Merge pull request #346 from mcdan/people/mcdan/issues/219

Fix additional mdadm parsing cases

Author: SuperQ

collector/fixtures/e2e-output.txt

@@ -713,10 +713,12 @@ 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="md00"} 4.186624e+06
 node_md_blocks{device="md10"} 3.14159265e+08
 node_md_blocks{device="md11"} 4.190208e+06
 node_md_blocks{device="md12"} 3.886394368e+09
 node_md_blocks{device="md127"} 3.12319552e+08
+node_md_blocks{device="md219"} 7932
 node_md_blocks{device="md3"} 5.853468288e+09
 node_md_blocks{device="md4"} 4.883648e+06
 node_md_blocks{device="md6"} 1.95310144e+08
@@ -726,10 +728,12 @@ 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="md00"} 4.186624e+06
 node_md_blocks_synced{device="md10"} 3.14159265e+08
 node_md_blocks_synced{device="md11"} 4.190208e+06
 node_md_blocks_synced{device="md12"} 3.886394368e+09
 node_md_blocks_synced{device="md127"} 3.12319552e+08
+node_md_blocks_synced{device="md219"} 7932
 node_md_blocks_synced{device="md3"} 5.853468288e+09
 node_md_blocks_synced{device="md4"} 4.883648e+06
 node_md_blocks_synced{device="md6"} 1.6775552e+07
@@ -739,10 +743,12 @@ 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="md00"} 1
 node_md_disks{device="md10"} 2
 node_md_disks{device="md11"} 2
 node_md_disks{device="md12"} 2
 node_md_disks{device="md127"} 2
+node_md_disks{device="md219"} 2
 node_md_disks{device="md3"} 8
 node_md_disks{device="md4"} 2
 node_md_disks{device="md6"} 2
@@ -752,10 +758,12 @@ 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="md00"} 1
 node_md_disks_active{device="md10"} 2
 node_md_disks_active{device="md11"} 2
 node_md_disks_active{device="md12"} 2
 node_md_disks_active{device="md127"} 2
+node_md_disks_active{device="md219"} 2
 node_md_disks_active{device="md3"} 8
 node_md_disks_active{device="md4"} 2
 node_md_disks_active{device="md6"} 1
@@ -765,10 +773,12 @@ 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="md00"} 1
 node_md_is_active{device="md10"} 1
 node_md_is_active{device="md11"} 1
 node_md_is_active{device="md12"} 1
 node_md_is_active{device="md127"} 1
+node_md_is_active{device="md219"} 0
 node_md_is_active{device="md3"} 1
 node_md_is_active{device="md4"} 0
 node_md_is_active{device="md6"} 1

collector/fixtures/proc/mdstat

@@ -1,13 +1,13 @@
 Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
 md3 : active raid6 sda1[8] sdh1[7] sdg1[6] sdf1[5] sde1[11] sdd1[3] sdc1[10] sdb1[9]
       5853468288 blocks super 1.2 level 6, 64k chunk, algorithm 2 [8/8] [UUUUUUUU]
-      
+
 md127 : active raid1 sdi2[0] sdj2[1]
       312319552 blocks [2/2] [UU]
-      
+
 md0 : active raid1 sdi1[0] sdj1[1]
       248896 blocks [2/2] [UU]
-      
+
 md4 : inactive raid1 sda3[0] sdb3[1]
       4883648 blocks [2/2] [UU]
 
@@ -18,7 +18,7 @@ md6 : active raid1 sdb2[2] sda2[0]
 md8 : active raid1 sdb1[1] sda1[0]
 	  195310144 blocks [2/2] [UU]
 	  [=>...................]  resync =  8.5% (16775552/195310144) finish=17.0min speed=259783K/sec
-	  
+
 md7 : active raid6 sdb1[0] sde1[3] sdd1[2] sdc1[1]
       7813735424 blocks super 1.2 level 6, 512k chunk, algorithm 2 [4/3] [U_UU]
       bitmap: 0/30 pages [0KB], 65536KB chunk
@@ -37,4 +37,10 @@ md11 : active (auto-read-only) raid1 sdb2[0] sdc2[1]
 md12 : active raid0 sdc2[0] sdd2[1]
       3886394368 blocks super 1.2 512k chunks
 
