Implement audit enrichments, TUI workflows, and production ISO scaffold
This commit is contained in:
@@ -28,6 +28,11 @@ func Run() schema.HardwareIngestRequest {
|
||||
snap.Memory = collectMemory()
|
||||
snap.Storage = collectStorage()
|
||||
snap.PCIeDevices = collectPCIe()
|
||||
snap.PCIeDevices = enrichPCIeWithNVIDIA(snap.PCIeDevices, snap.Board.SerialNumber)
|
||||
snap.PCIeDevices = enrichPCIeWithMellanox(snap.PCIeDevices)
|
||||
snap.PCIeDevices = enrichPCIeWithNICTelemetry(snap.PCIeDevices)
|
||||
snap.Storage = enrichStorageWithVROC(snap.Storage, snap.PCIeDevices)
|
||||
snap.Storage = appendUniqueStorage(snap.Storage, collectRAIDStorage(snap.PCIeDevices))
|
||||
snap.PowerSupplies = collectPSUs()
|
||||
|
||||
// remaining collectors added in steps 1.8 – 1.10
|
||||
|
||||
164
audit/internal/collector/nic_mellanox.go
Normal file
164
audit/internal/collector/nic_mellanox.go
Normal file
@@ -0,0 +1,164 @@
|
||||
package collector
|
||||
|
||||
import (
|
||||
"bee/audit/internal/schema"
|
||||
"log/slog"
|
||||
"os"
|
||||
"os/exec"
|
||||
"path/filepath"
|
||||
"strings"
|
||||
)
|
||||
|
||||
const mellanoxVendorID = 0x15b3
|
||||
|
||||
var (
|
||||
mstflintQuery = func(bdf string) (string, error) {
|
||||
out, err := exec.Command("mstflint", "-d", bdf, "q").Output()
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
return string(out), nil
|
||||
}
|
||||
|
||||
ethtoolInfoQuery = func(iface string) (string, error) {
|
||||
out, err := exec.Command("ethtool", "-i", iface).Output()
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
return string(out), nil
|
||||
}
|
||||
|
||||
netIfacesByBDF = listNetIfacesByBDF
|
||||
)
|
||||
|
||||
// enrichPCIeWithMellanox enriches Mellanox/NVIDIA Networking devices with
|
||||
// firmware/serial information from mstflint, with ethtool fallback for firmware.
|
||||
func enrichPCIeWithMellanox(devs []schema.HardwarePCIeDevice) []schema.HardwarePCIeDevice {
|
||||
enriched := 0
|
||||
for i := range devs {
|
||||
if !isMellanoxDevice(devs[i]) {
|
||||
continue
|
||||
}
|
||||
|
||||
bdf := ""
|
||||
if devs[i].BDF != nil {
|
||||
bdf = normalizePCIeBDF(*devs[i].BDF)
|
||||
}
|
||||
if bdf == "" {
|
||||
continue
|
||||
}
|
||||
|
||||
fw, serial := queryMellanoxFromMstflint(bdf)
|
||||
if fw == "" {
|
||||
fw = queryFirmwareFromEthtool(bdf)
|
||||
}
|
||||
|
||||
if fw != "" {
|
||||
devs[i].Firmware = &fw
|
||||
}
|
||||
if serial != "" {
|
||||
devs[i].SerialNumber = &serial
|
||||
}
|
||||
if fw != "" || serial != "" {
|
||||
enriched++
|
||||
}
|
||||
}
|
||||
|
||||
slog.Info("mellanox: enriched", "count", enriched)
|
||||
return devs
|
||||
}
|
||||
|
||||
func isMellanoxDevice(dev schema.HardwarePCIeDevice) bool {
|
||||
if dev.VendorID != nil && *dev.VendorID == mellanoxVendorID {
|
||||
return true
|
||||
}
|
||||
if dev.Manufacturer != nil {
|
||||
m := strings.ToLower(*dev.Manufacturer)
|
||||
if strings.Contains(m, "mellanox") || strings.Contains(m, "nvidia networking") {
|
||||
return true
|
||||
}
|
||||
}
|
||||
return false
|
||||
}
|
||||
|
||||
func queryMellanoxFromMstflint(bdf string) (firmware, serial string) {
|
||||
out, err := mstflintQuery(bdf)
|
||||
if err != nil {
|
||||
return "", ""
|
||||
}
|
||||
return parseMstflintQuery(out)
|
||||
}
|
||||
|
||||
func parseMstflintQuery(raw string) (firmware, serial string) {
|
||||
for _, line := range strings.Split(raw, "\n") {
|
||||
line = strings.TrimSpace(line)
|
||||
if line == "" {
|
||||
continue
|
||||
}
|
||||
idx := strings.Index(line, ":")
|
||||
if idx < 0 {
|
||||
continue
|
||||
}
|
||||
key := strings.ToLower(strings.TrimSpace(line[:idx]))
|
||||
val := strings.TrimSpace(line[idx+1:])
|
||||
switch key {
|
||||
case "fw version":
|
||||
if val != "" {
|
||||
firmware = val
|
||||
}
|
||||
case "board serial number":
|
||||
if val != "" {
|
||||
serial = val
|
||||
}
|
||||
}
|
||||
}
|
||||
return firmware, serial
|
||||
}
|
||||
|
||||
func queryFirmwareFromEthtool(bdf string) string {
|
||||
for _, iface := range netIfacesByBDF(bdf) {
|
||||
out, err := ethtoolInfoQuery(iface)
|
||||
if err != nil {
|
||||
continue
|
||||
}
|
||||
if fw := parseEthtoolFirmwareInfo(out); fw != "" {
|
||||
return fw
|
||||
}
|
||||
}
|
||||
return ""
|
||||
}
|
||||
|
||||
func parseEthtoolFirmwareInfo(raw string) string {
|
||||
for _, line := range strings.Split(raw, "\n") {
|
||||
line = strings.TrimSpace(line)
|
||||
if line == "" {
|
||||
continue
|
||||
}
|
||||
idx := strings.Index(line, ":")
|
||||
if idx < 0 {
|
||||
continue
|
||||
}
|
||||
key := strings.ToLower(strings.TrimSpace(line[:idx]))
|
||||
val := strings.TrimSpace(line[idx+1:])
|
||||
if key == "firmware-version" && val != "" {
|
||||
return val
|
||||
}
|
||||
}
|
||||
return ""
|
||||
}
|
||||
|
||||
func listNetIfacesByBDF(bdf string) []string {
|
||||
path := filepath.Join("/sys/bus/pci/devices", bdf, "net")
|
||||
entries, err := os.ReadDir(path)
|
||||
if err != nil {
|
||||
return nil
|
||||
}
|
||||
ifaces := make([]string, 0, len(entries))
|
||||
for _, e := range entries {
|
||||
if e.Name() == "" {
|
||||
continue
|
||||
}
|
||||
ifaces = append(ifaces, e.Name())
|
||||
}
|
||||
return ifaces
|
||||
}
|
||||
118
audit/internal/collector/nic_mellanox_test.go
Normal file
118
audit/internal/collector/nic_mellanox_test.go
Normal file
@@ -0,0 +1,118 @@
|
||||
package collector
|
||||
|
||||
import (
|
||||
"bee/audit/internal/schema"
|
||||
"fmt"
|
||||
"testing"
|
||||
)
|
||||
|
||||
func TestParseMstflintQuery(t *testing.T) {
|
||||
raw := `Device #1:
|
||||
----------
|
||||
FW Version: 28.39.1002
|
||||
Board Serial Number: MT1234ABC
|
||||
`
|
||||
fw, serial := parseMstflintQuery(raw)
|
||||
if fw != "28.39.1002" {
|
||||
t.Fatalf("firmware: got %q", fw)
|
||||
}
|
||||
if serial != "MT1234ABC" {
|
||||
t.Fatalf("serial: got %q", serial)
|
||||
}
|
||||
}
|
||||
|
||||
func TestParseEthtoolFirmwareInfo(t *testing.T) {
|
||||
raw := `driver: mlx5_core
|
||||
version: 6.6.31-0-lts
|
||||
firmware-version: 28.39.1002 (MT_0000000000)
|
||||
bus-info: 0000:18:00.0
|
||||
`
|
||||
fw := parseEthtoolFirmwareInfo(raw)
|
||||
if fw != "28.39.1002 (MT_0000000000)" {
|
||||
t.Fatalf("firmware: got %q", fw)
|
||||
}
|
||||
}
|
||||
|
||||
func TestEnrichPCIeWithMellanox_mstflint(t *testing.T) {
|
||||
origMst := mstflintQuery
|
||||
origEth := ethtoolInfoQuery
|
||||
origIfaces := netIfacesByBDF
|
||||
t.Cleanup(func() {
|
||||
mstflintQuery = origMst
|
||||
ethtoolInfoQuery = origEth
|
||||
netIfacesByBDF = origIfaces
|
||||
})
|
||||
|
||||
mstflintQuery = func(bdf string) (string, error) {
|
||||
if bdf != "0000:18:00.0" {
|
||||
t.Fatalf("unexpected bdf: %s", bdf)
|
||||
}
|
||||
return "FW Version: 28.39.1002\nBoard Serial Number: SN-MST-001\n", nil
|
||||
}
|
||||
ethtoolInfoQuery = func(string) (string, error) {
|
||||
t.Fatal("ethtool should not be called when mstflint succeeds")
|
||||
return "", nil
|
||||
}
|
||||
netIfacesByBDF = func(string) []string { return nil }
