drive-health-metrics/controller.go

package main

import (
	"regexp"
	"strconv"
	"strings"
)

// ctrlDrive holds the RAID-controller-side view of one physical drive — data
// smartctl cannot see (predictive-failure, firmware state, controller media/
// other error counters, physical enclosure:slot). Keyed for matching to a
// smartctl megaraid passthrough by DeviceID (== the megaraid,N index).
type ctrlDrive struct {
	DeviceID   string
	Enclosure  string
	Slot       string
	MediaErr   int
	OtherErr   int
	Predictive int
	SmartAlert bool
	FwState    string
	TempC      *int
	Inquiry    string // Inquiry is the legacy single-line MegaCLI/storcli inquiry (serial model fw).
	Model      string // Model is the structured identity (perccli2); used for controller-only drives.
	Serial     string
	Firmware   string
	Rotation   string // Rotation is "SSD"/"NVMe" derived from controller media/interface, when known.
}

// controllerIndex enumerates all RAID controllers found, preferring modern
// tools (storcli/perccli) then MegaCLI, and returns a DeviceID->ctrlDrive map.
// If no controller CLI is present (plain HBA / onboard SATA / NVMe) it returns
// an empty map — that's fine, smartctl still covers those drives directly.
func controllerIndex() map[string]ctrlDrive {
	idx := map[string]ctrlDrive{}

	// perccli2 (8.x) is JSON-native. Its plain-text "show all" adds a second
	// status column that breaks positional parsing, so query JSON ('J') and use
	// the dedicated parser. Tried first since it covers the newest controllers.
	for _, bin := range []string{"perccli2", "/opt/MegaRAID/perccli2/perccli2"} {
		p := lookPath(bin)
		if p == "" {
			continue
		}
		drives := parsePerccli2(run(p, "/call/eall/sall", "show", "all", "J"))
		for _, cd := range drives {
			mergeCtrl(idx, cd)
		}
		if len(drives) > 0 {
			break
		}
	}

	// storcli / perccli (classic) share the same text "show all" layout (perccli
	// is Dell's rebrand). Try each installed binary until one returns drives, so
	// a host with several tools present still resolves.
	for _, bin := range []string{"storcli64", "storcli", "perccli64", "perccli",
		"/opt/MegaRAID/storcli/storcli64", "/opt/MegaRAID/perccli/perccli64"} {
		p := lookPath(bin)
		if p == "" {
			continue
		}
		drives := parseStorcli(run(p, "/call/eall/sall", "show", "all"))
		for _, cd := range drives {
			mergeCtrl(idx, cd)
		}
		if len(drives) > 0 {
			break
		}
	}

	// MegaCLI (older controllers). Same try-until-data approach.
	for _, bin := range []string{"MegaCli64", "MegaCli", "megacli",
		"/opt/MegaRAID/MegaCli/MegaCli64", "/usr/sbin/megacli"} {
		p := lookPath(bin)
		if p == "" {
			continue
		}
		drives := parseMegacliPDList(run(p, "-PDList", "-aAll"))
		for _, cd := range drives {
			mergeCtrl(idx, cd)
		}
		if len(drives) > 0 {
			break
		}
	}
	return idx
}

// mergeCtrl records cd under its DeviceID, keeping the first writer so the
// preferred tool (queried earlier) wins and a later tool can't clobber it.
// Entries without a DeviceID are dropped — they can't be matched to a drive.
func mergeCtrl(idx map[string]ctrlDrive, cd ctrlDrive) {
	if cd.DeviceID == "" {
		return
	}
	if _, exists := idx[cd.DeviceID]; !exists {
		idx[cd.DeviceID] = cd
	}
}

// afterColon returns the trimmed text following the first colon, or "". It reads
// the "Key : Value" lines MegaCLI/storcli emit.
func afterColon(s string) string {
	if i := strings.Index(s, ":"); i >= 0 {
		return strings.TrimSpace(s[i+1:])
	}
	return ""
}

