Files
bee/audit/internal/collector/contract.go
T
Mikhail ChusavitinandClaude Sonnet 5 a34e823f82 fix(collector): stop NVIDIA enrichment from clobbering PCIe status
enrichPCIeWithNVIDIAData unconditionally overwrote dev.Status after
collectPCIe() had already flagged a Warning/Critical (e.g. PCIe link
speed degraded), so a clean ECC/remap/reset readout silently downgraded
that finding back to OK while leaving the stale ErrorDescription behind.

Add a severity-ordered merge (OK/Unknown < Warning < Critical) shared
via mergeDeviceStatus in contract.go, and route both the NVIDIA status
calculation and the driver-unavailable fallback through it.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-08-04 17:28:47 +03:00

108 lines
3.1 KiB
Go

package collector
import (
"bee/audit/internal/schema"
"strings"
)
const (
statusOK = "OK"
statusWarning = "Warning"
statusCritical = "Critical"
statusUnknown = "Unknown"
statusEmpty = "Empty"
)
func mapPCIeDeviceClass(raw string) string {
normalized := strings.ToLower(strings.TrimSpace(raw))
switch {
case normalized == "":
return ""
case strings.Contains(normalized, "ethernet controller"):
return "EthernetController"
case strings.Contains(normalized, "fibre channel"):
return "FibreChannelController"
case strings.Contains(normalized, "network controller"), strings.Contains(normalized, "infiniband controller"):
return "NetworkController"
case strings.Contains(normalized, "serial attached scsi"), strings.Contains(normalized, "storage controller"):
return "StorageController"
case strings.Contains(normalized, "raid"), strings.Contains(normalized, "mass storage"):
return "MassStorageController"
case strings.Contains(normalized, "display controller"):
return "DisplayController"
case strings.Contains(normalized, "vga"), strings.Contains(normalized, "3d controller"), strings.Contains(normalized, "video controller"):
return "VideoController"
case strings.Contains(normalized, "processing accelerators"), strings.Contains(normalized, "processing accelerator"):
return "ProcessingAccelerator"
default:
return raw
}
}
func isNICClass(class string) bool {
switch strings.TrimSpace(class) {
case "EthernetController", "NetworkController":
return true
default:
return false
}
}
func isGPUClass(class string) bool {
switch strings.TrimSpace(class) {
case "VideoController", "DisplayController", "ProcessingAccelerator":
return true
default:
return false
}
}
func isRAIDClass(class string) bool {
switch strings.TrimSpace(class) {
case "MassStorageController", "StorageController":
return true
default:
return false
}
}
// statusSeverity ranks component statuses so merges can only escalate, never
// downgrade. Unknown ranks with OK: it means "couldn't tell", not "healthy",
// so it must not silently clear a Warning/Critical raised by an earlier stage.
func statusSeverity(status string) int {
switch strings.TrimSpace(status) {
case statusCritical:
return 3
case statusWarning:
return 2
case statusOK:
return 1
case statusUnknown:
return 1
default:
return 0
}
}
// mergeDeviceStatus applies status/description to dev only if it is at least
// as severe as whatever is already set. This lets later enrichment stages
// (e.g. NVIDIA telemetry) report their own findings without silently
// clobbering a Warning/Critical raised earlier in the collector pipeline
// (e.g. a PCIe link-speed degradation).
func mergeDeviceStatus(dev *schema.HardwarePCIeDevice, status, description string) {
if dev == nil || status == "" {
return
}
current := ""
if dev.Status != nil {
current = strings.TrimSpace(*dev.Status)
}
if current != "" && current != statusUnknown && statusSeverity(status) <= statusSeverity(current) {
return
}
dev.Status = &status
if strings.TrimSpace(description) != "" {
dev.ErrorDescription = &description
}
}