|
||||
|
||||
vendorID := mellanoxVendorID
|
||||
bdf := "0000:18:00.0"
|
||||
manufacturer := "Mellanox Technologies"
|
||||
devs := []schema.HardwarePCIeDevice{{
|
||||
VendorID: &vendorID,
|
||||
BDF: &bdf,
|
||||
Manufacturer: &manufacturer,
|
||||
}}
|
||||
|
||||
out := enrichPCIeWithMellanox(devs)
|
||||
if out[0].Firmware == nil || *out[0].Firmware != "28.39.1002" {
|
||||
t.Fatalf("firmware: got %v", out[0].Firmware)
|
||||
}
|
||||
if out[0].SerialNumber == nil || *out[0].SerialNumber != "SN-MST-001" {
|
||||
t.Fatalf("serial: got %v", out[0].SerialNumber)
|
||||
}
|
||||
}
|
||||
|
||||
func TestEnrichPCIeWithMellanox_fallbackEthtool(t *testing.T) {
|
||||
origMst := mstflintQuery
|
||||
origEth := ethtoolInfoQuery
|
||||
origIfaces := netIfacesByBDF
|
||||
t.Cleanup(func() {
|
||||
mstflintQuery = origMst
|
||||
ethtoolInfoQuery = origEth
|
||||
netIfacesByBDF = origIfaces
|
||||
})
|
||||
|
||||
mstflintQuery = func(string) (string, error) {
|
||||
return "", fmt.Errorf("mstflint not found")
|
||||
}
|
||||
netIfacesByBDF = func(bdf string) []string {
|
||||
if bdf != "0000:18:00.0" {
|
||||
t.Fatalf("unexpected bdf: %s", bdf)
|
||||
}
|
||||
return []string{"eth0"}
|
||||
}
|
||||
ethtoolInfoQuery = func(iface string) (string, error) {
|
||||
if iface != "eth0" {
|
||||
t.Fatalf("unexpected iface: %s", iface)
|
||||
}
|
||||
return "driver: mlx5_core\nfirmware-version: 28.40.1000\n", nil
|
||||
}
|
||||
|
||||
vendorID := mellanoxVendorID
|
||||
bdf := "0000:18:00.0"
|
||||
manufacturer := "NVIDIA Networking"
|
||||
devs := []schema.HardwarePCIeDevice{{
|
||||
VendorID: &vendorID,
|
||||
BDF: &bdf,
|
||||
Manufacturer: &manufacturer,
|
||||
}}
|
||||
|
||||
out := enrichPCIeWithMellanox(devs)
|
||||
if out[0].Firmware == nil || *out[0].Firmware != "28.40.1000" {
|
||||
t.Fatalf("firmware: got %v", out[0].Firmware)
|
||||
}
|
||||
if out[0].SerialNumber != nil {
|
||||
t.Fatalf("serial should stay nil without mstflint, got %v", out[0].SerialNumber)
|
||||
}
|
||||
}
|
||||
172
audit/internal/collector/nic_telemetry.go
Normal file
172
audit/internal/collector/nic_telemetry.go
Normal file
@@ -0,0 +1,172 @@
|
||||
package collector
|
||||
|
||||
import (
|
||||
"bee/audit/internal/schema"
|
||||
"log/slog"
|
||||
"os"
|
||||
"path/filepath"
|
||||
"regexp"
|
||||
"strconv"
|
||||
"strings"
|
||||
)
|
||||
|
||||
var (
|
||||
ethtoolModuleQuery = func(iface string) (string, error) {
|
||||
out, err := raidToolQuery("ethtool", "-m", iface)
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
return string(out), nil
|
||||
}
|
||||
readNetStatFile = func(iface, key string) (int64, error) {
|
||||
path := filepath.Join("/sys/class/net", iface, "statistics", key)
|
||||
raw, err := os.ReadFile(path)
|
||||
if err != nil {
|
||||
return 0, err
|
||||
}
|
||||
v, err := strconv.ParseInt(strings.TrimSpace(string(raw)), 10, 64)
|
||||
if err != nil {
|
||||
return 0, err
|
||||
}
|
||||
return v, nil
|
||||
}
|
||||
)
|
||||
|
||||
func enrichPCIeWithNICTelemetry(devs []schema.HardwarePCIeDevice) []schema.HardwarePCIeDevice {
|
||||
enriched := 0
|
||||
for i := range devs {
|
||||
if !isNICDevice(devs[i]) || devs[i].BDF == nil {
|
||||
continue
|
||||
}
|
||||
bdf := normalizePCIeBDF(*devs[i].BDF)
|
||||
if bdf == "" {
|
||||
continue
|
||||
}
|
||||
ifaces := netIfacesByBDF(bdf)
|
||||
if len(ifaces) == 0 {
|
||||
continue
|
||||
}
|
||||
iface := ifaces[0]
|
||||
|
||||
if devs[i].Firmware == nil {
|
||||
if out, err := ethtoolInfoQuery(iface); err == nil {
|
||||
if fw := parseEthtoolFirmwareInfo(out); fw != "" {
|
||||
devs[i].Firmware = &fw
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if devs[i].Telemetry == nil {
|
||||
devs[i].Telemetry = map[string]any{}
|
||||
}
|
||||
injectNICPacketStats(devs[i].Telemetry, iface)
|
||||
if out, err := ethtoolModuleQuery(iface); err == nil {
|
||||
injectSFPDOMTelemetry(devs[i].Telemetry, out)
|
||||
}
|
||||
if len(devs[i].Telemetry) == 0 {
|
||||
devs[i].Telemetry = nil
|
||||
} else {
|
||||
enriched++
|
||||
}
|
||||
}
|
||||
slog.Info("nic: telemetry enriched", "count", enriched)
|
||||
return devs
|
||||
}
|
||||
|
||||
func isNICDevice(dev schema.HardwarePCIeDevice) bool {
|
||||
if dev.DeviceClass == nil {
|
||||
return false
|
||||
}
|
||||
c := strings.ToLower(strings.TrimSpace(*dev.DeviceClass))
|
||||
return strings.Contains(c, "ethernet controller") ||
|
||||
strings.Contains(c, "network controller") ||
|
||||
strings.Contains(c, "infiniband controller")
|
||||
}
|
||||
|
||||
func injectNICPacketStats(dst map[string]any, iface string) {
|
||||
for _, key := range []string{"rx_packets", "tx_packets", "rx_errors", "tx_errors"} {
|
||||
if v, err := readNetStatFile(iface, key); err == nil {
|
||||
dst[key] = v
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func injectSFPDOMTelemetry(dst map[string]any, raw string) {
|
||||
parsed := parseSFPDOM(raw)
|
||||
for k, v := range parsed {
|
||||
dst[k] = v
|
||||
}
|
||||
}
|
||||
|
||||
var floatRe = regexp.MustCompile(`[-+]?[0-9]*\.?[0-9]+`)
|
||||
|
||||
func parseSFPDOM(raw string) map[string]any {
|
||||
out := map[string]any{}
|
||||
for _, line := range strings.Split(raw, "\n") {
|
||||
trimmed := strings.TrimSpace(line)
|
||||
if trimmed == "" {
|
||||
continue
|
||||
}
|
||||
idx := strings.Index(trimmed, ":")
|
||||
if idx < 0 {
|
||||
continue
|
||||
}
|
||||
key := strings.ToLower(strings.TrimSpace(trimmed[:idx]))
|
||||
val := strings.TrimSpace(trimmed[idx+1:])
|
||||
|
||||
switch {
|
||||
case strings.Contains(key, "module temperature"):
|
||||
if f, ok := firstFloat(val); ok {
|
||||
out["sfp_temperature_c"] = f
|
||||
}
|
||||
case strings.Contains(key, "laser output power"):
|
||||
if f, ok := dbmValue(val); ok {
|
||||
out["sfp_tx_power_dbm"] = f
|
||||
}
|
||||
case strings.Contains(key, "receiver signal"):
|
||||
if f, ok := dbmValue(val); ok {
|
||||
out["sfp_rx_power_dbm"] = f
|
||||
}
|
||||
case strings.Contains(key, "module voltage"):
|
||||
if f, ok := firstFloat(val); ok {
|
||||
out["sfp_voltage_v"] = f
|
||||
}
|
||||
case strings.Contains(key, "laser bias current"):
|
||||
if f, ok := firstFloat(val); ok {
|
||||
out["sfp_bias_ma"] = f
|
||||
}
|
||||
}
|
||||
}
|
||||
return out
|
||||
}
|
||||
|
||||
func firstFloat(raw string) (float64, bool) {
|
||||
m := floatRe.FindString(raw)
|
||||
if m == "" {
|
||||
return 0, false
|
||||
}
|
||||
v, err := strconv.ParseFloat(m, 64)
|
||||
if err != nil {
|
||||
return 0, false
|
||||
}
|
||||
return v, true
|
||||
}
|
||||
|
||||
func dbmValue(raw string) (float64, bool) {
|
||||
parts := strings.Split(strings.ToLower(raw), "dbm")
|
||||
if len(parts) == 0 {
|
||||
return 0, false
|
||||
}
|
||||
for i := len(parts) - 1; i >= 0; i-- {
|
||||
candidate := parts[i]
|
||||
matches := floatRe.FindAllString(candidate, -1)
|
||||
if len(matches) == 0 {
|
||||
continue
|
||||
}
|
||||
v, err := strconv.ParseFloat(matches[len(matches)-1], 64)
|
||||
if err == nil {
