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
+17
View File
@@ -3,8 +3,25 @@ package inspur
import (
"testing"
"time"
"git.mchus.pro/mchus/logpile/internal/models"
)
func TestDedupSELEvents(t *testing.T) {
ts := time.Date(2026, 7, 27, 4, 52, 55, 0, time.UTC)
events := []models.Event{
{Source: "SEL", Timestamp: ts, EventType: "Asserted", Description: "PSU0 fail"},
{Source: "SEL", Timestamp: ts, EventType: "Asserted", Description: "PSU0 fail"},
{Source: "BMC", Timestamp: ts, EventType: "Assert", Description: "recurring alarm"},
{Source: "BMC", Timestamp: ts, EventType: "Assert", Description: "recurring alarm"},
}
out := dedupSELEvents(events)
if len(out) != 3 {
t.Fatalf("expected 3 events (1 SEL dup removed, BMC untouched), got %d: %+v", len(out), out)
}
}
func TestParseSELListWithLocation_UsesProvidedTimezone(t *testing.T) {
content := []byte("sel elist:\n1,02/28/2026,04:18:18,Sensor X,Event,Asserted\n")
shanghai, err := time.LoadLocation("Asia/Shanghai")