reanimator/bee
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity

Add NVIDIA stress loader selection and DCGM 4 support

Browse Source
This commit is contained in:
Mikhail Chusavitin
2026-03-31 11:15:15 +03:00
parent 20f834aa96
commit 6dee8f3509
31 changed files with 789 additions and 111 deletions
Download Patch File Download Diff File Expand all files Collapse all files

65
audit/internal/webui/tasks.go
View File

@@ -24,22 +24,31 @@ const (
TaskCancelled = "cancelled"
)
// taskNames maps target → human-readable name.
// taskNames maps target → human-readable name for validate (SAT) runs.
var taskNames = map[string]string{
"nvidia": "NVIDIA SAT",
"memory": "Memory SAT",
"storage": "Storage SAT",
"cpu": "CPU SAT",
"amd": "AMD GPU SAT",
"amd-mem": "AMD GPU MEM Integrity",
"amd-bandwidth": "AMD GPU MEM Bandwidth",
"amd-stress": "AMD GPU Burn-in",
"memory-stress": "Memory Burn-in",
"sat-stress": "SAT Stress (stressapptest)",
"nvidia": "NVIDIA SAT",
"nvidia-stress": "NVIDIA GPU Stress",
"memory": "Memory SAT",
"storage": "Storage SAT",
"cpu": "CPU SAT",
"amd": "AMD GPU SAT",
"amd-mem": "AMD GPU MEM Integrity",
"amd-bandwidth": "AMD GPU MEM Bandwidth",
"amd-stress": "AMD GPU Burn-in",
"memory-stress": "Memory Burn-in",
"sat-stress": "SAT Stress (stressapptest)",
"platform-stress": "Platform Thermal Cycling",
"audit": "Audit",
"install": "Install to Disk",
"install-to-ram": "Install to RAM",
"audit": "Audit",
"install": "Install to Disk",
"install-to-ram": "Install to RAM",
}
// burnNames maps target → human-readable name when a burn profile is set.
var burnNames = map[string]string{
"nvidia": "NVIDIA Burn-in",
"memory": "Memory Burn-in",
"cpu": "CPU Burn-in",
"amd": "AMD GPU Burn-in",
}
// Task represents one unit of work in the queue.
@@ -62,12 +71,14 @@ type Task struct {
// taskParams holds optional parameters parsed from the run request.
type taskParams struct {
Duration int `json:"duration,omitempty"`
DiagLevel int `json:"diag_level,omitempty"`
GPUIndices []int `json:"gpu_indices,omitempty"`
BurnProfile string `json:"burn_profile,omitempty"`
DisplayName string `json:"display_name,omitempty"`
Device string `json:"device,omitempty"` // for install
Duration int `json:"duration,omitempty"`
DiagLevel int `json:"diag_level,omitempty"`
GPUIndices []int `json:"gpu_indices,omitempty"`
ExcludeGPUIndices []int `json:"exclude_gpu_indices,omitempty"`
Loader string `json:"loader,omitempty"`
BurnProfile string `json:"burn_profile,omitempty"`
DisplayName string `json:"display_name,omitempty"`
Device string `json:"device,omitempty"` // for install
}
type persistedTask struct {
@@ -162,6 +173,9 @@ var (
runAMDMemBandwidthPackCtx = func(a *app.App, ctx context.Context, baseDir string, logFunc func(string)) (string, error) {
return a.RunAMDMemBandwidthPackCtx(ctx, baseDir, logFunc)
}
runNvidiaStressPackCtx = func(a *app.App, ctx context.Context, baseDir string, opts platform.NvidiaStressOptions, logFunc func(string)) (string, error) {
return a.RunNvidiaStressPackCtx(ctx, baseDir, opts, logFunc)
}
runAMDStressPackCtx = func(a *app.App, ctx context.Context, baseDir string, durationSec int, logFunc func(string)) (string, error) {
return a.RunAMDStressPackCtx(ctx, baseDir, durationSec, logFunc)
}
@@ -403,6 +417,17 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
} else {
archive, err = a.RunNvidiaAcceptancePack("", j.append)
}
case "nvidia-stress":
dur := t.params.Duration
if t.params.BurnProfile != "" && dur <= 0 {
dur = resolveBurnPreset(t.params.BurnProfile).DurationSec
}
archive, err = runNvidiaStressPackCtx(a, ctx, "", platform.NvidiaStressOptions{
DurationSec: dur,
Loader: t.params.Loader,
GPUIndices: t.params.GPUIndices,
ExcludeGPUIndices: t.params.ExcludeGPUIndices,
}, j.append)
case "memory":
archive, err = runMemoryAcceptancePackCtx(a, ctx, "", j.append)
case "storage":