|
||||
return v, true
|
||||
}
|
||||
}
|
||||
return 0, false
|
||||
}
|
||||
51
audit/internal/collector/nic_telemetry_test.go
Normal file
51
audit/internal/collector/nic_telemetry_test.go
Normal file
@@ -0,0 +1,51 @@
|
||||
package collector
|
||||
|
||||
import "testing"
|
||||
|
||||
func TestParseSFPDOM(t *testing.T) {
|
||||
raw := `
|
||||
Module temperature : 41.23 C
|
||||
Module voltage : 3.30 V
|
||||
Laser bias current : 6.12 mA
|
||||
Laser output power : 0.4712 mW / -3.27 dBm
|
||||
Receiver signal average optical power : 0.4123 mW / -3.85 dBm
|
||||
`
|
||||
got := parseSFPDOM(raw)
|
||||
|
||||
if v, ok := got["sfp_temperature_c"].(float64); !ok || v != 41.23 {
|
||||
t.Fatalf("sfp_temperature_c mismatch: %#v", got["sfp_temperature_c"])
|
||||
}
|
||||
if v, ok := got["sfp_voltage_v"].(float64); !ok || v != 3.30 {
|
||||
t.Fatalf("sfp_voltage_v mismatch: %#v", got["sfp_voltage_v"])
|
||||
}
|
||||
if v, ok := got["sfp_bias_ma"].(float64); !ok || v != 6.12 {
|
||||
t.Fatalf("sfp_bias_ma mismatch: %#v", got["sfp_bias_ma"])
|
||||
}
|
||||
if v, ok := got["sfp_tx_power_dbm"].(float64); !ok || v != -3.27 {
|
||||
t.Fatalf("sfp_tx_power_dbm mismatch: %#v", got["sfp_tx_power_dbm"])
|
||||
}
|
||||
if v, ok := got["sfp_rx_power_dbm"].(float64); !ok || v != -3.85 {
|
||||
t.Fatalf("sfp_rx_power_dbm mismatch: %#v", got["sfp_rx_power_dbm"])
|
||||
}
|
||||
}
|
||||
|
||||
func TestDBMValue(t *testing.T) {
|
||||
tests := []struct {
|
||||
in string
|
||||
want float64
|
||||
ok bool
|
||||
}{
|
||||
{"0.4123 mW / -3.85 dBm", -3.85, true},
|
||||
{"-1.23 dBm", -1.23, true},
|
||||
{"not supported", 0, false},
|
||||
}
|
||||
for _, tt := range tests {
|
||||
got, ok := dbmValue(tt.in)
|
||||
if ok != tt.ok {
|
||||
t.Fatalf("dbmValue(%q) ok=%v want %v", tt.in, ok, tt.ok)
|
||||
}
|
||||
if ok && got != tt.want {
|
||||
t.Fatalf("dbmValue(%q)=%v want %v", tt.in, got, tt.want)
|
||||
}
|
||||
}
|
||||
}
|
||||
245
audit/internal/collector/nvidia.go
Normal file
245
audit/internal/collector/nvidia.go
Normal file
@@ -0,0 +1,245 @@
|
||||
package collector
|
||||
|
||||
import (
|
||||
"bee/audit/internal/schema"
|
||||
"encoding/csv"
|
||||
"fmt"
|
||||
"log/slog"
|
||||
"os/exec"
|
||||
"strconv"
|
||||
"strings"
|
||||
)
|
||||
|
||||
const nvidiaVendorID = 0x10de
|
||||
|
||||
type nvidiaGPUInfo struct {
|
||||
BDF string
|
||||
Serial string
|
||||
VBIOS string
|
||||
TemperatureC *float64
|
||||
PowerW *float64
|
||||
ECCUncorrected *int64
|
||||
ECCCorrected *int64
|
||||
HWSlowdown *bool
|
||||
}
|
||||
|
||||
// enrichPCIeWithNVIDIA enriches NVIDIA PCIe devices with data from nvidia-smi.
|
||||
// If the driver/tool is unavailable, NVIDIA devices get UNKNOWN status and
|
||||
// a stable serial fallback based on board serial + slot.
|
||||
func enrichPCIeWithNVIDIA(devs []schema.HardwarePCIeDevice, boardSerial string) []schema.HardwarePCIeDevice {
|
||||
gpuByBDF, err := queryNVIDIAGPUs()
|
||||
if err != nil {
|
||||
slog.Info("nvidia: enrichment skipped", "err", err)
|
||||
return enrichPCIeWithNVIDIAData(devs, nil, boardSerial, false)
|
||||
}
|
||||
return enrichPCIeWithNVIDIAData(devs, gpuByBDF, boardSerial, true)
|
||||
}
|
||||
|
||||
func enrichPCIeWithNVIDIAData(devs []schema.HardwarePCIeDevice, gpuByBDF map[string]nvidiaGPUInfo, boardSerial string, driverLoaded bool) []schema.HardwarePCIeDevice {
|
||||
enriched := 0
|
||||
for i := range devs {
|
||||
if !isNVIDIADevice(devs[i]) {
|
||||
continue
|
||||
}
|
||||
|
||||
if !driverLoaded {
|
||||
setPCIeFallback(&devs[i], boardSerial)
|
||||
continue
|
||||
}
|
||||
|
||||
bdf := ""
|
||||
if devs[i].BDF != nil {
|
||||
bdf = normalizePCIeBDF(*devs[i].BDF)
|
||||
}
|
||||
info, ok := gpuByBDF[bdf]
|
||||
if !ok {
|
||||
setPCIeFallback(&devs[i], boardSerial)
|
||||
continue
|
||||
}
|
||||
|
||||
if v := strings.TrimSpace(info.Serial); v != "" {
|
||||
devs[i].SerialNumber = &v
|
||||
} else {
|
||||
setPCIeFallbackSerial(&devs[i], boardSerial)
|
||||
}
|
||||
if v := strings.TrimSpace(info.VBIOS); v != "" {
|
||||
devs[i].Firmware = &v
|
||||
}
|
||||
|
||||
status := "OK"
|
||||
if info.ECCUncorrected != nil && *info.ECCUncorrected > 0 {
|
||||
status = "WARNING"
|
||||
}
|
||||
devs[i].Status = &status
|
||||
injectNVIDIATelemetry(&devs[i], info)
|
||||
enriched++
|
||||
}
|
||||
|
||||
if driverLoaded {
|
||||
slog.Info("nvidia: enriched", "count", enriched)
|
||||
}
|
||||
return devs
|
||||
}
|
||||
|
||||
func queryNVIDIAGPUs() (map[string]nvidiaGPUInfo, error) {
|
||||
out, err := exec.Command(
|
||||
"nvidia-smi",
|
||||
"--query-gpu=index,pci.bus_id,serial,vbios_version,temperature.gpu,power.draw,ecc.errors.uncorrected.aggregate.total,ecc.errors.corrected.aggregate.total,clocks_throttle_reasons.hw_slowdown",
|
||||
"--format=csv,noheader,nounits",
|
||||
).Output()
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return parseNVIDIASMIQuery(string(out))
|
||||
}
|
||||
|
||||
func parseNVIDIASMIQuery(raw string) (map[string]nvidiaGPUInfo, error) {
|
||||
r := csv.NewReader(strings.NewReader(raw))
|
||||
r.TrimLeadingSpace = true
|
||||
r.FieldsPerRecord = -1
|
||||
|
||||
records, err := r.ReadAll()
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
result := make(map[string]nvidiaGPUInfo)
|
||||
for _, rec := range records {
|
||||
if len(rec) == 0 {
|
||||
continue
|
||||
}
|
||||
if len(rec) < 9 {
|
||||
return nil, fmt.Errorf("unexpected nvidia-smi columns: got %d, want 9", len(rec))
|
||||
}
|
||||
|
||||
bdf := normalizePCIeBDF(rec[1])
|
||||
if bdf == "" {
|
||||
continue
|
||||
}
|
||||
|
||||
info := nvidiaGPUInfo{
|
||||
BDF: bdf,
|
||||
Serial: strings.TrimSpace(rec[2]),
|
||||
VBIOS: strings.TrimSpace(rec[3]),
|
||||
TemperatureC: parseMaybeFloat(rec[4]),
|
||||
PowerW: parseMaybeFloat(rec[5]),
|
||||
ECCUncorrected: parseMaybeInt64(rec[6]),
|
||||
ECCCorrected: parseMaybeInt64(rec[7]),
|
||||
HWSlowdown: parseMaybeBool(rec[8]),
|
||||
}
|
||||
result[bdf] = info
|
||||
}
|
||||
|
||||
return result, nil
|
||||
}
|
||||
|
||||
func parseMaybeFloat(v string) *float64 {
|
||||
v = strings.TrimSpace(v)
|
||||
if v == "" || strings.EqualFold(v, "n/a") || strings.EqualFold(v, "not supported") || strings.EqualFold(v, "[not supported]") {
|
||||
return nil
|
||||
}
|
||||
n, err := strconv.ParseFloat(v, 64)
|
||||
if err != nil {
|
||||
return nil
|
||||
}
|
||||
return &n
|
||||
}
|
||||
|
||||
func parseMaybeInt64(v string) *int64 {
|
||||
v = strings.TrimSpace(v)
|
||||
if v == "" || strings.EqualFold(v, "n/a") || strings.EqualFold(v, "not supported") || strings.EqualFold(v, "[not supported]") {
|
||||
return nil
|
||||
}
|
||||
n, err := strconv.ParseInt(v, 10, 64)
|
||||
if err != nil {
|
||||
return nil
|
||||
}
|
||||
return &n
|
||||
}
|
||||
|
||||
func parseMaybeBool(v string) *bool {
|
||||
v = strings.TrimSpace(strings.ToLower(v))
|
||||
switch v {
|
||||
case "active", "enabled", "true", "1":
|
||||
b := true
|
||||
return &b
|
||||
case "not active", "disabled", "false", "0":
