Moved system package inputs out to top level (#4406)
This commit is contained in:
Steve Domino
committed by
Daniel Nelson
Daniel Nelson
parent
9a14d1f074
commit
7b73b0db3a
65
plugins/inputs/processes/README.md
Normal file
65
plugins/inputs/processes/README.md
Normal file
@@ -0,0 +1,65 @@
|
||||
# Processes Input Plugin
|
||||
|
||||
This plugin gathers info about the total number of processes and groups
|
||||
them by status (zombie, sleeping, running, etc.)
|
||||
|
||||
On linux this plugin requires access to procfs (/proc), on other OSes
|
||||
it requires access to execute `ps`.
|
||||
|
||||
### Configuration:
|
||||
|
||||
```toml
|
||||
# Get the number of processes and group them by status
|
||||
[[inputs.processes]]
|
||||
# no configuration
|
||||
```
|
||||
|
||||
Another possible configuration is to define an alternative path for resolving the /proc location.
|
||||
Using the environment variable `HOST_PROC` the plugin will retrieve process information from the specified location.
|
||||
|
||||
`docker run -v /proc:/rootfs/proc:ro -e HOST_PROC=/rootfs/proc`
|
||||
|
||||
### Measurements & Fields:
|
||||
|
||||
- processes
|
||||
- blocked (aka disk sleep or uninterruptible sleep)
|
||||
- running
|
||||
- sleeping
|
||||
- stopped
|
||||
- total
|
||||
- zombie
|
||||
- dead
|
||||
- wait (freebsd only)
|
||||
- idle (bsd and Linux 4+ only)
|
||||
- paging (linux only)
|
||||
- total_threads (linux only)
|
||||
|
||||
### Process State Mappings
|
||||
|
||||
Different OSes use slightly different State codes for their processes, these
|
||||
state codes are documented in `man ps`, and I will give a mapping of what major
|
||||
OS state codes correspond to in telegraf metrics:
|
||||
|
||||
```
|
||||
Linux FreeBSD Darwin meaning
|
||||
R R R running
|
||||
S S S sleeping
|
||||
Z Z Z zombie
|
||||
X none none dead
|
||||
T T T stopped
|
||||
I I I idle (sleeping for longer than about 20 seconds)
|
||||
D D,L U blocked (waiting in uninterruptible sleep, or locked)
|
||||
W W none paging (linux kernel < 2.6 only), wait (freebsd)
|
||||
```
|
||||
|
||||
### Tags:
|
||||
|
||||
None
|
||||
|
||||
### Example Output:
|
||||
|
||||
```
|
||||
$ telegraf --config ~/ws/telegraf.conf --input-filter processes --test
|
||||
* Plugin: processes, Collection 1
|
||||
> processes blocked=8i,running=1i,sleeping=265i,stopped=0i,total=274i,zombie=0i,dead=0i,paging=0i,total_threads=687i 1457478636980905042
|
||||
```
|
||||
237
plugins/inputs/processes/processes.go
Normal file
237
plugins/inputs/processes/processes.go
Normal file
@@ -0,0 +1,237 @@
|
||||
// +build !windows
|
||||
|
||||
package processes
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"fmt"
|
||||
"io/ioutil"
|
||||
"log"
|
||||
"os"
|
||||
"os/exec"
|
||||
"path/filepath"
|
||||
"runtime"
|
||||
"strconv"
|
||||
"syscall"
|
||||
|
||||
"github.com/influxdata/telegraf"
|
||||
"github.com/influxdata/telegraf/plugins/inputs"
|
||||
"github.com/influxdata/telegraf/plugins/inputs/linux_sysctl_fs"
|
||||
)
|
||||
|
||||
type Processes struct {
|
||||
execPS func() ([]byte, error)
|
||||
readProcFile func(filename string) ([]byte, error)
|
||||
|
||||
forcePS bool
|
||||
forceProc bool
|
||||
}
|
||||
|
||||
func (p *Processes) Description() string {
|
||||
return "Get the number of processes and group them by status"
|
||||
}
|
||||
|
||||
func (p *Processes) SampleConfig() string { return "" }
|
||||
|
||||
func (p *Processes) Gather(acc telegraf.Accumulator) error {
|
||||
// Get an empty map of metric fields
|
||||
fields := getEmptyFields()
|
||||
|
||||
// Decide if we will use 'ps' to get stats (use procfs otherwise)
|
||||
usePS := true
|
||||
if runtime.GOOS == "linux" {
|
||||
usePS = false
|
||||
}
|
||||
if p.forcePS {
|
||||
usePS = true
|
||||
} else if p.forceProc {
|
||||
usePS = false
|
||||
}
|
||||
|
||||
// Gather stats from 'ps' or procfs
|
||||
if usePS {
|
||||
if err := p.gatherFromPS(fields); err != nil {
|
||||
return err
|
||||
}
|
||||
} else {
|
||||
if err := p.gatherFromProc(fields); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
|
||||
acc.AddGauge("processes", fields, nil)
|
||||
return nil
|
||||
}
|
||||
|
||||
// Gets empty fields of metrics based on the OS
|
||||
func getEmptyFields() map[string]interface{} {
|
||||
fields := map[string]interface{}{
|
