feat(inputs.smart): Collect SSD endurance information where available in smartctl (#11391)

plugins/inputs/smart/README.md

@@ -180,10 +180,13 @@ execute this script.
   - fields:
     - exit_status
     - health_ok
+    - media_wearout_indicator
+    - percent_lifetime_remain
     - read_error_rate
     - seek_error
     - temp_c
     - udma_crc_errors
+    - wear_leveling_count
 
 - smart_attribute:
   - tags:

plugins/inputs/smart/smart.go

@@ -87,6 +87,14 @@ var (
 		"199": "udma_crc_errors",
 	}
 
+	// There are some fields we're interested in which use the vendor specific device ids
+	// so we need to be able to match on name instead
+	deviceFieldNames = map[string]string{
+		"Percent_Lifetime_Remain": "percent_lifetime_remain",
+		"Wear_Leveling_Count":     "wear_leveling_count",
+		"Media_Wearout_Indicator": "media_wearout_indicator",
+	}
+
 	// to obtain metrics from smartctl
 	sasNVMeAttributes = map[string]struct {
 		ID    string
@@ -150,6 +158,10 @@ var (
 			Name:  "Percentage_Used",
 			Parse: parsePercentageInt,
 		},
+		"Percentage used endurance indicator": {
+			Name:  "Percentage_Used",
+			Parse: parsePercentageInt,
+		},
 		"Data Units Read": {
 			Name:  "Data_Units_Read",
 			Parse: parseDataUnits,
@@ -817,6 +829,16 @@ func (m *Smart) gatherDisk(acc telegraf.Accumulator, device string, wg *sync.Wai
 					deviceFields[field] = val
 				}
 			}
+
+			if len(attr) > 4 {
+				// If the attribute name matches on in deviceFieldNames
+				// save the value to a field
+				if field, ok := deviceFieldNames[attr[2]]; ok {
+					if val, err := parseRawValue(attr[4]); err == nil {
+						deviceFields[field] = val
+					}
+				}
+			}
 		} else {
 			// what was found is not a vendor attribute
 			if matches := sasNVMeAttr.FindStringSubmatch(line); len(matches) > 2 {

plugins/inputs/smart/smart_test.go

@@ -52,7 +52,7 @@ func TestGatherAttributes(t *testing.T) {
 			err := s.Gather(&acc)
 
 			require.NoError(t, err)
-			assert.Equal(t, 65, acc.NFields(), "Wrong number of fields gathered")
+			assert.Equal(t, 66, acc.NFields(), "Wrong number of fields gathered")
 
 			for _, test := range testsAda0Attributes {
 				acc.AssertContainsTaggedFields(t, "smart_attribute", test.fields, test.tags)
@@ -171,7 +171,7 @@ func TestGatherNoAttributes(t *testing.T) {
 		err := s.Gather(&acc)
 
 		require.NoError(t, err)
-		assert.Equal(t, 8, acc.NFields(), "Wrong number of fields gathered")
+		assert.Equal(t, 9, acc.NFields(), "Wrong number of fields gathered")
 		acc.AssertDoesNotContainMeasurement(t, "smart_attribute")
 
 		for _, test := range testsAda0Device {
@@ -264,6 +264,23 @@ func TestGatherHtSAS(t *testing.T) {
 	testutil.RequireMetricsEqual(t, testHtsasAtributtes, acc.GetTelegrafMetrics(), testutil.SortMetrics(), testutil.IgnoreTime())
 }
 