|
||||
b := false
|
||||
return &b
|
||||
default:
|
||||
return nil
|
||||
}
|
||||
}
|
||||
|
||||
func normalizePCIeBDF(bdf string) string {
|
||||
bdf = strings.TrimSpace(strings.ToLower(bdf))
|
||||
if bdf == "" {
|
||||
return ""
|
||||
}
|
||||
parts := strings.Split(bdf, ":")
|
||||
if len(parts) == 3 {
|
||||
domain := parts[0]
|
||||
if len(domain) > 4 {
|
||||
domain = domain[len(domain)-4:]
|
||||
}
|
||||
return domain + ":" + parts[1] + ":" + parts[2]
|
||||
}
|
||||
if len(parts) == 2 {
|
||||
return "0000:" + parts[0] + ":" + parts[1]
|
||||
}
|
||||
return bdf
|
||||
}
|
||||
|
||||
func isNVIDIADevice(dev schema.HardwarePCIeDevice) bool {
|
||||
if dev.VendorID != nil && *dev.VendorID == nvidiaVendorID {
|
||||
return true
|
||||
}
|
||||
if dev.Manufacturer != nil && strings.Contains(strings.ToLower(*dev.Manufacturer), "nvidia") {
|
||||
return true
|
||||
}
|
||||
return false
|
||||
}
|
||||
|
||||
func setPCIeFallback(dev *schema.HardwarePCIeDevice, boardSerial string) {
|
||||
setPCIeFallbackSerial(dev, boardSerial)
|
||||
status := "UNKNOWN"
|
||||
dev.Status = &status
|
||||
}
|
||||
|
||||
func setPCIeFallbackSerial(dev *schema.HardwarePCIeDevice, boardSerial string) {
|
||||
if strings.TrimSpace(boardSerial) == "" || dev.SerialNumber != nil {
|
||||
return
|
||||
}
|
||||
slot := "unknown"
|
||||
if dev.BDF != nil && strings.TrimSpace(*dev.BDF) != "" {
|
||||
slot = strings.TrimSpace(*dev.BDF)
|
||||
} else if dev.Slot != nil && strings.TrimSpace(*dev.Slot) != "" {
|
||||
slot = strings.TrimSpace(*dev.Slot)
|
||||
}
|
||||
fb := fmt.Sprintf("%s-PCIE-%s", boardSerial, slot)
|
||||
dev.SerialNumber = &fb
|
||||
}
|
||||
|
||||
func injectNVIDIATelemetry(dev *schema.HardwarePCIeDevice, info nvidiaGPUInfo) {
|
||||
if dev.Telemetry == nil {
|
||||
dev.Telemetry = map[string]any{}
|
||||
}
|
||||
if info.TemperatureC != nil {
|
||||
dev.Telemetry["temperature_c"] = *info.TemperatureC
|
||||
}
|
||||
if info.PowerW != nil {
|
||||
dev.Telemetry["power_w"] = *info.PowerW
|
||||
}
|
||||
if info.ECCUncorrected != nil {
|
||||
dev.Telemetry["ecc_uncorrected_total"] = *info.ECCUncorrected
|
||||
}
|
||||
if info.ECCCorrected != nil {
|
||||
dev.Telemetry["ecc_corrected_total"] = *info.ECCCorrected
|
||||
}
|
||||
if info.HWSlowdown != nil {
|
||||
dev.Telemetry["hw_slowdown_active"] = *info.HWSlowdown
|
||||
}
|
||||
if len(dev.Telemetry) == 0 {
|
||||
dev.Telemetry = nil
|
||||
}
|
||||
}
|
||||
116
audit/internal/collector/nvidia_test.go
Normal file
116
audit/internal/collector/nvidia_test.go
Normal file
@@ -0,0 +1,116 @@
|
||||
package collector
|
||||
|
||||
import (
|
||||
"bee/audit/internal/schema"
|
||||
"testing"
|
||||
)
|
||||
|
||||
func TestParseNVIDIASMIQuery(t *testing.T) {
|
||||
raw := "0, 00000000:65:00.0, GPU-SERIAL-1, 96.00.1F.00.02, 54, 210.33, 0, 5, Not Active\n"
|
||||
byBDF, err := parseNVIDIASMIQuery(raw)
|
||||
if err != nil {
|
||||
t.Fatalf("parse failed: %v", err)
|
||||
}
|
||||
|
||||
gpu, ok := byBDF["0000:65:00.0"]
|
||||
if !ok {
|
||||
t.Fatalf("gpu by normalized bdf not found")
|
||||
}
|
||||
if gpu.Serial != "GPU-SERIAL-1" {
|
||||
t.Fatalf("serial: got %q", gpu.Serial)
|
||||
}
|
||||
if gpu.VBIOS != "96.00.1F.00.02" {
|
||||
t.Fatalf("vbios: got %q", gpu.VBIOS)
|
||||
}
|
||||
if gpu.ECCUncorrected == nil || *gpu.ECCUncorrected != 0 {
|
||||
t.Fatalf("ecc uncorrected: got %v", gpu.ECCUncorrected)
|
||||
}
|
||||
if gpu.HWSlowdown == nil || *gpu.HWSlowdown {
|
||||
t.Fatalf("hw slowdown: got %v, want false", gpu.HWSlowdown)
|
||||
}
|
||||
}
|
||||
|
||||
func TestNormalizePCIeBDF(t *testing.T) {
|
||||
tests := []struct {
|
||||
in string
|
||||
want string
|
||||
}{
|
||||
{"00000000:17:00.0", "0000:17:00.0"},
|
||||
{"0000:17:00.0", "0000:17:00.0"},
|
||||
{"17:00.0", "0000:17:00.0"},
|
||||
}
|
||||
for _, tt := range tests {
|
||||
got := normalizePCIeBDF(tt.in)
|
||||
if got != tt.want {
|
||||
t.Fatalf("normalizePCIeBDF(%q)=%q want %q", tt.in, got, tt.want)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func TestEnrichPCIeWithNVIDIAData_driverLoaded(t *testing.T) {
|
||||
vendorID := nvidiaVendorID
|
||||
bdf := "0000:65:00.0"
|
||||
manufacturer := "NVIDIA Corporation"
|
||||
status := "OK"
|
||||
devices := []schema.HardwarePCIeDevice{
|
||||
{
|
||||
VendorID: &vendorID,
|
||||
BDF: &bdf,
|
||||
Manufacturer: &manufacturer,
|
||||
Status: &status,
|
||||
},
|
||||
}
|
||||
|
||||
byBDF := map[string]nvidiaGPUInfo{
|
||||
"0000:65:00.0": {
|
||||
BDF: "0000:65:00.0",
|
||||
Serial: "GPU-ABC",
|
||||
VBIOS: "96.00.1F.00.02",
|
||||
ECCUncorrected: ptrInt64(2),
|
||||
ECCCorrected: ptrInt64(10),
|
||||
TemperatureC: ptrFloat(55.5),
|
||||
PowerW: ptrFloat(230.2),
|
||||
},
|
||||
}
|
||||
|
||||
out := enrichPCIeWithNVIDIAData(devices, byBDF, "BOARD-001", true)
|
||||
if out[0].SerialNumber == nil || *out[0].SerialNumber != "GPU-ABC" {
|
||||
t.Fatalf("serial: got %v", out[0].SerialNumber)
|
||||
}
|
||||
if out[0].Firmware == nil || *out[0].Firmware != "96.00.1F.00.02" {
|
||||
t.Fatalf("firmware: got %v", out[0].Firmware)
|
||||
}
|
||||
if out[0].Status == nil || *out[0].Status != "WARNING" {
|
||||
t.Fatalf("status: got %v", out[0].Status)
|
||||
}
|
||||
if out[0].Telemetry == nil {
|
||||
t.Fatal("expected telemetry")
|
||||
}
|
||||
if got, ok := out[0].Telemetry["ecc_uncorrected_total"].(int64); !ok || got != 2 {
|
||||
t.Fatalf("ecc_uncorrected_total: got %#v", out[0].Telemetry["ecc_uncorrected_total"])
|
||||
}
|
||||
}
|
||||
|
||||
func TestEnrichPCIeWithNVIDIAData_driverMissingFallback(t *testing.T) {
|
||||
vendorID := nvidiaVendorID
|
||||
bdf := "0000:17:00.0"
|
||||
manufacturer := "NVIDIA Corporation"
|
||||
devices := []schema.HardwarePCIeDevice{
|
||||
{
|
||||
VendorID: &vendorID,
|
||||
BDF: &bdf,
|
||||
Manufacturer: &manufacturer,
|
||||
},
|
||||
}
|
||||
|
||||
out := enrichPCIeWithNVIDIAData(devices, nil, "BOARD-123", false)
|
||||
if out[0].SerialNumber == nil || *out[0].SerialNumber != "BOARD-123-PCIE-0000:17:00.0" {
|
||||
t.Fatalf("fallback serial: got %v", out[0].SerialNumber)
|
||||
}
|
||||
if out[0].Status == nil || *out[0].Status != "UNKNOWN" {
|
||||
t.Fatalf("fallback status: got %v", out[0].Status)
|
||||
}
|
||||
}
|
||||
|
||||
func ptrInt64(v int64) *int64 { return &v }
|
||||
func ptrFloat(v float64) *float64 { return &v }
|
||||
@@ -37,12 +37,44 @@ func parseLspci(output string) []schema.HardwarePCIeDevice {
|
||||
val := strings.TrimSpace(line[idx+2:])
|
||||
fields[key] = val
|
||||
}
|
||||
if !shouldIncludePCIeDevice(fields["Class"]) {
|
||||
continue
|
||||
}
|
||||
dev := parseLspciDevice(fields)
|
||||
devs = append(devs, dev)
|
||||
}
|
||||
return devs
|
||||
}
|
||||
|
||||
func shouldIncludePCIeDevice(class string) bool {
|
||||
c := strings.ToLower(strings.TrimSpace(class))
|
||||
if c == "" {
|
||||
return true
|
||||
}
|
||||
|
||||
// Keep inventory focused on useful replaceable components, not chipset/virtual noise.