||||
"blocked": int64(0),
|
||||
"zombies": int64(0),
|
||||
"stopped": int64(0),
|
||||
"running": int64(0),
|
||||
"sleeping": int64(0),
|
||||
"total": int64(0),
|
||||
"unknown": int64(0),
|
||||
}
|
||||
switch runtime.GOOS {
|
||||
case "freebsd":
|
||||
fields["idle"] = int64(0)
|
||||
fields["wait"] = int64(0)
|
||||
case "darwin":
|
||||
fields["idle"] = int64(0)
|
||||
case "openbsd":
|
||||
fields["idle"] = int64(0)
|
||||
case "linux":
|
||||
fields["dead"] = int64(0)
|
||||
fields["paging"] = int64(0)
|
||||
fields["total_threads"] = int64(0)
|
||||
fields["idle"] = int64(0)
|
||||
}
|
||||
return fields
|
||||
}
|
||||
|
||||
// exec `ps` to get all process states
|
||||
func (p *Processes) gatherFromPS(fields map[string]interface{}) error {
|
||||
out, err := p.execPS()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
for i, status := range bytes.Fields(out) {
|
||||
if i == 0 && string(status) == "STAT" {
|
||||
// This is a header, skip it
|
||||
continue
|
||||
}
|
||||
switch status[0] {
|
||||
case 'W':
|
||||
fields["wait"] = fields["wait"].(int64) + int64(1)
|
||||
case 'U', 'D', 'L':
|
||||
// Also known as uninterruptible sleep or disk sleep
|
||||
fields["blocked"] = fields["blocked"].(int64) + int64(1)
|
||||
case 'Z':
|
||||
fields["zombies"] = fields["zombies"].(int64) + int64(1)
|
||||
case 'X':
|
||||
fields["dead"] = fields["dead"].(int64) + int64(1)
|
||||
case 'T':
|
||||
fields["stopped"] = fields["stopped"].(int64) + int64(1)
|
||||
case 'R':
|
||||
fields["running"] = fields["running"].(int64) + int64(1)
|
||||
case 'S':
|
||||
fields["sleeping"] = fields["sleeping"].(int64) + int64(1)
|
||||
case 'I':
|
||||
fields["idle"] = fields["idle"].(int64) + int64(1)
|
||||
case '?':
|
||||
fields["unknown"] = fields["unknown"].(int64) + int64(1)
|
||||
default:
|
||||
log.Printf("I! processes: Unknown state [ %s ] from ps",
|
||||
string(status[0]))
|
||||
}
|
||||
fields["total"] = fields["total"].(int64) + int64(1)
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// get process states from /proc/(pid)/stat files
|
||||
func (p *Processes) gatherFromProc(fields map[string]interface{}) error {
|
||||
filenames, err := filepath.Glob(linux_sysctl_fs.GetHostProc() + "/[0-9]*/stat")
|
||||
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
for _, filename := range filenames {
|
||||
_, err := os.Stat(filename)
|
||||
data, err := p.readProcFile(filename)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
if data == nil {
|
||||
continue
|
||||
}
|
||||
|
||||
// Parse out data after (<cmd name>)
|
||||
i := bytes.LastIndex(data, []byte(")"))
|
||||
if i == -1 {
|
||||
continue
|
||||
}
|
||||
data = data[i+2:]
|
||||
|
||||
stats := bytes.Fields(data)
|
||||
if len(stats) < 3 {
|
||||
return fmt.Errorf("Something is terribly wrong with %s", filename)
|
||||
}
|
||||
switch stats[0][0] {
|
||||
case 'R':
|
||||
fields["running"] = fields["running"].(int64) + int64(1)
|
||||
case 'S':
|
||||
fields["sleeping"] = fields["sleeping"].(int64) + int64(1)
|
||||
case 'D':
|
||||
fields["blocked"] = fields["blocked"].(int64) + int64(1)
|
||||
case 'Z':
|
||||
fields["zombies"] = fields["zombies"].(int64) + int64(1)
|
||||
case 'X':
|
||||
fields["dead"] = fields["dead"].(int64) + int64(1)
|
||||
case 'T', 't':
|
||||
fields["stopped"] = fields["stopped"].(int64) + int64(1)
|
||||
case 'W':
|
||||
fields["paging"] = fields["paging"].(int64) + int64(1)
|
||||
case 'I':
|
||||
fields["idle"] = fields["idle"].(int64) + int64(1)
|
||||
default:
|
||||
log.Printf("I! processes: Unknown state [ %s ] in file %s",
|
||||
string(stats[0][0]), filename)
|
||||
}
|
||||
fields["total"] = fields["total"].(int64) + int64(1)
|
||||
|
||||
threads, err := strconv.Atoi(string(stats[17]))
|
||||
if err != nil {
|
||||
log.Printf("I! processes: Error parsing thread count: %s", err)
|
||||
continue
|
||||
}
|
||||
fields["total_threads"] = fields["total_threads"].(int64) + int64(threads)
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func readProcFile(filename string) ([]byte, error) {
|
||||
data, err := ioutil.ReadFile(filename)
|
||||
if err != nil {
|
||||
if os.IsNotExist(err) {
|
||||
return nil, nil
|
||||
}
|
||||
|
||||
// Reading from /proc/<PID> fails with ESRCH if the process has
|
||||
// been terminated between open() and read().