+func TestGatherLongFormEnduranceAttrib(t *testing.T) {
+	runCmd = func(timeout config.Duration, sudo bool, command string, args ...string) ([]byte, error) {
+		return []byte(mockHGST), nil
+	}
+
+	var (
+		acc = &testutil.Accumulator{}
+		wg  = &sync.WaitGroup{}
+	)
+
+	wg.Add(1)
+
+	sampleSmart.gatherDisk(acc, "", wg)
+	assert.Equal(t, 7, acc.NFields(), "Wrong number of fields gathered")
+	assert.Equal(t, uint64(5), acc.NMetrics(), "Wrong number of metrics gathered")
+}
+
 func TestGatherSSD(t *testing.T) {
 	runCmd = func(timeout config.Duration, sudo bool, command string, args ...string) ([]byte, error) {
 		return []byte(ssdInfoData), nil
@@ -276,7 +293,7 @@ func TestGatherSSD(t *testing.T) {
 
 	wg.Add(1)
 	sampleSmart.gatherDisk(acc, "", wg)
-	assert.Equal(t, 105, acc.NFields(), "Wrong number of fields gathered")
+	assert.Equal(t, 106, acc.NFields(), "Wrong number of fields gathered")
 	assert.Equal(t, uint64(26), acc.NMetrics(), "Wrong number of metrics gathered")
 }
 
@@ -292,7 +309,7 @@ func TestGatherSSDRaid(t *testing.T) {
 
 	wg.Add(1)
 	sampleSmart.gatherDisk(acc, "", wg)
-	assert.Equal(t, 74, acc.NFields(), "Wrong number of fields gathered")
+	assert.Equal(t, 75, acc.NFields(), "Wrong number of fields gathered")
 	assert.Equal(t, uint64(15), acc.NMetrics(), "Wrong number of metrics gathered")
 }
 
@@ -1394,11 +1411,12 @@ var (
 	}{
 		{
 			map[string]interface{}{
-				"exit_status":     int(0),
+				"exit_status":         int(0),
-				"health_ok":       bool(true),
+				"health_ok":           bool(true),
-				"read_error_rate": int64(0),
+				"read_error_rate":     int64(0),
-				"temp_c":          int64(34),
+				"temp_c":              int64(34),
-				"udma_crc_errors": int64(0),
+				"udma_crc_errors":     int64(0),
+				"wear_leveling_count": int64(185),
 			},
 			map[string]string{
 				"device":    "ada0",
@@ -1810,6 +1828,52 @@ ID# ATTRIBUTE_NAME          FLAGS    VALUE WORST THRESH FAIL RAW_VALUE
                             |||____ S speed/performance
                             ||_____ O updated online
                             |______ P prefailure warning
+`
+
+	mockHGST = `
+smartctl 6.6 2016-05-31 r4324 [x86_64-linux-4.9.0-3-amd64] (local build)
+Copyright (C) 2002-16, Bruce Allen, Christian Franke, www.smartmontools.org
+
+=== START OF INFORMATION SECTION ===
+Vendor:               HGST
+Product:              HUSMM1640ASS200
+Revision:             A360
+Compliance:           SPC-4
+User Capacity:        400,088,457,216 bytes [400 GB]
+Logical block size:   512 bytes
+Physical block size:  4096 bytes
+LU is resource provisioned, LBPRZ=1
+Rotation Rate:        Solid State Device
+Form Factor:          2.5 inches
+Logical Unit id:      0x5000cca04ec26364
+Serial number:        ZZZZZZZZZ
+Device type:          disk
+Transport protocol:   SAS (SPL-3)
+Local Time is:        Mon Nov  6 10:20:33 2017 CET
+SMART support is:     Available - device has SMART capability.
+SMART support is:     Enabled
+Temperature Warning:  Enabled
+Read Cache is:        Enabled
+Writeback Cache is:   Enabled
+
+=== START OF READ SMART DATA SECTION ===
+SMART Health Status: OK
+
+Percentage used endurance indicator: 0%
+Current Drive Temperature:     28 C
+Drive Trip Temperature:        70 C
+
+Manufactured in week 30 of year 2017
+Specified cycle count over device lifetime:  0
+Accumulated start-stop cycles:  0
+Specified load-unload count over device lifetime:  0
+Accumulated load-unload cycles:  0
+defect list format 6 unknown
+Elements in grown defect list: 0
+
+Vendor (Seagate) cache information
+  Blocks sent to initiator = 3400674574336
+
 `
 
 	htSASInfoData = `smartctl 6.6 2016-05-31 r4324 [x86_64-linux-4.15.18-12-pve] (local build)