|
||||
excluded := []string{
|
||||
"host bridge",
|
||||
"isa bridge",
|
||||
"pci bridge",
|
||||
"ram memory",
|
||||
"system peripheral",
|
||||
"communication controller",
|
||||
"signal processing controller",
|
||||
"usb controller",
|
||||
"smbus",
|
||||
"audio device",
|
||||
"serial bus controller",
|
||||
"unassigned class",
|
||||
}
|
||||
for _, bad := range excluded {
|
||||
if strings.Contains(c, bad) {
|
||||
return false
|
||||
}
|
||||
}
|
||||
return true
|
||||
}
|
||||
|
||||
func parseLspciDevice(fields map[string]string) schema.HardwarePCIeDevice {
|
||||
dev := schema.HardwarePCIeDevice{}
|
||||
present := true
|
||||
|
||||
41
audit/internal/collector/pcie_filter_test.go
Normal file
41
audit/internal/collector/pcie_filter_test.go
Normal file
@@ -0,0 +1,41 @@
|
||||
package collector
|
||||
|
||||
import "testing"
|
||||
|
||||
func TestShouldIncludePCIeDevice(t *testing.T) {
|
||||
tests := []struct {
|
||||
class string
|
||||
want bool
|
||||
}{
|
||||
{"USB controller", false},
|
||||
{"System peripheral", false},
|
||||
{"Audio device", false},
|
||||
{"Host bridge", false},
|
||||
{"PCI bridge", false},
|
||||
{"SMBus", false},
|
||||
{"Ethernet controller", true},
|
||||
{"RAID bus controller", true},
|
||||
{"Non-Volatile memory controller", true},
|
||||
{"VGA compatible controller", true},
|
||||
}
|
||||
|
||||
for _, tt := range tests {
|
||||
got := shouldIncludePCIeDevice(tt.class)
|
||||
if got != tt.want {
|
||||
t.Fatalf("class %q include=%v want %v", tt.class, got, tt.want)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func TestParseLspci_filtersExcludedClasses(t *testing.T) {
|
||||
input := "Slot:\t0000:00:14.0\nClass:\tUSB controller\nVendor:\tIntel Corporation\nDevice:\tUSB 3.0\n\n" +
|
||||
"Slot:\t0000:65:00.0\nClass:\tVGA compatible controller\nVendor:\tNVIDIA Corporation\nDevice:\tH100\n\n"
|
||||
|
||||
devs := parseLspci(input)
|
||||
if len(devs) != 1 {
|
||||
t.Fatalf("expected 1 filtered device, got %d", len(devs))
|
||||
}
|
||||
if devs[0].DeviceClass == nil || *devs[0].DeviceClass != "VGA compatible controller" {
|
||||
t.Fatalf("unexpected remaining class: %v", devs[0].DeviceClass)
|
||||
}
|
||||
}
|
||||
748
audit/internal/collector/raid.go
Normal file
748
audit/internal/collector/raid.go
Normal file
@@ -0,0 +1,748 @@
|
||||
package collector
|
||||
|
||||
import (
|
||||
"bee/audit/internal/schema"
|
||||
"encoding/json"
|
||||
"log/slog"
|
||||
"os"
|
||||
"os/exec"
|
||||
"regexp"
|
||||
"sort"
|
||||
"strconv"
|
||||
"strings"
|
||||
)
|
||||
|
||||
const (
|
||||
vendorBroadcomLSI = 0x1000
|
||||
vendorAdaptec = 0x9005
|
||||
vendorHPE = 0x103c
|
||||
vendorIntel = 0x8086
|
||||
)
|
||||
|
||||
var raidToolQuery = func(name string, args ...string) ([]byte, error) {
|
||||
return exec.Command(name, args...).Output()
|
||||
}
|
||||
|
||||
var readMDStat = func() ([]byte, error) {
|
||||
return os.ReadFile("/proc/mdstat")
|
||||
}
|
||||
|
||||
// collectRAIDStorage collects physical disks behind RAID controllers that may
|
||||
// not be exposed as regular block devices.
|
||||
func collectRAIDStorage(pcie []schema.HardwarePCIeDevice) []schema.HardwareStorage {
|
||||
vendors := detectRAIDVendors(pcie)
|
||||
if len(vendors) == 0 {
|
||||
return nil
|
||||
}
|
||||
|
||||
var out []schema.HardwareStorage
|
||||
|
||||
if vendors[vendorBroadcomLSI] {
|
||||
if drives := collectStorcliDrives(); len(drives) > 0 {
|
||||
out = append(out, drives...)
|
||||
}
|
||||
if drives := collectSASIrcuDrives("sas3ircu"); len(drives) > 0 {
|
||||
out = append(out, drives...)
|
||||
}
|
||||
if drives := collectSASIrcuDrives("sas2ircu"); len(drives) > 0 {
|
||||
out = append(out, drives...)
|
||||
}
|
||||
}
|
||||
|
||||
if vendors[vendorAdaptec] {
|
||||
if drives := collectArcconfDrives(); len(drives) > 0 {
|
||||
out = append(out, drives...)
|
||||
}
|
||||
}
|
||||
if vendors[vendorHPE] {
|
||||
if drives := collectSSACLIDrives(); len(drives) > 0 {
|
||||
out = append(out, drives...)
|
||||
}
|
||||
}
|
||||
|
||||
if len(out) > 0 {
|
||||
slog.Info("raid: collected physical drives", "count", len(out))
|
||||
}
|
||||
return out
|
||||
}
|
||||
|
||||
func detectRAIDVendors(pcie []schema.HardwarePCIeDevice) map[int]bool {
|
||||
out := map[int]bool{}
|
||||
for _, dev := range pcie {
|
||||
if dev.VendorID == nil {
|
||||
continue
|
||||
}
|
||||
if isLikelyRAIDController(dev) {
|
||||
out[*dev.VendorID] = true
|
||||
}
|
||||
}
|
||||
return out
|
||||
}
|
||||
|
||||
func isLikelyRAIDController(dev schema.HardwarePCIeDevice) bool {
|
||||
if dev.DeviceClass == nil {
|
||||
return false
|
||||
}
|
||||
c := strings.ToLower(*dev.DeviceClass)
|
||||
return strings.Contains(c, "raid") ||
|
||||
strings.Contains(c, "sas") ||
|
||||
strings.Contains(c, "mass storage") ||
|
||||
strings.Contains(c, "serial attached scsi")
|
||||
}
|
||||
|
||||
func collectStorcliDrives() []schema.HardwareStorage {
|
||||
out, err := raidToolQuery("storcli64", "/call/eall/sall", "show", "all", "J")
|
||||
if err != nil {
|
||||
slog.Info("raid: storcli unavailable", "err", err)
|
||||
return nil
|
||||
}
|
||||
drives := parseStorcliDrivesJSON(out)
|
||||
if len(drives) == 0 {
|
||||
slog.Info("raid: storcli returned no drives")
|
||||
}
|
||||
return drives
|
||||
}
|
||||
|
||||
func collectSASIrcuDrives(tool string) []schema.HardwareStorage {
|
||||
out, err := raidToolQuery(tool, "list")
|
||||
if err != nil {
|
||||
slog.Info("raid: "+tool+" unavailable", "err", err)
|
||||
return nil
|
||||
}
|
||||
|
||||
var drives []schema.HardwareStorage
|
||||
for _, ctlID := range parseSASIrcuControllerIDs(string(out)) {
|
||||
raw, err := raidToolQuery(tool, strconv.Itoa(ctlID), "display")
|
||||
if err != nil {
|
||||
continue
|
||||
}
|
||||
drives = append(drives, parseSASIrcuDisplay(string(raw))...)
