mchus/logpile
1
0
Fork 0
Code Issues 5 Pull Requests Actions Packages Projects Releases 13 Wiki Activity

fix(redfish): read PCIeInterface link width for GPU PCIe devices

Browse Source 
parseGPUWithSupplementalDocs did not read PCIeInterface from the device
doc, only from function docs. xFusion GPU PCIeCard entries carry link
width/speed in PCIeInterface (LanesInUse/Maxlanes/PCIeType/MaxPCIeType)
so GPU link width was always empty for xFusion servers.

Also apply the xFusion OEM function-level fallback for GPU function docs,
consistent with the NIC and PCIeDevice paths.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Mikhail Chusavitin
2026-04-12 12:48:52 +03:00
parent e10440ae32
commit becdca1d7e
2 changed files with 40 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
internal/collector/redfish.go
View File

@@ -4308,6 +4308,21 @@ func parseGPUWithSupplementalDocs(doc map[string]interface{}, functionDocs []map
gpu.DeviceID = asHexOrInt(doc["DeviceId"])
}
if pcieIf, ok := doc["PCIeInterface"].(map[string]interface{}); ok {
if gpu.CurrentLinkWidth == 0 {
gpu.CurrentLinkWidth = asInt(pcieIf["LanesInUse"])
}
if gpu.MaxLinkWidth == 0 {
gpu.MaxLinkWidth = firstNonZeroInt(asInt(pcieIf["MaxLanes"]), asInt(pcieIf["Maxlanes"]))
}
if gpu.CurrentLinkSpeed == "" {
gpu.CurrentLinkSpeed = firstNonEmpty(asString(pcieIf["PCIeType"]), asString(pcieIf["CurrentLinkSpeedGTs"]), asString(pcieIf["CurrentLinkSpeed"]))
}
if gpu.MaxLinkSpeed == "" {
gpu.MaxLinkSpeed = firstNonEmpty(asString(pcieIf["MaxPCIeType"]), asString(pcieIf["MaxLinkSpeedGTs"]), asString(pcieIf["MaxLinkSpeed"]))
}
}
for _, fn := range functionDocs {
if gpu.BDF == "" {
gpu.BDF = sanitizeRedfishBDF(asString(fn["FunctionId"]))
@@ -4330,6 +4345,9 @@ func parseGPUWithSupplementalDocs(doc map[string]interface{}, functionDocs []map
if gpu.CurrentLinkSpeed == "" {
gpu.CurrentLinkSpeed = firstNonEmpty(asString(fn["CurrentLinkSpeedGTs"]), asString(fn["CurrentLinkSpeed"]))
}
if gpu.CurrentLinkWidth == 0 || gpu.MaxLinkWidth == 0 || gpu.CurrentLinkSpeed == "" || gpu.MaxLinkSpeed == "" {
redfishEnrichFromOEMxFusionPCIeLink(fn, &gpu.CurrentLinkWidth, &gpu.MaxLinkWidth, &gpu.CurrentLinkSpeed, &gpu.MaxLinkSpeed)
}
}
if isMissingOrRawPCIModel(gpu.Model) {

22
internal/collector/redfish_test.go
View File

@@ -2815,6 +2815,28 @@ func TestReplayCollectGPUs_DedupUsesRedfishPathBeforeHeuristics(t *testing.T) {
}
}
func TestParseGPU_xFusionPCIeInterfaceMaxlanes(t *testing.T) {
// xFusion GPU PCIeDevices (PCIeCard1..N) carry link width in PCIeInterface
// with "Maxlanes" (lowercase 'l') rather than "MaxLanes".
doc := map[string]interface{}{
"Id": "PCIeCard1",
"Model": "RTX PRO 6000",
"PCIeInterface": map[string]interface{}{
"LanesInUse": 16,
"Maxlanes": 16,
"PCIeType": "Gen5",
"MaxPCIeType": "Gen5",
},
}
gpu := parseGPU(doc, nil, 1)
if gpu.CurrentLinkWidth != 16 || gpu.MaxLinkWidth != 16 {
t.Fatalf("expected link widths 16/16 from PCIeInterface, got current=%d max=%d", gpu.CurrentLinkWidth, gpu.MaxLinkWidth)
}
if gpu.CurrentLinkSpeed != "Gen5" || gpu.MaxLinkSpeed != "Gen5" {
t.Fatalf("expected link speeds Gen5/Gen5 from PCIeInterface, got current=%q max=%q", gpu.CurrentLinkSpeed, gpu.MaxLinkSpeed)
}
}
func TestParseGPU_UsesNestedOemSerialNumber(t *testing.T) {
doc := map[string]interface{}{
"Id": "GPU4",