|
||||
if perr, ok := err.(*os.PathError); ok && perr.Err == syscall.ESRCH {
|
||||
return nil, nil
|
||||
}
|
||||
|
||||
return nil, err
|
||||
}
|
||||
|
||||
return data, nil
|
||||
}
|
||||
|
||||
func execPS() ([]byte, error) {
|
||||
bin, err := exec.LookPath("ps")
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
out, err := exec.Command(bin, "axo", "state").Output()
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
return out, err
|
||||
}
|
||||
|
||||
func init() {
|
||||
inputs.Add("processes", func() telegraf.Input {
|
||||
return &Processes{
|
||||
execPS: execPS,
|
||||
readProcFile: readProcFile,
|
||||
}
|
||||
})
|
||||
}
|
||||
186
plugins/inputs/processes/processes_test.go
Normal file
186
plugins/inputs/processes/processes_test.go
Normal file
@@ -0,0 +1,186 @@
|
||||
// +build !windows
|
||||
|
||||
package processes
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"runtime"
|
||||
"testing"
|
||||
|
||||
"github.com/influxdata/telegraf/testutil"
|
||||
"github.com/stretchr/testify/assert"
|
||||
"github.com/stretchr/testify/require"
|
||||
)
|
||||
|
||||
func TestProcesses(t *testing.T) {
|
||||
processes := &Processes{
|
||||
execPS: execPS,
|
||||
readProcFile: readProcFile,
|
||||
}
|
||||
var acc testutil.Accumulator
|
||||
|
||||
err := processes.Gather(&acc)
|
||||
require.NoError(t, err)
|
||||
|
||||
assert.True(t, acc.HasInt64Field("processes", "running"))
|
||||
assert.True(t, acc.HasInt64Field("processes", "sleeping"))
|
||||
assert.True(t, acc.HasInt64Field("processes", "stopped"))
|
||||
assert.True(t, acc.HasInt64Field("processes", "total"))
|
||||
total, ok := acc.Get("processes")
|
||||
require.True(t, ok)
|
||||
assert.True(t, total.Fields["total"].(int64) > 0)
|
||||
}
|
||||
|
||||
func TestFromPS(t *testing.T) {
|
||||
processes := &Processes{
|
||||
execPS: testExecPS,
|
||||
forcePS: true,
|
||||
}
|
||||
|
||||
var acc testutil.Accumulator
|
||||
err := processes.Gather(&acc)
|
||||
require.NoError(t, err)
|
||||
|
||||
fields := getEmptyFields()
|
||||
fields["blocked"] = int64(4)
|
||||
fields["zombies"] = int64(1)
|
||||
fields["running"] = int64(4)
|
||||
fields["sleeping"] = int64(34)
|
||||
fields["idle"] = int64(2)
|
||||
fields["total"] = int64(45)
|
||||
|
||||
acc.AssertContainsTaggedFields(t, "processes", fields, map[string]string{})
|
||||
}
|
||||
|
||||
func TestFromPSError(t *testing.T) {
|
||||
processes := &Processes{
|
||||
execPS: testExecPSError,
|
||||
forcePS: true,
|
||||
}
|
||||
|
||||
var acc testutil.Accumulator
|
||||
err := processes.Gather(&acc)
|
||||
require.Error(t, err)
|
||||
}
|
||||
|
||||
func TestFromProcFiles(t *testing.T) {
|
||||
if runtime.GOOS != "linux" {
|
||||
t.Skip("This test only runs on linux")
|
||||
}
|
||||
tester := tester{}
|
||||
processes := &Processes{
|
||||
readProcFile: tester.testProcFile,
|
||||
forceProc: true,