// parseMegacliPDList parses `MegaCli -PDList -aAll`. Record boundary is the
// "Enclosure Device ID" line.
func parseMegacliPDList(text string) []ctrlDrive {
	var drives []ctrlDrive
	var cur ctrlDrive
	have := false
	flush := func() {
		if have && (cur.DeviceID != "" || cur.Slot != "") {
			drives = append(drives, cur)
		}
	}
	for _, raw := range strings.Split(text, "\n") {
		s := strings.TrimSpace(raw)
		switch {
		case strings.HasPrefix(s, "Enclosure Device ID"):
			flush()
			cur = ctrlDrive{Enclosure: afterColon(s)}
			have = true
		case strings.HasPrefix(s, "Slot Number"):
			cur.Slot = afterColon(s)
		case strings.HasPrefix(s, "Device Id"):
			cur.DeviceID = afterColon(s)
		case strings.HasPrefix(s, "Media Error Count"):
			cur.MediaErr = atoiSafe(afterColon(s))
		case strings.HasPrefix(s, "Other Error Count"):
			cur.OtherErr = atoiSafe(afterColon(s))
		case strings.HasPrefix(s, "Predictive Failure Count"):
			cur.Predictive = atoiSafe(afterColon(s))
		// MegaCLI phrases this as "Drive has flagged a S.M.A.R.T alert : No";
		// the older "S.M.A.R.T alert flagged by drive" form is kept for safety.
		case strings.HasPrefix(s, "Drive has flagged a S.M.A.R.T alert"),
			strings.HasPrefix(s, "S.M.A.R.T alert flagged by drive"):
			cur.SmartAlert = strings.Contains(s, "Yes")
		case strings.HasPrefix(s, "Firmware state"):
			cur.FwState = afterColon(s)
		case strings.HasPrefix(s, "Drive Temperature"):
			if m := regexp.MustCompile(`(\d+)\s*C`).FindStringSubmatch(s); m != nil {
				cur.TempC = pInt(atoiSafe(m[1]))
			}
		case strings.HasPrefix(s, "Inquiry Data"):
			cur.Inquiry = afterColon(s)
		}
	}
	flush()
	return drives
}

// parseStorcli parses `storcli /call/eall/sall show all`. A physical drive is
// introduced by a bare summary header ("Drive /c0/e64/s0 :") followed by a
// table row ("64:0  22 Onln ...") that carries the DID (== the megaraid index
// smartctl uses) and the controller state. The same drive then repeats sub-
// section headers ("Drive .../s0 - Detailed Information :", "... State :",
// "... Device attributes :") that must NOT open a new record — only the bare
// summary header does — so detail fields ("Key = Value") accumulate into one
// record across those sections.
func parseStorcli(text string) []ctrlDrive {
	var drives []ctrlDrive
	var cur ctrlDrive
	have := false
	// hdr matches only the bare summary header (path then ": " at end), not the
	// "- Detailed Information"/"State"/"Device attributes" sub-section headers.
	hdr := regexp.MustCompile(`^Drive /c\d+/e(\d+)/s(\d+)\s*:$`)
	// row matches the summary table data row "EID:Slt DID State ..."; this
	// storcli version reports the DID here, never as a "DID = N" line.
	row := regexp.MustCompile(`^(\d+):(\d+)\s+(\d+)\s+(\S+)`)
	flush := func() {
		if have && (cur.DeviceID != "" || cur.Slot != "") {
			drives = append(drives, cur)
		}
	}
	kv := func(s string) (string, string, bool) {
		if i := strings.Index(s, "="); i >= 0 {
			return strings.TrimSpace(s[:i]), strings.TrimSpace(s[i+1:]), true
		}
		return "", "", false
	}
	for _, raw := range strings.Split(text, "\n") {
		s := strings.TrimSpace(raw)

		// New drive record: only the bare summary header opens one.
		if m := hdr.FindStringSubmatch(s); m != nil {
			flush()
			cur = ctrlDrive{Enclosure: m[1], Slot: m[2]}
			have = true
			continue
		}
		if !have {
			continue
		}
		// Summary table row supplies the DID and controller state.
		if m := row.FindStringSubmatch(s); m != nil && cur.DeviceID == "" {
			cur.DeviceID = m[3]
			cur.FwState = m[4]
			continue
		}
		k, v, ok := kv(s)
		if !ok {
			continue
		}
		switch k {
		case "DID":
			cur.DeviceID = v
		case "Media Error Count":
			cur.MediaErr = atoiSafe(v)
		case "Other Error Count":
			cur.OtherErr = atoiSafe(v)
		case "Predictive Failure Count":
			cur.Predictive = atoiSafe(v)
		case "S.M.A.R.T alert flagged by drive":
			cur.SmartAlert = strings.EqualFold(v, "Yes")
		case "Firmware state", "State":
			if cur.FwState == "" {
				cur.FwState = v
			}
		case "Drive Temperature":
			if m := regexp.MustCompile(`(\d+)\s*C`).FindStringSubmatch(v); m != nil {
				cur.TempC = pInt(atoiSafe(m[1]))
			}
		case "Model Number", "Manufacturer Identification":
			if cur.Inquiry == "" {
				cur.Inquiry = v
			}
		}
	}
	flush()
	return drives
}

