Merge pull request #253 from prometheus/superq/md_raid0

Add support for raid0 devices in mdadm_linux collector.

collector/fixtures/e2e-output.txt

@@ -486,6 +486,7 @@ node_load5 0.37
 # HELP node_md_blocks Total number of blocks on device.
 # TYPE node_md_blocks gauge
 node_md_blocks{device="md0"} 248896
+node_md_blocks{device="md10"} 3.14159265e+08
 node_md_blocks{device="md127"} 3.12319552e+08
 node_md_blocks{device="md3"} 5.853468288e+09
 node_md_blocks{device="md4"} 4.883648e+06
@@ -496,6 +497,7 @@ node_md_blocks{device="md9"} 523968
 # HELP node_md_blocks_synced Number of blocks synced on device.
 # TYPE node_md_blocks_synced gauge
 node_md_blocks_synced{device="md0"} 248896
+node_md_blocks_synced{device="md10"} 3.14159265e+08
 node_md_blocks_synced{device="md127"} 3.12319552e+08
 node_md_blocks_synced{device="md3"} 5.853468288e+09
 node_md_blocks_synced{device="md4"} 4.883648e+06
@@ -506,6 +508,7 @@ node_md_blocks_synced{device="md9"} 523968
 # HELP node_md_disks Total number of disks of device.
 # TYPE node_md_disks gauge
 node_md_disks{device="md0"} 2
+node_md_disks{device="md10"} 2
 node_md_disks{device="md127"} 2
 node_md_disks{device="md3"} 8
 node_md_disks{device="md4"} 2
@@ -516,6 +519,7 @@ node_md_disks{device="md9"} 4
 # HELP node_md_disks_active Number of active disks of device.
 # TYPE node_md_disks_active gauge
 node_md_disks_active{device="md0"} 2
+node_md_disks_active{device="md10"} 2
 node_md_disks_active{device="md127"} 2
 node_md_disks_active{device="md3"} 8
 node_md_disks_active{device="md4"} 2
@@ -526,6 +530,7 @@ node_md_disks_active{device="md9"} 4
 # HELP node_md_is_active Indicator whether the md-device is active or not.
 # TYPE node_md_is_active gauge
 node_md_is_active{device="md0"} 1
+node_md_is_active{device="md10"} 1
 node_md_is_active{device="md127"} 1
 node_md_is_active{device="md3"} 1
 node_md_is_active{device="md4"} 0

collector/fixtures/proc/mdstat

@@ -27,5 +27,7 @@ md9 : active raid1 sdc2[2] sdd2[3] sdb2[1] sda2[0]
       523968 blocks super 1.2 [4/4] [UUUU]
             resync=DELAYED
 
+md10 : active raid0 sda1[0] sdb1[1]
+      314159265 blocks 64k chunks
 
 unused devices: <none>

collector/fixtures/proc/mdstat_invalid

@@ -0,0 +1,5 @@
+Personalities : [invalid]
+md3 : invalid
+      314159265 blocks 64k chunks
+
+unused devices: <none>

collector/mdadm_linux.go

@@ -29,6 +29,7 @@ import (
 
 var (
 	statuslineRE = regexp.MustCompile(`(\d+) blocks .*\[(\d+)/(\d+)\] \[[U_]+\]`)
+	raid0lineRE  = regexp.MustCompile(`(\d+) blocks \d+k chunks`)
 	buildlineRE  = regexp.MustCompile(`\((\d+)/\d+\)`)
 )
 
@@ -76,6 +77,21 @@ func evalStatusline(statusline string) (active, total, size int64, err error) {
 	return active, total, size, nil
 }
 
+func evalRaid0line(statusline string) (size int64, err error) {
+	matches := raid0lineRE.FindStringSubmatch(statusline)
+
+	if len(matches) != 2 {
+		return 0, fmt.Errorf("invalid raid0 status line: %s", statusline)
+	}
+
+	size, err = strconv.ParseInt(matches[1], 10, 64)
+	if err != nil {
+		return 0, fmt.Errorf("%s in statusline: %s", err, statusline)
+	}
+
+	return size, nil
+}
+
 // Gets the size that has already been synced out of the sync-line.
 func evalBuildline(buildline string) (int64, error) {
 	matches := buildlineRE.FindStringSubmatch(buildline)
@@ -108,7 +124,11 @@ func parseMdstat(mdStatusFilePath string) ([]mdStatus, error) {
 	mdStatusFile := string(content)
 
 	lines := strings.Split(mdStatusFile, "\n")
-	var currentMD string
+	var (
+		currentMD           string
+		personality         string
+		active, total, size int64
+	)
 
 	// Each md has at least the deviceline, statusline and one empty line afterwards
 	// so we will have probably something of the order len(lines)/3 devices
@@ -133,17 +153,25 @@ func parseMdstat(mdStatusFilePath string) ([]mdStatus, error) {
 		}
 
 		mainLine := strings.Split(l, " ")
-		if len(mainLine) < 3 {
+		if len(mainLine) < 4 {
 			return mdStates, fmt.Errorf("error parsing mdline: %s", l)
 		}
-		currentMD = mainLine[0]               // name of md-device
-		isActive := (mainLine[2] == "active") // activity status of said md-device
+		currentMD = mainLine[0]               // The name of the md-device.
+		isActive := (mainLine[2] == "active") // The activity status of the md-device.
+		personality = mainLine[3]             // The personality type of the md-device.
 
 		if len(lines) <= i+3 {
 			return mdStates, fmt.Errorf("error parsing mdstat: entry for %s has fewer lines than expected", currentMD)
 		}
 
-		active, total, size, err := evalStatusline(lines[i+1]) // parse statusline, always present
+		switch personality {
+		case "raid0":
+			active = int64(len(mainLine) - 4)     // Get the number of devices from the main line.
+			total = active                        // Raid0 active and total is always the same if active.
+			size, err = evalRaid0line(lines[i+1]) // Parse statusline, always present.
+		default:
+			active, total, size, err = evalStatusline(lines[i+1]) // Parse statusline, always present.
+		}
 
 		if err != nil {
 			return mdStates, fmt.Errorf("error parsing mdstat: %s", err)

collector/mdadm_linux_test.go

@@ -33,6 +33,7 @@ func TestMdadm(t *testing.T) {
 		"md8":   {"md8", true, 2, 2, 195310144, 16775552},
 		"md7":   {"md7", true, 3, 4, 7813735424, 7813735424},
 		"md9":   {"md9", true, 4, 4, 523968, 523968},
+		"md10":  {"md10", true, 2, 2, 314159265, 314159265},
 	}
 
 	for _, md := range mdStates {
@@ -45,3 +46,11 @@ func TestMdadm(t *testing.T) {
 		t.Errorf("expected number of parsed md-device to be %d, but was %d", len(refs), len(mdStates))
 	}
 }
+
+func TestInvalidMdstat(t *testing.T) {
+	_, err := parseMdstat("fixtures/proc/mdstat_invalid")
+
+	if err == nil {
+		t.Fatalf("parsing of invalid reference file did not find any errors")
+	}
+}