fix(parsers/nagios): metrics will always return a supported status co…

plugins/inputs/exec/exec.go

@@ -127,9 +127,9 @@ func (e *Exec) ProcessCommand(command string, acc telegraf.Accumulator, wg *sync
 	defer wg.Done()
 	_, isNagios := e.parser.(*nagios.NagiosParser)
 
-	out, errbuf, runErr := e.runner.Run(command, e.Environment, time.Duration(e.Timeout))
+	out, errBuf, runErr := e.runner.Run(command, e.Environment, time.Duration(e.Timeout))
 	if !isNagios && runErr != nil {
-		err := fmt.Errorf("exec: %s for command '%s': %s", runErr, command, string(errbuf))
+		err := fmt.Errorf("exec: %s for command '%s': %s", runErr, command, string(errBuf))
 		acc.AddError(err)
 		return
 	}
@@ -141,10 +141,7 @@ func (e *Exec) ProcessCommand(command string, acc telegraf.Accumulator, wg *sync
 	}
 
 	if isNagios {
-		metrics, err = nagios.TryAddState(runErr, metrics)
-		if err != nil {
-			e.Log.Errorf("Failed to add nagios state: %s", err)
-		}
+		metrics = nagios.AddState(runErr, errBuf, metrics)
 	}
 
 	for _, m := range metrics {

plugins/parsers/nagios/parser.go

@@ -4,7 +4,6 @@ import (
 	"bufio"
 	"bytes"
 	"errors"
-	"fmt"
 	"os/exec"
 	"regexp"
 	"strconv"
@@ -16,6 +15,10 @@ import (
 	"github.com/influxdata/telegraf/metric"
 )
 
+// unknownExitCode is the nagios unknown status code
+// the exit code should be used if an error occurs or something unexpected happens
+const unknownExitCode = 3
+
 // getExitCode get the exit code from an error value which is the result
 // of running a command through exec package api.
 func getExitCode(err error) (int, error) {
@@ -25,10 +28,7 @@ func getExitCode(err error) (int, error) {
 
 	ee, ok := err.(*exec.ExitError)
 	if !ok {
-		// If it is not an *exec.ExitError, then it must be
-		// an io error, but docs do not say anything about the
-		// exit code in this case.
-		return 0, err
+		return unknownExitCode, err
 	}
 
 	ws, ok := ee.Sys().(syscall.WaitStatus)
@@ -39,19 +39,35 @@ func getExitCode(err error) (int, error) {
 	return ws.ExitStatus(), nil
 }
 
-// TryAddState attempts to add a state derived from the runErr.
-// If any error occurs, it is guaranteed to be returned along with
-// the initial metric slice.
-func TryAddState(runErr error, metrics []telegraf.Metric) ([]telegraf.Metric, error) {
-	state, err := getExitCode(runErr)
-	if err != nil {
-		return metrics, fmt.Errorf("exec: get exit code: %s", err)
+// AddState adds a state derived from the runErr. Unknown state will be set as fallback.
+// If any error occurs, it is guaranteed to be added to the service output.
+// An updated slice of metrics will be returned.
+func AddState(runErr error, errMessage []byte, metrics []telegraf.Metric) []telegraf.Metric {
+	state, exitErr := getExitCode(runErr)
+	// This will ensure that in every error case the valid nagios state 'unknown' will be returned.
+	// No error needs to be thrown because the output will contain the error information.
+	// Description found at 'Plugin Return Codes' https://nagios-plugins.org/doc/guidelines.html
+	if exitErr != nil || state < 0 || state > unknownExitCode {
+		state = unknownExitCode
 	}
 
 	for _, m := range metrics {
 		if m.Name() == "nagios_state" {
 			m.AddField("state", state)
-			return metrics, nil
+
+			if state == unknownExitCode {
+				errorMessage := string(errMessage)
+				if exitErr != nil && exitErr.Error() != "" {
+					errorMessage = exitErr.Error()
+				}
+				value, ok := m.GetField("service_output")
+				if !ok || value == "" {
+					// By adding the error message as output, the metric contains all needed information to understand
+					// the problem and fix it
+					m.AddField("service_output", errorMessage)
+				}
+			}
+			return metrics
 		}
 	}
 
@@ -66,8 +82,7 @@ func TryAddState(runErr error, metrics []telegraf.Metric) ([]telegraf.Metric, er
 	}
 	m := metric.New("nagios_state", nil, f, ts)
 
