fix(parser): support Inspur HGX dump_<serial>_<timestamp>/ onekeylog layout

Fixes #20. This onekeylog variant has no devicefrusdr.log at all: FRU/sensors
come from raw ipmitool text output, PCIe/GPU presence has a dedicated
structural snapshot, SEL lives at a different path, and BMC component
failures are logged separately from SEL/IDL.

- Fall back to component/fru.txt (same FRU block format as devicefrusdr.log)
  and component/sensor.txt / sdr.txt (ipmitool sensor list / sdr elist)
  when devicefrusdr.log is absent.
- Parse log/bmc/diagnose/OtrdDiagnoseComponent.json's PCIe Device Info array
  for GPU/PCIe presence and link state, independent of SEL/IDL alarm
  history; flag devices running below their negotiated max link
  speed/width as degraded with a Warning event.
- Fall back to log/sel.csv (same format as selelist.csv) when selelist.csv
  is absent.
- Parse log/bmc/commer-comp/{commerslot,commerhmc,commerswvr,commerswcpld}
  logs (including rotated *.tar.gz.N parts) into failure events, filtering
  known-noisy lines.
- Collapse SEL events duplicated across sources by
  (timestamp, event_type, description).
- Surface a CollectionError when FRU/sensors are still empty after all
  fallbacks, instead of silently returning an empty inventory.
- Fix ParseFRU: a later placeholder "Product Serial : 0" / "Product Part
  Number : NULL" line in the same FRU block (e.g. SCM_FRU) was overwriting
  an already-parsed real Board Serial/Part Number.

Verified against dump_23DB01633_20260727-1359.tar.gz (HGX B200, KR9288-X3):
fru 0→21, sensors 0→303, 8 GPUs present at Gen5 x16 in slots 100-107.

Deferred (not covered by this change): BIOS-change-settings context and
BIOS POST codes from the same layout — see bible-local/10-decisions.md
ADL-048.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This commit is contained in:
2026-07-29 10:05:58 +03:00
co-authored by Claude Sonnet 5
parent 6e1a8232ec
commit 74e6bf0578
12 changed files with 708 additions and 4 deletions
+104
View File
@@ -0,0 +1,104 @@
// Fallback inventory parsing for the per-file dump_<serial>_<timestamp>/ onekeylog
// layout, which has no combined devicefrusdr.log. Instead FRU and sensors live in
// separate ipmitool-output text files under component/.
package inspur
import (
"bufio"
"strconv"
"strings"
"git.mchus.pro/mchus/logpile/internal/models"
)
// ParseSensorList parses component/sensor.txt ("ipmitool sensor list" output).
// Line format: Name | Value | Unit | Status | lnr | lcr | lnc | unc | ucr | unr
func ParseSensorList(content []byte) []models.SensorReading {
var readings []models.SensorReading
scanner := bufio.NewScanner(strings.NewReader(string(content)))
for scanner.Scan() {
line := strings.TrimSpace(scanner.Text())
if line == "" || !strings.Contains(line, "|") {
continue
}
fields := strings.Split(line, "|")
if len(fields) < 4 {
continue
}
name := strings.TrimSpace(fields[0])
valueStr := strings.TrimSpace(fields[1])
unit := strings.TrimSpace(fields[2])
status := strings.TrimSpace(fields[3])
if name == "" {
continue
}
reading := models.SensorReading{
Name: name,
Status: status,
Unit: unit,
Type: determineSensorType(name),
}
if valueStr != "" && valueStr != "na" {
if v, err := strconv.ParseFloat(valueStr, 64); err == nil {
reading.Value = v
reading.RawValue = valueStr + " " + unit
} else {
reading.RawValue = valueStr
}
}
readings = append(readings, reading)
}
return readings
}
// ParseSDRElist parses component/sdr.txt ("ipmitool sdr elist" output), used as a
// fallback when component/sensor.txt is unavailable.
// Line format: Name | ID | Status | Entity | Reading (e.g. "21 degrees C" or "no reading")
func ParseSDRElist(content []byte) []models.SensorReading {
var readings []models.SensorReading
scanner := bufio.NewScanner(strings.NewReader(string(content)))
for scanner.Scan() {
line := strings.TrimSpace(scanner.Text())
if line == "" || !strings.Contains(line, "|") {
continue
}
fields := strings.Split(line, "|")
if len(fields) < 3 {
continue
}
name := strings.TrimSpace(fields[0])
status := strings.TrimSpace(fields[2])
if name == "" {
continue
}
reading := models.SensorReading{
Name: name,
Status: status,
Type: determineSensorType(name),
}
if len(fields) >= 5 {
readingStr := strings.TrimSpace(fields[4])
if readingStr != "" && readingStr != "no reading" {
reading.RawValue = readingStr
if m := valueRegex.FindStringSubmatch(readingStr); m != nil {
if v, err := strconv.ParseFloat(m[1], 64); err == nil {
reading.Value = v
reading.Unit = strings.TrimSpace(m[2])
}
}
}
}
readings = append(readings, reading)
}
return readings
}