+md219 : inactive sdb[2](S) sdc[1](S) sda[0](S)
+    7932 blocks super external:imsm
+
+md00 : active raid0 xvdb[0]
+      4186624 blocks super 1.2 256k chunks
+
 unused devices: <none>

collector/mdadm_linux.go

@@ -28,9 +28,11 @@ import (
 )
 
 var (
-	statuslineRE = regexp.MustCompile(`(\d+) blocks .*\[(\d+)/(\d+)\] \[[U_]+\]`)
-	raid0lineRE  = regexp.MustCompile(`(\d+) blocks .*\d+k chunks`)
-	buildlineRE  = regexp.MustCompile(`\((\d+)/\d+\)`)
+	statuslineRE             = regexp.MustCompile(`(\d+) blocks .*\[(\d+)/(\d+)\] \[[U_]+\]`)
+	raid0lineRE              = regexp.MustCompile(`(\d+) blocks( super ([0-9\.])*)? \d+k chunks`)
+	buildlineRE              = regexp.MustCompile(`\((\d+)/\d+\)`)
+	unknownPersonalityLineRE = regexp.MustCompile(`(\d+) blocks (.*)`)
+	raidPersonalityRE        = regexp.MustCompile(`raid[0-9]+`)
 )
 
 type mdStatus struct {
@@ -80,7 +82,7 @@ func evalStatusline(statusline string) (active, total, size int64, err error) {
 func evalRaid0line(statusline string) (size int64, err error) {
 	matches := raid0lineRE.FindStringSubmatch(statusline)
 
-	if len(matches) != 2 {
+	if len(matches) < 2 {
 		return 0, fmt.Errorf("invalid raid0 status line: %s", statusline)
 	}
 
@@ -92,6 +94,21 @@ func evalRaid0line(statusline string) (size int64, err error) {
 	return size, nil
 }
 
+func evalUnknownPersonalitylineRE(statusline string) (size int64, err error) {
+	matches := unknownPersonalityLineRE.FindStringSubmatch(statusline)
+
+	if len(matches) != 2+1 {
+		return 0, fmt.Errorf("invalid unknown personality 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)
@@ -158,19 +175,28 @@ func parseMdstat(mdStatusFilePath string) ([]mdStatus, error) {
 		}
 		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.
+		personality = ""
+		for _, possiblePersonality := range mainLine[3:] {
+			if raidPersonalityRE.MatchString(possiblePersonality) {
+				personality = possiblePersonality
+				break
+			}
+		}
 
 		if len(lines) <= i+3 {
 			return mdStates, fmt.Errorf("error parsing mdstat: entry for %s has fewer lines than expected", currentMD)
 		}
 
-		switch personality {
-		case "raid0":
+		switch {
+		case personality == "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:
+		case raidPersonalityRE.MatchString(personality):
 			active, total, size, err = evalStatusline(lines[i+1]) // Parse statusline, always present.
+		default:
+			log.Infof("Personality unknown: %s\n", mainLine)
+			size, err = evalUnknownPersonalitylineRE(lines[i+1]) // Parse statusline, always present.
 		}
 
 		if err != nil {

collector/mdadm_linux_test.go

@@ -25,6 +25,7 @@ func TestMdadm(t *testing.T) {
 	}
 
 	refs := map[string]mdStatus{
+		// { "<name>", <active?>, <numDisksActive>, <totalNumDisks>, <amountSynced>, <totalSize>}
 		"md3":   {"md3", true, 8, 8, 5853468288, 5853468288},
 		"md127": {"md127", true, 2, 2, 312319552, 312319552},
 		"md0":   {"md0", true, 2, 2, 248896, 248896},
@@ -36,6 +37,8 @@ func TestMdadm(t *testing.T) {
 		"md10":  {"md10", true, 2, 2, 314159265, 314159265},
 		"md11":  {"md11", true, 2, 2, 4190208, 4190208},
 		"md12":  {"md12", true, 2, 2, 3886394368, 3886394368},
+		"md219": {"md219", false, 2, 2, 7932, 7932},
+		"md00":  {"md00", true, 1, 1, 4186624, 4186624},
 	}
 
 	for _, md := range mdStates {