|
||||
}
|
||||
|
||||
var acc testutil.Accumulator
|
||||
err := processes.Gather(&acc)
|
||||
require.NoError(t, err)
|
||||
|
||||
fields := getEmptyFields()
|
||||
fields["sleeping"] = tester.calls
|
||||
fields["total_threads"] = tester.calls * 2
|
||||
fields["total"] = tester.calls
|
||||
|
||||
acc.AssertContainsTaggedFields(t, "processes", fields, map[string]string{})
|
||||
}
|
||||
|
||||
func TestFromProcFilesWithSpaceInCmd(t *testing.T) {
|
||||
if runtime.GOOS != "linux" {
|
||||
t.Skip("This test only runs on linux")
|
||||
}
|
||||
tester := tester{}
|
||||
processes := &Processes{
|
||||
readProcFile: tester.testProcFile2,
|
||||
forceProc: true,
|
||||
}
|
||||
|
||||
var acc testutil.Accumulator
|
||||
err := processes.Gather(&acc)
|
||||
require.NoError(t, err)
|
||||
|
||||
fields := getEmptyFields()
|
||||
fields["sleeping"] = tester.calls
|
||||
fields["total_threads"] = tester.calls * 2
|
||||
fields["total"] = tester.calls
|
||||
|
||||
acc.AssertContainsTaggedFields(t, "processes", fields, map[string]string{})
|
||||
}
|
||||
|
||||
func testExecPS() ([]byte, error) {
|
||||
return []byte(testPSOut), nil
|
||||
}
|
||||
|
||||
// struct for counting calls to testProcFile
|
||||
type tester struct {
|
||||
calls int64
|
||||
}
|
||||
|
||||
func (t *tester) testProcFile(_ string) ([]byte, error) {
|
||||
t.calls++
|
||||
return []byte(fmt.Sprintf(testProcStat, "S", "2")), nil
|
||||
}
|
||||
|
||||
func (t *tester) testProcFile2(_ string) ([]byte, error) {
|
||||
t.calls++
|
||||
return []byte(fmt.Sprintf(testProcStat2, "S", "2")), nil
|
||||
}
|
||||
|
||||
func testExecPSError() ([]byte, error) {
|
||||
return []byte(testPSOut), fmt.Errorf("ERROR!")
|
||||
}
|
||||
|
||||
const testPSOut = `
|
||||
STAT
|
||||
S
|
||||
S
|
||||
S
|
||||
S
|
||||
R
|
||||
R
|
||||
S
|
||||
S
|
||||
Ss
|
||||
Ss
|
||||
S
|
||||
SNs
|
||||
Ss
|
||||
Ss
|
||||
S
|
||||
R+
|
||||
S
|
||||
U
|
||||
S
|
||||
S
|
||||
S
|
||||
S
|
||||
Ss
|
||||
S+
|
||||
Ss
|
||||
S
|
||||
S+
|
||||
S+
|
||||
Ss
|
||||
S+
|
||||
Ss
|
||||
S
|
||||
R+
|
||||
Ss
|
||||
S
|
||||
S+
|
||||
S+
|
||||
Ss
|
||||
L
|
||||
U
|
||||
Z
|
||||
D
|
||||
S+
|
||||
I
|
||||
I
|
||||
`
|
||||
|
||||
const testProcStat = `10 (rcuob/0) %s 2 0 0 0 -1 2129984 0 0 0 0 0 0 0 0 20 0 %s 0 11 0 0 18446744073709551615 0 0 0 0 0 0 0 2147483647 0 18446744073709551615 0 0 17 0 0 0 0 0 0 0 0 0 0 0 0 0 0
|
||||
`
|
||||
|
||||
const testProcStat2 = `10 (rcuob 0) %s 2 0 0 0 -1 2129984 0 0 0 0 0 0 0 0 20 0 %s 0 11 0 0 18446744073709551615 0 0 0 0 0 0 0 2147483647 0 18446744073709551615 0 0 17 0 0 0 0 0 0 0 0 0 0 0 0 0 0
|
||||
`
|
||||
3
plugins/inputs/processes/processes_windows.go
Normal file
3
plugins/inputs/processes/processes_windows.go
Normal file
@@ -0,0 +1,3 @@
|
||||
// +build windows
|
||||
|
||||
package processes
|
||||
Reference in New Issue
Block a user