Add GPU serial number extraction for NVIDIA diagnostics

Parse inventory/output.log to extract GPU serial numbers from lspci output,
expose them via serials API, and add GPU category to web UI.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
This commit is contained in:
2026-02-10 22:50:46 +03:00
parent 8b065c6cca
commit bcce975fd6
7 changed files with 536 additions and 1 deletions

View File

@@ -0,0 +1,145 @@
package nvidia
import (
"os"
"path/filepath"
"testing"
"git.mchus.pro/mchus/logpile/internal/parser"
)
func TestNVIDIAParser_RealArchive(t *testing.T) {
// Test with the real archive that was reported as problematic
archivePath := filepath.Join("../../../../example", "A514359X5A09844_logs-20260115-151707.tar")
// Check if file exists
if _, err := os.Stat(archivePath); os.IsNotExist(err) {
t.Skip("Test archive not found, skipping test")
}
// Extract files from archive
files, err := parser.ExtractArchive(archivePath)
if err != nil {
t.Fatalf("Failed to extract archive: %v", err)
}
// Check if inventory/output.log exists
hasInventoryLog := false
for _, f := range files {
if filepath.Base(f.Path) == "output.log" {
t.Logf("Found file: %s", f.Path)
}
if f.Path == "./inventory/output.log" || f.Path == "inventory/output.log" {
hasInventoryLog = true
t.Logf("Found inventory/output.log with %d bytes", len(f.Content))
}
}
if !hasInventoryLog {
t.Error("inventory/output.log not found in extracted files")
}
// Create parser and parse
p := &Parser{}
result, err := p.Parse(files)
if err != nil {
t.Fatalf("Failed to parse archive: %v", err)
}
// Verify basic system info
if result.Hardware.BoardInfo.Manufacturer == "" {
t.Error("Expected Manufacturer to be set")
}
if result.Hardware.BoardInfo.ProductName == "" {
t.Error("Expected ProductName to be set")
}
if result.Hardware.BoardInfo.SerialNumber == "" {
t.Error("Expected SerialNumber to be set")
}
t.Logf("System Info:")
t.Logf(" Manufacturer: %s", result.Hardware.BoardInfo.Manufacturer)
t.Logf(" Product: %s", result.Hardware.BoardInfo.ProductName)
t.Logf(" Serial: %s", result.Hardware.BoardInfo.SerialNumber)
// Verify GPUs were found
if len(result.Hardware.GPUs) == 0 {
t.Error("Expected to find GPUs")
}
t.Logf("\nFound %d GPUs:", len(result.Hardware.GPUs))
gpusWithSerials := 0
for _, gpu := range result.Hardware.GPUs {
t.Logf(" %s: %s (Firmware: %s, Serial: %s, BDF: %s)",
gpu.Slot, gpu.Model, gpu.Firmware, gpu.SerialNumber, gpu.BDF)
if gpu.SerialNumber != "" {
gpusWithSerials++
}
}
// Verify that GPU serial numbers were extracted
if gpusWithSerials == 0 {
t.Error("Expected at least some GPUs to have serial numbers")
}
t.Logf("\nGPUs with serial numbers: %d/%d", gpusWithSerials, len(result.Hardware.GPUs))
// Check events for SXM2 failures
t.Logf("\nTotal events: %d", len(result.Events))
// Look for the specific serial or SXM2
sxm2Events := 0
for _, event := range result.Events {
desc := event.Description + " " + event.RawData + " " + event.EventType
if contains(desc, "SXM2") || contains(desc, "1653925025827") {
t.Logf(" SXM2 Event: [%s] %s (Severity: %s)", event.EventType, event.Description, event.Severity)
sxm2Events++
}
}
if sxm2Events == 0 {
t.Error("Expected to find events for SXM2 (faulty GPU 1653925025827)")
}
t.Logf("\nSXM2 failure events: %d", sxm2Events)
}
func contains(s, substr string) bool {
return len(s) >= len(substr) && (s == substr || len(s) > len(substr) &&
(s[:len(substr)] == substr || s[len(s)-len(substr):] == substr ||
findSubstring(s, substr)))
}
func findSubstring(s, substr string) bool {
for i := 0; i <= len(s)-len(substr); i++ {
if s[i:i+len(substr)] == substr {
return true
}
}
return false
}
func TestFormatGPUSerial(t *testing.T) {
tests := []struct {
input string
expected string
}{
{
input: "14-17-dc-65-77-2d-b0-48",
expected: "48:B0:2D:77:65:DC:17:14",
},
{
input: "f2-fd-85-e0-2f-2d-b0-48",
expected: "48:B0:2D:2F:E0:85:FD:F2",
},
}
for _, tt := range tests {
t.Run(tt.input, func(t *testing.T) {
result := formatGPUSerial(tt.input)
if result != tt.expected {
t.Errorf("formatGPUSerial(%s) = %s, want %s", tt.input, result, tt.expected)
}
})
}
}