|
||||
}
|
||||
return drives
|
||||
}
|
||||
|
||||
func parseSASIrcuControllerIDs(raw string) []int {
|
||||
lines := strings.Split(raw, "\n")
|
||||
idsMap := map[int]bool{}
|
||||
for _, line := range lines {
|
||||
fields := strings.Fields(strings.TrimSpace(line))
|
||||
if len(fields) == 0 {
|
||||
continue
|
||||
}
|
||||
id, err := strconv.Atoi(fields[0])
|
||||
if err != nil {
|
||||
continue
|
||||
}
|
||||
idsMap[id] = true
|
||||
}
|
||||
var ids []int
|
||||
for id := range idsMap {
|
||||
ids = append(ids, id)
|
||||
}
|
||||
sort.Ints(ids)
|
||||
return ids
|
||||
}
|
||||
|
||||
func parseSASIrcuDisplay(raw string) []schema.HardwareStorage {
|
||||
var blocks []map[string]string
|
||||
var cur map[string]string
|
||||
var currentType string
|
||||
|
||||
for _, line := range strings.Split(raw, "\n") {
|
||||
trimmed := strings.TrimSpace(line)
|
||||
if strings.HasPrefix(trimmed, "Device is a ") {
|
||||
if cur != nil {
|
||||
cur["__device_type"] = currentType
|
||||
blocks = append(blocks, cur)
|
||||
}
|
||||
cur = map[string]string{}
|
||||
currentType = strings.TrimSpace(strings.TrimPrefix(trimmed, "Device is a "))
|
||||
continue
|
||||
}
|
||||
if cur == nil {
|
||||
continue
|
||||
}
|
||||
if idx := strings.Index(trimmed, ":"); idx > 0 {
|
||||
key := strings.TrimSpace(trimmed[:idx])
|
||||
val := strings.TrimSpace(trimmed[idx+1:])
|
||||
cur[key] = val
|
||||
}
|
||||
}
|
||||
if cur != nil {
|
||||
cur["__device_type"] = currentType
|
||||
blocks = append(blocks, cur)
|
||||
}
|
||||
|
||||
var out []schema.HardwareStorage
|
||||
for _, b := range blocks {
|
||||
dt := strings.ToLower(b["__device_type"])
|
||||
if !strings.Contains(dt, "hard disk") && !strings.Contains(dt, "ssd") && !strings.Contains(dt, "nvme") {
|
||||
continue
|
||||
}
|
||||
|
||||
present := true
|
||||
status := mapRAIDDriveStatus(b["State"])
|
||||
s := schema.HardwareStorage{Present: &present, Status: &status}
|
||||
|
||||
enclosure := strings.TrimSpace(b["Enclosure #"])
|
||||
slot := strings.TrimSpace(b["Slot #"])
|
||||
if enclosure != "" || slot != "" {
|
||||
v := enclosure + ":" + slot
|
||||
v = strings.Trim(v, ":")
|
||||
s.Slot = &v
|
||||
}
|
||||
|
||||
if v := strings.TrimSpace(b["Model Number"]); v != "" {
|
||||
s.Model = &v
|
||||
}
|
||||
if v := strings.TrimSpace(b["Serial No"]); v != "" {
|
||||
s.SerialNumber = &v
|
||||
}
|
||||
if v := strings.ToUpper(strings.TrimSpace(b["Protocol"])); v != "" {
|
||||
s.Interface = &v
|
||||
}
|
||||
|
||||
media := strings.ToUpper(strings.TrimSpace(b["Drive Type"]))
|
||||
if media == "" {
|
||||
media = strings.ToUpper(dt)
|
||||
}
|
||||
intf := ""
|
||||
if s.Interface != nil {
|
||||
intf = *s.Interface
|
||||
}
|
||||
devType := inferDriveType(media, intf)
|
||||
s.Type = &devType
|
||||
|
||||
if mb := parseSASIrcuMB(b["Size (in MB)/(in sectors)"]); mb > 0 {
|
||||
gb := mb / 1000
|
||||
if gb == 0 {
|
||||
gb = 1
|
||||
}
|
||||
s.SizeGB = &gb
|
||||
}
|
||||
|
||||
if s.Slot != nil || s.SerialNumber != nil || s.Model != nil {
|
||||
out = append(out, s)
|
||||
}
|
||||
}
|
||||
return out
|
||||
}
|
||||
|
||||
func parseSASIrcuMB(raw string) int {
|
||||
raw = strings.TrimSpace(raw)
|
||||
if raw == "" {
|
||||
return 0
|
||||
}
|
||||
head := strings.SplitN(raw, "/", 2)[0]
|
||||
n, err := strconv.Atoi(strings.TrimSpace(head))
|
||||
if err != nil {
|
||||
return 0
|
||||
}
|
||||
return n
|
||||
}
|
||||
|
||||
func collectArcconfDrives() []schema.HardwareStorage {
|
||||
raw, err := raidToolQuery("arcconf", "getconfig", "1", "pd")
|
||||
if err != nil {
|
||||
slog.Info("raid: arcconf unavailable", "err", err)
|
||||
return nil
|
||||
}
|
||||
return parseArcconfPhysicalDrives(string(raw))
|
||||
}
|
||||
|
||||
func parseArcconfPhysicalDrives(raw string) []schema.HardwareStorage {
|
||||
lines := strings.Split(raw, "\n")
|
||||
var blocks []map[string]string
|
||||
var cur map[string]string
|
||||
|
||||
for _, line := range lines {
|
||||
trimmed := strings.TrimSpace(line)
|
||||
if strings.HasPrefix(strings.ToLower(trimmed), "device #") {
|
||||
if cur != nil {
|
||||
blocks = append(blocks, cur)
|
||||
}
|
||||
cur = map[string]string{}
|
||||
continue
|
||||
}
|
||||
if cur == nil {
|
||||
continue
|
||||
}
|
||||
if idx := strings.Index(trimmed, ":"); idx > 0 {
|
||||
key := strings.TrimSpace(trimmed[:idx])
|
||||
val := strings.TrimSpace(trimmed[idx+1:])
|
||||
cur[key] = val
|
||||
}
|
||||
}
|
||||
if cur != nil {
|
||||
blocks = append(blocks, cur)
|
||||
}
|
||||
|
||||
var out []schema.HardwareStorage
|
||||
for _, b := range blocks {
|
||||
present := true
|
||||
status := mapRAIDDriveStatus(b["State"])
|
||||
s := schema.HardwareStorage{Present: &present, Status: &status}
|
||||
|
||||
if v := strings.TrimSpace(b["Reported Location"]); v != "" {
|
||||
s.Slot = &v
|
||||
}
|
||||
if v := strings.TrimSpace(b["Model"]); v != "" {
|
||||
s.Model = &v
|
||||
}
|
||||
if v := strings.TrimSpace(b["Serial number"]); v != "" {
|
||||
s.SerialNumber = &v
|
||||
}
|
||||
if gb := parseHumanSizeToGB(b["Total Size"]); gb > 0 {
|
||||
s.SizeGB = &gb
|
||||
}
|
||||
|
||||
intf := parseArcconfInterface(b["Transfer Speed"])
|
||||
if intf != "" {
|
||||
s.Interface = &intf
|
||||
}
|
||||
media := strings.ToUpper(strings.TrimSpace(b["SSD"]))
|
||||
if media == "YES" || media == "TRUE" {
|
||||
media = "SSD"
|
||||
}
|
||||
devType := inferDriveType(media, intf)
|
||||
s.Type = &devType
|
||||
|
||||
if s.Slot != nil || s.SerialNumber != nil || s.Model != nil {
|
||||
out = append(out, s)
|
||||
}
|
||||
}
|
||||
return out
|
||||
}
|
||||
|
||||
func parseArcconfInterface(raw string) string {
|
||||
u := strings.ToUpper(raw)
|
||||
switch {
|
||||
case strings.Contains(u, "SAS"):
|
||||
return "SAS"
|
||||
case strings.Contains(u, "SATA"):
|
||||
return "SATA"
|
||||
case strings.Contains(u, "NVME"):
|
||||
return "NVME"
|
||||
default:
|
||||
return ""
|
||||
}
|
||||
}
|
||||
|
||||
var ssacliPhysicalDriveLine = regexp.MustCompile(`(?i)^physicaldrive\s+(\S+)\s+\(([^)]*)\)$`)
|
||||
|
||||
func collectSSACLIDrives() []schema.HardwareStorage {
|
||||
raw, err := raidToolQuery("ssacli", "ctrl", "all", "show", "config", "detail")
|
||||
if err != nil {
|
||||
slog.Info("raid: ssacli unavailable", "err", err)
|
||||
return nil
|
||||
}
|
||||
return parseSSACLIPhysicalDrives(string(raw))
|
||||
}
|
||||
|
||||
func parseSSACLIPhysicalDrives(raw string) []schema.HardwareStorage {
|
||||
lines := strings.Split(raw, "\n")
|
||||
var out []schema.HardwareStorage
|
||||
var cur *schema.HardwareStorage
|
||||
|
||||
flush := func() {
|
||||
if cur == nil {
|
||||
return
|
||||
}
|
||||
if cur.Slot != nil || cur.SerialNumber != nil || cur.Model != nil {
|
||||
out = append(out, *cur)
|
||||
}
|
||||
cur = nil
|
||||
}
|
||||
|
||||
for _, line := range lines {
|
||||
trimmed := strings.TrimSpace(line)
|
||||
if trimmed == "" {
|
||||
continue
|
||||
}
|
||||
if m := ssacliPhysicalDriveLine.FindStringSubmatch(trimmed); len(m) == 3 {
|
||||
flush()
|
||||
present := true
|
||||
status := "UNKNOWN"
|
||||
s := schema.HardwareStorage{Present: &present, Status: &status}
|
||||
slot := m[1]
|
||||
s.Slot = &slot
|
||||
|
||||
meta := strings.Split(m[2], ",")
|
||||
if len(meta) > 0 {
|
||||
if gb := parseHumanSizeToGB(strings.TrimSpace(meta[0])); gb > 0 {
|
||||
s.SizeGB = &gb
|
||||
}
|
||||
}
|
||||
if len(meta) > 1 {
|
||||
intf := parseSSACLIInterface(meta[1])
|
||||
if intf != "" {
|
||||
s.Interface = &intf
|
||||
}
|
||||
devType := inferDriveType(strings.ToUpper(meta[1]), intf)
|
||||
s.Type = &devType
|
||||
}
|
||||
if len(meta) > 2 {
|
||||
st := mapRAIDDriveStatus(meta[len(meta)-1])
|
||||
s.Status = &st
|
||||
}
|
||||