-	metrics = append(metrics, m)
-	return metrics, nil
+	return append(metrics, m)
 }
 
 type NagiosParser struct {

plugins/parsers/nagios/parser_test.go

@@ -32,7 +32,7 @@ func TestGetExitCode(t *testing.T) {
 			errF: func() error {
 				return errors.New("I am not *exec.ExitError")
 			},
-			expCode: 0,
+			expCode: 3,
 			expErr:  errors.New("I am not *exec.ExitError"),
 		},
 	}
@@ -89,10 +89,11 @@ func assertEqual(t *testing.T, exp, actual []telegraf.Metric) {
 
 func TestTryAddState(t *testing.T) {
 	tests := []struct {
-		name    string
-		runErrF func() error
-		metrics []telegraf.Metric
-		assertF func(*testing.T, []telegraf.Metric, error)
+		name          string
+		runErrF       func() error
+		runErrMessage []byte
+		metrics       []telegraf.Metric
+		assertF       func(*testing.T, []telegraf.Metric)
 	}{
 		{
 			name: "should append state=0 field to existing metric",
@@ -107,7 +108,7 @@ func TestTryAddState(t *testing.T) {
 					n("nagios_state").
 					f("service_output", "OK: system working").b(),
 			},
-			assertF: func(t *testing.T, metrics []telegraf.Metric, err error) {
+			assertF: func(t *testing.T, metrics []telegraf.Metric) {
 				exp := []telegraf.Metric{
 					mb().
 						n("nagios").
@@ -118,7 +119,6 @@ func TestTryAddState(t *testing.T) {
 						f("state", 0).b(),
 				}
 				assertEqual(t, exp, metrics)
-				require.NoError(t, err)
 			},
 		},
 		{
@@ -131,7 +131,7 @@ func TestTryAddState(t *testing.T) {
 					n("nagios").
 					f("perfdata", 0).b(),
 			},
-			assertF: func(t *testing.T, metrics []telegraf.Metric, err error) {
+			assertF: func(t *testing.T, metrics []telegraf.Metric) {
 				exp := []telegraf.Metric{
 					mb().
 						n("nagios").
@@ -141,7 +141,6 @@ func TestTryAddState(t *testing.T) {
 						f("state", 0).b(),
 				}
 				assertEqual(t, exp, metrics)
-				require.NoError(t, err)
 			},
 		},
 		{
@@ -150,45 +149,62 @@ func TestTryAddState(t *testing.T) {
 				return nil
 			},
 			metrics: []telegraf.Metric{},
-			assertF: func(t *testing.T, metrics []telegraf.Metric, err error) {
+			assertF: func(t *testing.T, metrics []telegraf.Metric) {
 				require.Len(t, metrics, 1)
 				m := metrics[0]
 				require.Equal(t, "nagios_state", m.Name())
 				s, ok := m.GetField("state")
 				require.True(t, ok)
 				require.Equal(t, int64(0), s)
 				require.WithinDuration(t, time.Now().UTC(), m.Time(), 10*time.Second)
-				require.NoError(t, err)
 			},
 		},
 		{
-			name: "should return original metrics and an error",
+			name: "should return metrics with state unknown and thrown error is service_output",
 			runErrF: func() error {
 				return errors.New("non parsable error")
 			},
 			metrics: []telegraf.Metric{
 				mb().
-					n("nagios").
-					f("perfdata", 0).b(),
+					n("nagios_state").b(),
 			},
-			assertF: func(t *testing.T, metrics []telegraf.Metric, err error) {
+			assertF: func(t *testing.T, metrics []telegraf.Metric) {
 				exp := []telegraf.Metric{
 					mb().
-						n("nagios").
-						f("perfdata", 0).b(),
+						n("nagios_state").
+						f("state", 3).
+						f("service_output", "non parsable error").b(),
 				}
-				expErr := "exec: get exit code: non parsable error"
 
 				assertEqual(t, exp, metrics)
-				require.Equal(t, expErr, err.Error())
+			},
+		},
+		{
+			name: "should return metrics with state unknown and service_output error from error message parameter",
+			runErrF: func() error {
+				return errors.New("")
+			},
+			runErrMessage: []byte("some error message"),
+			metrics: []telegraf.Metric{
+				mb().
+					n("nagios_state").b(),
+			},
+			assertF: func(t *testing.T, metrics []telegraf.Metric) {
+				exp := []telegraf.Metric{
+					mb().
+						n("nagios_state").
+						f("state", 3).
+						f("service_output", "some error message").b(),
+				}
+				assertEqual(t, exp, metrics)
 			},
 		},
 	}
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			metrics, err := TryAddState(tt.runErrF(), tt.metrics)
-			tt.assertF(t, metrics, err)
+			metrics := AddState(tt.runErrF(), tt.runErrMessage, tt.metrics)
+			tt.assertF(t, metrics)
 		})
 	}
 }