fix(inputs.exec): Clean up grandchildren processes (#13937)

internal/exec_unix.go

@@ -19,14 +19,22 @@ const KillGrace = 5 * time.Second
 func WaitTimeout(c *exec.Cmd, timeout time.Duration) error {
 	var kill *time.Timer
 	term := time.AfterFunc(timeout, func() {
-		err := c.Process.Signal(syscall.SIGTERM)
+		err := syscall.Kill(-c.Process.Pid, syscall.SIGTERM)
+		if err != nil {
+			log.Printf("E! [agent] Error terminating process children: %s", err)
+		}
+		err = c.Process.Signal(syscall.SIGTERM)
 		if err != nil {
 			log.Printf("E! [agent] Error terminating process: %s", err)
 			return
 		}
 
 		kill = time.AfterFunc(KillGrace, func() {
-			err := c.Process.Kill()
+			err := syscall.Kill(-c.Process.Pid, syscall.SIGKILL)
+			if err != nil {
+				log.Printf("E! [agent] Error terminating process children: %s", err)
+			}
+			err = c.Process.Kill()
 			if err != nil {
 				log.Printf("E! [agent] Error killing process: %s", err)
 				return

plugins/inputs/exec/exec.go

@@ -7,19 +7,14 @@ import (
 	"errors"
 	"fmt"
 	"io"
-	"os"
-	osExec "os/exec"
 	"path/filepath"
 	"runtime"
 	"strings"
 	"sync"
 	"time"
 
-	"github.com/kballard/go-shellquote"
-
 	"github.com/influxdata/telegraf"
 	"github.com/influxdata/telegraf/config"
-	"github.com/influxdata/telegraf/internal"
 	"github.com/influxdata/telegraf/models"
 	"github.com/influxdata/telegraf/plugins/inputs"
 	"github.com/influxdata/telegraf/plugins/parsers/nagios"
@@ -61,40 +56,6 @@ type Runner interface {
 
 type CommandRunner struct{}
 
-func (c CommandRunner) Run(
-	command string,
-	environments []string,
-	timeout time.Duration,
-) ([]byte, []byte, error) {
-	splitCmd, err := shellquote.Split(command)
-	if err != nil || len(splitCmd) == 0 {
-		return nil, nil, fmt.Errorf("exec: unable to parse command: %w", err)
-	}
-
-	cmd := osExec.Command(splitCmd[0], splitCmd[1:]...)
-
-	if len(environments) > 0 {
-		cmd.Env = append(os.Environ(), environments...)
-	}
-
-	var (
-		out    bytes.Buffer
-		stderr bytes.Buffer
-	)
-	cmd.Stdout = &out
-	cmd.Stderr = &stderr
-
-	runErr := internal.RunTimeout(cmd, timeout)
-
-	out = removeWindowsCarriageReturns(out)
-	if stderr.Len() > 0 && !telegraf.Debug {
-		stderr = removeWindowsCarriageReturns(stderr)
-		stderr = c.truncate(stderr)
-	}
-
-	return out.Bytes(), stderr.Bytes(), runErr
-}
-
 func (c CommandRunner) truncate(buf bytes.Buffer) bytes.Buffer {
 	// Limit the number of bytes.
 	didTruncate := false

plugins/inputs/exec/run_notwinodws.go

@@ -0,0 +1,51 @@
+//go:build !windows
+
+package exec
+
+import (
+	"bytes"
+	"fmt"
+	"os"
+	osExec "os/exec"
+	"syscall"
+	"time"
+
+	"github.com/influxdata/telegraf"
+	"github.com/influxdata/telegraf/internal"
+	"github.com/kballard/go-shellquote"
+)
+
+func (c CommandRunner) Run(
+	command string,
+	environments []string,
+	timeout time.Duration,
+) ([]byte, []byte, error) {
+	splitCmd, err := shellquote.Split(command)
+	if err != nil || len(splitCmd) == 0 {
+		return nil, nil, fmt.Errorf("exec: unable to parse command: %w", err)
+	}
+
+	cmd := osExec.Command(splitCmd[0], splitCmd[1:]...)
+	cmd.SysProcAttr = &syscall.SysProcAttr{Setpgid: true}
+
+	if len(environments) > 0 {
+		cmd.Env = append(os.Environ(), environments...)
+	}
+
+	var (
+		out    bytes.Buffer
+		stderr bytes.Buffer
+	)
+	cmd.Stdout = &out
+	cmd.Stderr = &stderr
+
+	runErr := internal.RunTimeout(cmd, timeout)
+
+	out = removeWindowsCarriageReturns(out)
+	if stderr.Len() > 0 && !telegraf.Debug {
+		stderr = removeWindowsCarriageReturns(stderr)
+		stderr = c.truncate(stderr)
+	}
+
+	return out.Bytes(), stderr.Bytes(), runErr
+}

plugins/inputs/exec/run_windows.go

@@ -0,0 +1,53 @@
+//go:build windows
+
+package exec
+
+import (
+	"bytes"
+	"fmt"
+	"os"
+	osExec "os/exec"
+	"syscall"
+	"time"
+
+	"github.com/influxdata/telegraf"
+	"github.com/influxdata/telegraf/internal"
+	"github.com/kballard/go-shellquote"
+)
+
+func (c CommandRunner) Run(
+	command string,
+	environments []string,
+	timeout time.Duration,
+) ([]byte, []byte, error) {
+	splitCmd, err := shellquote.Split(command)
+	if err != nil || len(splitCmd) == 0 {
+		return nil, nil, fmt.Errorf("exec: unable to parse command: %w", err)
+	}
+
+	cmd := osExec.Command(splitCmd[0], splitCmd[1:]...)
+	cmd.SysProcAttr = &syscall.SysProcAttr{
+		CreationFlags: syscall.CREATE_NEW_PROCESS_GROUP,
+	}
+
+	if len(environments) > 0 {
+		cmd.Env = append(os.Environ(), environments...)
+	}
+
+	var (
+		out    bytes.Buffer
+		stderr bytes.Buffer
+	)
+	cmd.Stdout = &out
+	cmd.Stderr = &stderr
+
+	runErr := internal.RunTimeout(cmd, timeout)
+
+	out = removeWindowsCarriageReturns(out)
+	if stderr.Len() > 0 && !telegraf.Debug {
+		stderr = removeWindowsCarriageReturns(stderr)
+		stderr = c.truncate(stderr)
+	}
+
+	return out.Bytes(), stderr.Bytes(), runErr
+}