cur = &s
|
||||
continue
|
||||
}
|
||||
if cur == nil {
|
||||
continue
|
||||
}
|
||||
if idx := strings.Index(trimmed, ":"); idx > 0 {
|
||||
key := strings.ToLower(strings.TrimSpace(trimmed[:idx]))
|
||||
val := strings.TrimSpace(trimmed[idx+1:])
|
||||
switch key {
|
||||
case "serial number":
|
||||
if val != "" {
|
||||
cur.SerialNumber = &val
|
||||
}
|
||||
case "model":
|
||||
if val != "" {
|
||||
cur.Model = &val
|
||||
}
|
||||
case "status":
|
||||
st := mapRAIDDriveStatus(val)
|
||||
cur.Status = &st
|
||||
}
|
||||
}
|
||||
}
|
||||
flush()
|
||||
return out
|
||||
}
|
||||
|
||||
func parseSSACLIInterface(raw string) string {
|
||||
u := strings.ToUpper(raw)
|
||||
switch {
|
||||
case strings.Contains(u, "SAS"):
|
||||
return "SAS"
|
||||
case strings.Contains(u, "SATA"):
|
||||
return "SATA"
|
||||
case strings.Contains(u, "NVME"):
|
||||
return "NVME"
|
||||
default:
|
||||
return ""
|
||||
}
|
||||
}
|
||||
|
||||
func parseStorcliDrivesJSON(raw []byte) []schema.HardwareStorage {
|
||||
var doc struct {
|
||||
Controllers []struct {
|
||||
ResponseData struct {
|
||||
DriveInformation []struct {
|
||||
EIDSlt string `json:"EID:Slt"`
|
||||
State string `json:"State"`
|
||||
Size string `json:"Size"`
|
||||
Intf string `json:"Intf"`
|
||||
Med string `json:"Med"`
|
||||
Model string `json:"Model"`
|
||||
SN string `json:"SN"`
|
||||
Sp string `json:"Sp"`
|
||||
Type string `json:"Type"`
|
||||
} `json:"Drive Information"`
|
||||
} `json:"Response Data"`
|
||||
} `json:"Controllers"`
|
||||
}
|
||||
if err := json.Unmarshal(raw, &doc); err != nil {
|
||||
slog.Warn("raid: parse storcli json failed", "err", err)
|
||||
return nil
|
||||
}
|
||||
|
||||
var drives []schema.HardwareStorage
|
||||
for _, ctl := range doc.Controllers {
|
||||
for _, d := range ctl.ResponseData.DriveInformation {
|
||||
if s := storcliDriveToStorage(d); s != nil {
|
||||
drives = append(drives, *s)
|
||||
}
|
||||
}
|
||||
}
|
||||
return drives
|
||||
}
|
||||
|
||||
func storcliDriveToStorage(d struct {
|
||||
EIDSlt string `json:"EID:Slt"`
|
||||
State string `json:"State"`
|
||||
Size string `json:"Size"`
|
||||
Intf string `json:"Intf"`
|
||||
Med string `json:"Med"`
|
||||
Model string `json:"Model"`
|
||||
SN string `json:"SN"`
|
||||
Sp string `json:"Sp"`
|
||||
Type string `json:"Type"`
|
||||
}) *schema.HardwareStorage {
|
||||
present := true
|
||||
status := mapRAIDDriveStatus(d.State)
|
||||
s := schema.HardwareStorage{
|
||||
Present: &present,
|
||||
Status: &status,
|
||||
}
|
||||
|
||||
if v := strings.TrimSpace(d.EIDSlt); v != "" {
|
||||
s.Slot = &v
|
||||
}
|
||||
if v := strings.TrimSpace(d.Model); v != "" {
|
||||
s.Model = &v
|
||||
}
|
||||
if v := strings.TrimSpace(d.SN); v != "" {
|
||||
s.SerialNumber = &v
|
||||
}
|
||||
if v := strings.TrimSpace(strings.ToUpper(d.Intf)); v != "" {
|
||||
s.Interface = &v
|
||||
}
|
||||
|
||||
devType := inferDriveType(strings.TrimSpace(strings.ToUpper(d.Med)), strings.TrimSpace(strings.ToUpper(d.Intf)))
|
||||
if devType != "" {
|
||||
s.Type = &devType
|
||||
}
|
||||
|
||||
if gb := parseHumanSizeToGB(d.Size); gb > 0 {
|
||||
s.SizeGB = &gb
|
||||
}
|
||||
|
||||
// return only meaningful records
|
||||
if s.Model == nil && s.SerialNumber == nil && s.Slot == nil {
|
||||
return nil
|
||||
}
|
||||
return &s
|
||||
}
|
||||
|
||||
func inferDriveType(med, intf string) string {
|
||||
switch {
|
||||
case strings.Contains(med, "SSD"):
|
||||
return "SSD"
|
||||
case strings.Contains(intf, "NVME"):
|
||||
return "NVMe"
|
||||
case strings.Contains(med, "HDD"):
|
||||
return "HDD"
|
||||
case strings.Contains(intf, "SAS") || strings.Contains(intf, "SATA"):
|
||||
return "HDD"
|
||||
default:
|
||||
return "Unknown"
|
||||
}
|
||||
}
|
||||
|
||||
func mapRAIDDriveStatus(raw string) string {
|
||||
u := strings.ToUpper(strings.TrimSpace(raw))
|
||||
switch {
|
||||
case strings.Contains(u, "OK"), strings.Contains(u, "OPTIMAL"), strings.Contains(u, "READY"):
|
||||
return "OK"
|
||||
case strings.Contains(u, "ONLN"), strings.Contains(u, "ONLINE"):
|
||||
return "OK"
|
||||
case strings.Contains(u, "RBLD"), strings.Contains(u, "REBUILD"):
|
||||
return "WARNING"
|
||||
case strings.Contains(u, "FAIL"), strings.Contains(u, "OFFLINE"):
|
||||
return "CRITICAL"
|
||||
default:
|
||||
return "UNKNOWN"
|
||||
}
|
||||
}
|
||||
|
||||
func parseHumanSizeToGB(raw string) int {
|
||||
parts := strings.Fields(strings.TrimSpace(raw))
|
||||
if len(parts) < 2 {
|
||||
return 0
|
||||
}
|
||||
value, err := strconv.ParseFloat(strings.TrimSpace(parts[0]), 64)
|
||||
if err != nil {
|
||||
return 0
|
||||
}
|
||||
unit := strings.ToUpper(parts[1])
|
||||
switch {
|
||||
case strings.HasPrefix(unit, "TB"):
|
||||
return int(value * 1000)
|
||||
case strings.HasPrefix(unit, "GB"):
|
||||
return int(value)
|
||||
case strings.HasPrefix(unit, "MB"):
|
||||
return int(value / 1000)
|
||||
default:
|
||||
return 0
|
||||
}
|
||||
}
|
||||
|
||||
func appendUniqueStorage(base, extra []schema.HardwareStorage) []schema.HardwareStorage {
|
||||
if len(extra) == 0 {
|
||||
return base
|
||||
}
|
||||
seen := map[string]bool{}
|
||||
for _, d := range base {
|
||||
seen[storageIdentityKey(d)] = true
|
||||
}
|
||||
for _, d := range extra {
|
||||
key := storageIdentityKey(d)
|
||||
if key == "" || seen[key] {
|
||||
continue
|
||||
}
|
||||
base = append(base, d)
|
||||
seen[key] = true
|
||||
}
|
||||
return base
|
||||
}
|
||||
|
||||
func storageIdentityKey(d schema.HardwareStorage) string {
|
||||
if d.SerialNumber != nil && strings.TrimSpace(*d.SerialNumber) != "" {
|
||||
return "sn:" + strings.ToLower(strings.TrimSpace(*d.SerialNumber))
|
||||
}
|
||||
if d.Model != nil && d.Slot != nil {
|
||||
return "modelslot:" + strings.ToLower(strings.TrimSpace(*d.Model)) + ":" + strings.ToLower(strings.TrimSpace(*d.Slot))
|
||||
}
|
||||
return ""
|
||||
}
|
||||
|
||||
type mdArray struct {
|
||||
Name string
|
||||
Degraded bool
|
||||
Members []string
|
||||
}
|
||||
|
||||
func enrichStorageWithVROC(storage []schema.HardwareStorage, pcie []schema.HardwarePCIeDevice) []schema.HardwareStorage {
|
||||
if !hasVROCController(pcie) {
|
||||
return storage
|
||||
}
|
||||
|
||||
raw, err := readMDStat()
|
||||
if err != nil {
|
||||
slog.Info("vroc: cannot read /proc/mdstat", "err", err)
|
||||
return storage
|
||||
}
|
||||
arrays := parseMDStatArrays(string(raw))
|
||||
if len(arrays) == 0 {
|
||||
slog.Info("vroc: no md arrays found")
|
||||
return storage
|
||||
}
|
||||
|
||||
serialToArray := map[string]mdArray{}
|
||||
for _, arr := range arrays {
|
||||
for _, member := range arr.Members {
|
||||
serial := queryDeviceSerial("/dev/" + member)
|
||||
if serial == "" {
|
||||
continue
|
||||
}
|
||||
serialToArray[strings.ToLower(serial)] = arr
|
||||
}
|
||||
}
|
||||
if len(serialToArray) == 0 {
|
||||
return storage
|
||||
}
|
||||
|
||||
updated := 0
|
||||
for i := range storage {
|
||||
if storage[i].SerialNumber == nil || strings.TrimSpace(*storage[i].SerialNumber) == "" {
|
||||
continue
|
||||
}
|
||||
arr, ok := serialToArray[strings.ToLower(strings.TrimSpace(*storage[i].SerialNumber))]
|
||||
if !ok {
|
||||
continue
|
||||
}
|
||||
if storage[i].Telemetry == nil {
|
||||
storage[i].Telemetry = map[string]any{}
|
||||
}
|
||||
storage[i].Telemetry["vroc_array"] = arr.Name
|
||||
storage[i].Telemetry["vroc_degraded"] = arr.Degraded
|
||||
if arr.Degraded {
|
||||
status := "WARNING"
|
||||
storage[i].Status = &status
|
||||
}
|
||||
updated++
|
||||
}
|
||||
|
||||
slog.Info("vroc: enriched storage members", "count", updated)
|
||||
return storage
|
||||
}
|
||||
|
||||
func hasVROCController(pcie []schema.HardwarePCIeDevice) bool {
|
||||