// parsePerccli2 parses `perccli2 /call/eall/sall show all J` (JSON). perccli2
// (8.x) renames the classic DID to PID and splits the single State column into
// State (RAID role: Conf/UConf/GHS/JBOD) and Status (health: Online/Offline/
// Failed/Missing); the latter is what maps to FwState. Drives nest under
// Controllers[].Response Data.Drives List[]; health counters sit directly in
// "Drive Detailed Information" (SAS/SATA) or under its "LU/NS Properties" for
// NVMe namespaces.
func parsePerccli2(text string) []ctrlDrive {
	m := loadJSON(text)
	if m == nil {
		return nil
	}
	controllers, ok := m["Controllers"].([]interface{})
	if !ok {
		return nil
	}
	var drives []ctrlDrive
	for _, c := range controllers {
		cm, ok := c.(map[string]interface{})
		if !ok {
			continue
		}
		list, ok := jLeaf(cm, "Response Data", "Drives List").([]interface{})
		if !ok {
			continue
		}
		for _, it := range list {
			dm, ok := it.(map[string]interface{})
			if !ok {
				continue
			}
			info := jObj(dm, "Drive Information")
			if info == nil {
				continue
			}
			cd := ctrlDrive{}

			// Location + identity from the summary block.
			if es := jStr(info, "EID:Slt"); es != "" {
				if i := strings.Index(es, ":"); i >= 0 {
					cd.Enclosure = strings.TrimSpace(es[:i])
					cd.Slot = strings.TrimSpace(es[i+1:])
				}
			}
			if pid := jInt(info, "PID"); pid != nil {
				cd.DeviceID = strconv.Itoa(*pid)
			}
			cd.FwState = jStr(info, "Status") // Health verdict, not the RAID role.
			cd.Model = jStr(info, "Model")
			cd.Rotation = perccli2Rotation(jStr(info, "Intf"), jStr(info, "Med"))

			// Detail block: identity fallbacks, temperature, error counters.
			if detail := jObj(dm, "Drive Detailed Information"); detail != nil {
				cd.Serial = jStr(detail, "Serial Number")
				cd.Firmware = jStr(detail, "Firmware Revision Level")
				if cd.Model == "" {
					cd.Model = jStr(detail, "Model")
				}
				if t := jInt(detail, "Temperature(C)"); t != nil {
					cd.TempC = t
				}
				// Counters live in detail (SAS/SATA) or LU/NS Properties (NVMe).
				props := jObj(detail, "LU/NS Properties")
				ci := func(key string) int {
					if v := jInt(detail, key); v != nil {
						return *v
					}
					if props != nil {
						if v := jInt(props, key); v != nil {
							return *v
						}
					}
					return 0
				}
				cd.MediaErr = ci("Media Error Count")
				cd.OtherErr = ci("Other Error Count")
				cd.Predictive = ci("Predictive Failure Count")
			}

			if cd.DeviceID != "" || cd.Slot != "" {
				drives = append(drives, cd)
			}
		}
	}
	return drives
}

// perccli2Rotation maps a perccli2 interface/media pair to a rotation label.
func perccli2Rotation(intf, med string) string {
	switch {
	case strings.EqualFold(intf, "NVMe"):
		return "NVMe"
	case strings.EqualFold(med, "SSD"):
		return "SSD"
	default:
		return ""
	}
}

// atoiSafe extracts the first integer found in s (leading sign allowed) and
// returns 0 when none is present, since controller output often wraps the number
// in units or surrounding labels.
func atoiSafe(s string) int {
	s = strings.TrimSpace(s)
	if m := regexp.MustCompile(`-?\d+`).FindString(s); m != "" {
		if n, err := strconv.Atoi(m); err == nil {
			return n
		}
	}
	return 0
}