655 10444977 175 0 101 0 0 0
+intr 1472736 57 10 0 0 0 0 0 0 0 0 0 0 156 0 0 0 0 0 0 111551 42541 12356 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
+ctxt 2626618
+btime
+processes 10673
+procs_running 2
+procs_blocked 0
+softirq 1031662 0 649485 20946 111071 11620 0 1 0 994 237545
+page 5741 1808
+swap 1 0
+`
+
+// missing second page measurement
+const statFile_Invalid2 = `cpu  6796 252 5655 10444977 175 0 101 0 0 0
+cpu0 6796 252 5655 10444977 175 0 101 0 0 0
+intr 1472736 57 10 0 0 0 0 0 0 0 0 0 0 156 0 0 0 0 0 0 111551 42541 12356 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
+ctxt 2626618
+processes 10673
+procs_running 2
+page 5741
+procs_blocked 0
+softirq 1031662 0 649485 20946 111071 11620 0 1 0 994 237545
+`
+
+func makeFakeStatFile(content []byte) string {
+	tmpfile, err := ioutil.TempFile("", "kerneltest")
+	if err != nil {
+		panic(err)
+	}
+
+	if _, err := tmpfile.Write(content); err != nil {
+		panic(err)
+	}
+	if err := tmpfile.Close(); err != nil {
+		panic(err)
+	}
+
+	return tmpfile.Name()
+}