for _, dev := range pcie {
|
||||
if dev.VendorID == nil || *dev.VendorID != vendorIntel {
|
||||
continue
|
||||
}
|
||||
|
||||
class := ""
|
||||
if dev.DeviceClass != nil {
|
||||
class = strings.ToLower(*dev.DeviceClass)
|
||||
}
|
||||
model := ""
|
||||
if dev.Model != nil {
|
||||
model = strings.ToLower(*dev.Model)
|
||||
}
|
||||
|
||||
if strings.Contains(class, "raid") ||
|
||||
strings.Contains(model, "vroc") ||
|
||||
strings.Contains(model, "volume management device") ||
|
||||
strings.Contains(model, "vmd") {
|
||||
return true
|
||||
}
|
||||
}
|
||||
return false
|
||||
}
|
||||
|
||||
var mdHealthPattern = regexp.MustCompile(`\[[U_]+\]`)
|
||||
|
||||
func parseMDStatArrays(raw string) []mdArray {
|
||||
lines := strings.Split(raw, "\n")
|
||||
var arrays []mdArray
|
||||
var current *mdArray
|
||||
|
||||
for _, line := range lines {
|
||||
trimmed := strings.TrimSpace(line)
|
||||
if trimmed == "" {
|
||||
continue
|
||||
}
|
||||
|
||||
if strings.Contains(line, " : ") && !strings.HasPrefix(strings.TrimLeft(line, " \t"), "[") {
|
||||
left := strings.TrimSpace(strings.SplitN(line, " : ", 2)[0])
|
||||
if strings.EqualFold(left, "Personalities") || strings.EqualFold(left, "unused devices") {
|
||||
continue
|
||||
}
|
||||
if current != nil {
|
||||
arrays = append(arrays, *current)
|
||||
}
|
||||
|
||||
name := left
|
||||
fields := strings.Fields(strings.SplitN(line, " : ", 2)[1])
|
||||
|
||||
arr := mdArray{Name: name}
|
||||
for _, f := range fields {
|
||||
if i := strings.IndexByte(f, '['); i > 0 {
|
||||
member := strings.TrimSpace(f[:i])
|
||||
if member != "" {
|
||||
arr.Members = append(arr.Members, member)
|
||||
}
|
||||
}
|
||||
}
|
||||
current = &arr
|
||||
continue
|
||||
}
|
||||
|
||||
if current == nil {
|
||||
continue
|
||||
}
|
||||
if m := mdHealthPattern.FindString(trimmed); m != "" && strings.Contains(m, "_") {
|
||||
current.Degraded = true
|
||||
}
|
||||
}
|
||||
if current != nil {
|
||||
arrays = append(arrays, *current)
|
||||
}
|
||||
return arrays
|
||||
}
|
||||
|
||||
func queryDeviceSerial(devPath string) string {
|
||||
if out, err := exec.Command("nvme", "id-ctrl", devPath, "-o", "json").Output(); err == nil {
|
||||
var ctrl nvmeIDCtrl
|
||||
if json.Unmarshal(out, &ctrl) == nil {
|
||||
if v := cleanDMIValue(strings.TrimSpace(ctrl.SerialNumber)); v != "" {
|
||||
return v
|
||||
}
|
||||
}
|
||||
}
|
||||
if out, err := exec.Command("smartctl", "-j", "-i", devPath).Output(); err == nil {
|
||||
var info smartctlInfo
|
||||
if json.Unmarshal(out, &info) == nil {
|
||||
if v := cleanDMIValue(strings.TrimSpace(info.SerialNumber)); v != "" {
|
||||
return v
|
||||
}
|
||||
}
|
||||
}
|
||||
return ""
|
||||
}
|
||||
96
audit/internal/collector/raid_parsers_test.go
Normal file
96
audit/internal/collector/raid_parsers_test.go
Normal file
@@ -0,0 +1,96 @@
|
||||
package collector
|
||||
|
||||
import "testing"
|
||||
|
||||
func TestParseSASIrcuControllerIDs(t *testing.T) {
|
||||
raw := `LSI Corporation SAS2 IR Configuration Utility.
|
||||
Adapter List
|
||||
==============
|
||||
0 SAS2008(B2)
|
||||
1 SAS2308_2(D1)
|
||||
`
|
||||
ids := parseSASIrcuControllerIDs(raw)
|
||||
if len(ids) != 2 || ids[0] != 0 || ids[1] != 1 {
|
||||
t.Fatalf("unexpected ids: %#v", ids)
|
||||
}
|
||||
}
|
||||
|
||||
func TestParseSASIrcuDisplay(t *testing.T) {
|
||||
raw := `Device is a Hard disk
|
||||
Enclosure # : 32
|
||||
Slot # : 7
|
||||
State : Onln
|
||||
Size (in MB)/(in sectors) : 953869/1953525168
|
||||
Model Number : ST1000NM0033
|
||||
Serial No : Z1D12345
|
||||
Protocol : SAS
|
||||
Drive Type : HDD
|
||||
|
||||
Device is a Enclosure services device
|
||||
Enclosure # : 32
|
||||
`
|
||||
drives := parseSASIrcuDisplay(raw)
|
||||
if len(drives) != 1 {
|
||||
t.Fatalf("expected 1 drive, got %d", len(drives))
|
||||
}
|
||||
d := drives[0]
|
||||
if d.Slot == nil || *d.Slot != "32:7" {
|
||||
t.Fatalf("slot: %v", d.Slot)
|
||||
}
|
||||
if d.SerialNumber == nil || *d.SerialNumber != "Z1D12345" {
|
||||
t.Fatalf("serial: %v", d.SerialNumber)
|
||||
}
|
||||
if d.Interface == nil || *d.Interface != "SAS" {
|
||||
t.Fatalf("interface: %v", d.Interface)
|
||||
}
|
||||
if d.Status == nil || *d.Status != "OK" {
|
||||
t.Fatalf("status: %v", d.Status)
|
||||
}
|
||||
}
|
||||
|
||||
func TestParseArcconfPhysicalDrives(t *testing.T) {
|
||||
raw := `Device #0
|
||||
Reported Location : Channel 0, Device 3
|
||||
Model : Micron_5300
|
||||
Serial number : ARC12345
|
||||
State : Online
|
||||
Total Size : 894 GB
|
||||
Transfer Speed : SATA 6.0Gb/s
|
||||
SSD : Yes
|
||||
`
|
||||
drives := parseArcconfPhysicalDrives(raw)
|
||||
if len(drives) != 1 {
|
||||
t.Fatalf("expected 1 drive, got %d", len(drives))
|
||||
}
|
||||
d := drives[0]
|
||||
if d.Type == nil || *d.Type != "SSD" {
|
||||
t.Fatalf("type: %v", d.Type)
|
||||
}
|
||||
if d.Interface == nil || *d.Interface != "SATA" {
|
||||
t.Fatalf("interface: %v", d.Interface)
|
||||
}
|
||||
if d.Status == nil || *d.Status != "OK" {
|
||||
t.Fatalf("status: %v", d.Status)
|
||||
}
|
||||
}
|
||||
|
||||
func TestParseSSACLIPhysicalDrives(t *testing.T) {
|
||||
raw := `physicaldrive 1I:1:1 (894 GB, SAS HDD, OK)
|
||||
Serial Number: SSACLI001
|
||||
Model: MB8000JVYZQ
|
||||
|
||||
physicaldrive 1I:1:2 (894 GB, SAS HDD, Failed)
|
||||
Serial Number: SSACLI002
|
||||
Model: MB8000JVYZQ
|
||||
`
|
||||
drives := parseSSACLIPhysicalDrives(raw)
|
||||
if len(drives) != 2 {
|
||||
t.Fatalf("expected 2 drives, got %d", len(drives))
|
||||
}
|
||||
if drives[0].Status == nil || *drives[0].Status != "OK" {
|
||||
t.Fatalf("drive0 status: %v", drives[0].Status)
|
||||
}
|
||||
if drives[1].Status == nil || *drives[1].Status != "CRITICAL" {
|
||||
t.Fatalf("drive1 status: %v", drives[1].Status)
|
||||
}
|
||||
}
|
||||
57
audit/internal/collector/vroc_test.go
Normal file
57
audit/internal/collector/vroc_test.go
Normal file
@@ -0,0 +1,57 @@
|
||||
package collector
|
||||
|
||||
import (
|
||||
"bee/audit/internal/schema"
|
||||
"testing"
|
||||
)
|
||||
|
||||
func TestParseMDStatArrays(t *testing.T) {
|
||||
raw := `Personalities : [raid1]
|
||||
md126 : active raid1 nvme0n1[0] nvme1n1[1]
|
||||
976630464 blocks super external:/md127/0 [2/2] [UU]
|
||||
|
||||
md125 : active raid1 nvme2n1[0] nvme3n1[1]
|
||||
976630464 blocks super external:/md127/1 [2/1] [U_]
|
||||
`
|
||||
arrays := parseMDStatArrays(raw)
|
||||
if len(arrays) != 2 {
|
||||
t.Fatalf("expected 2 arrays, got %d", len(arrays))
|
||||
}
|
||||
if arrays[0].Name != "md126" || arrays[0].Degraded {
|
||||
t.Fatalf("unexpected array0: %+v", arrays[0])
|
||||
}
|
||||
if len(arrays[0].Members) != 2 || arrays[0].Members[0] != "nvme0n1" {
|
||||
t.Fatalf("unexpected members array0: %+v", arrays[0].Members)
|
||||
}
|
||||
if arrays[1].Name != "md125" || !arrays[1].Degraded {
|
||||
t.Fatalf("unexpected array1: %+v", arrays[1])
|
||||
}
|
||||
}
|
||||
|
||||
func TestHasVROCController(t *testing.T) {
|
||||
intel := vendorIntel
|
||||
model := "Volume Management Device NVMe RAID Controller"
|
||||
class := "RAID bus controller"
|
||||
tests := []struct {
|
||||
name string
|
||||
pcie []schema.HardwarePCIeDevice
|
||||
want bool
|
||||
}{
|
||||
{
|
||||
name: "intel vroc",
|
||||
pcie: []schema.HardwarePCIeDevice{{VendorID: &intel, Model: &model, DeviceClass: &class}},
|
||||
want: true,
|
||||
},
|
||||
{
|
||||
name: "non-intel raid",
|
||||
pcie: []schema.HardwarePCIeDevice{{}},
|
||||
want: false,
|
||||
},
|
||||
}
|
||||
for _, tt := range tests {
|
||||
got := hasVROCController(tt.pcie)
|
||||
if got != tt.want {
|
||||
t.Fatalf("%s: got %v want %v", tt.name, got, tt.want)
|
||||
}
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user