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

Refactor validate modes, fix benchmark report and IPMI power

Browse Source 
- Replace diag level 1-4 dropdown with Validate/Stress radio buttons
- Validate: dcgmi L2, 60s CPU, 256MB/1p memtester, SMART short
- Stress: dcgmi L3 + targeted_stress in Run All, 30min CPU, 1GB/3p memtester, SMART long/NVMe extended
- Parallel GPU mode: spawn single task for all GPUs instead of splitting per model
- Benchmark table: per-GPU columns for sequential runs, server-wide column for parallel
- Benchmark report converted to Markdown with server model, GPU model, version in header; only steady-state charts
- Fix IPMI power parsing in benchmark (was looking for 'Current Power', correct field is 'Instantaneous power reading')

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Michael Chus
2026-04-08 00:42:12 +03:00
parent dd26e03b2d
commit b2f8626fee
12 changed files with 332 additions and 164 deletions
Download Patch File Download Diff File Expand all files Collapse all files

67
audit/internal/webui/api.go
View File

@@ -222,7 +222,21 @@ func formatSplitTaskName(baseName, selectionLabel string) string {
}
func buildNvidiaTaskSet(target string, priority int, createdAt time.Time, params taskParams, baseName string, appRef *app.App, idPrefix string) ([]*Task, error) {
if !shouldSplitHomogeneousNvidiaTarget(target) {
if !shouldSplitHomogeneousNvidiaTarget(target) || params.ParallelGPUs {
// Parallel mode (or non-splittable target): one task for all selected GPUs.
if params.ParallelGPUs && shouldSplitHomogeneousNvidiaTarget(target) {
// Resolve the selected GPU indices so ExcludeGPUIndices is applied.
gpus, err := apiListNvidiaGPUs(appRef)
if err != nil {
return nil, err
}
resolved, err := expandSelectedGPUIndices(gpus, params.GPUIndices, params.ExcludeGPUIndices)
if err != nil {
return nil, err
}
params.GPUIndices = resolved
params.ExcludeGPUIndices = nil
}
t := &Task{
ID: newJobID(idPrefix),
Name: baseName,
@@ -262,6 +276,53 @@ func buildNvidiaTaskSet(target string, priority int, createdAt time.Time, params
return tasks, nil
}
// expandSelectedGPUIndices returns the sorted list of selected GPU indices after
// applying include/exclude filters, without splitting by model.
func expandSelectedGPUIndices(gpus []platform.NvidiaGPU, include, exclude []int) ([]int, error) {
indexed := make(map[int]struct{}, len(gpus))
allIndices := make([]int, 0, len(gpus))
for _, gpu := range gpus {
indexed[gpu.Index] = struct{}{}
allIndices = append(allIndices, gpu.Index)
}
sort.Ints(allIndices)
selected := allIndices
if len(include) > 0 {
selected = make([]int, 0, len(include))
seen := make(map[int]struct{}, len(include))
for _, idx := range include {
if _, ok := indexed[idx]; !ok {
continue
}
if _, dup := seen[idx]; dup {
continue
}
seen[idx] = struct{}{}
selected = append(selected, idx)
}
sort.Ints(selected)
}
if len(exclude) > 0 {
skip := make(map[int]struct{}, len(exclude))
for _, idx := range exclude {
skip[idx] = struct{}{}
}
filtered := selected[:0]
for _, idx := range selected {
if _, ok := skip[idx]; ok {
continue
}
filtered = append(filtered, idx)
}
selected = filtered
}
if len(selected) == 0 {
return nil, fmt.Errorf("no NVIDIA GPUs selected")
}
return selected, nil
}
// ── SSE helpers ───────────────────────────────────────────────────────────────
func sseWrite(w http.ResponseWriter, event, data string) bool {
@@ -423,7 +484,7 @@ func (h *handler) handleAPISATRun(target string) http.HandlerFunc {
var body struct {
Duration int `json:"duration"`
DiagLevel int `json:"diag_level"`
StressMode bool `json:"stress_mode"`
GPUIndices []int `json:"gpu_indices"`
ExcludeGPUIndices []int `json:"exclude_gpu_indices"`
Loader string `json:"loader"`
@@ -444,7 +505,7 @@ func (h *handler) handleAPISATRun(target string) http.HandlerFunc {
}
params := taskParams{
Duration: body.Duration,
DiagLevel: body.DiagLevel,
StressMode: body.StressMode,
GPUIndices: body.GPUIndices,
ExcludeGPUIndices: body.ExcludeGPUIndices,
Loader: body.Loader,

4
audit/internal/webui/server_test.go
View File

@@ -693,8 +693,8 @@ func TestBenchmarkPageRendersSavedResultsTable(t *testing.T) {
for _, needle := range []string{
`Benchmark Results`,
`Composite score by saved benchmark run and GPU.`,
`NVIDIA H100 PCIe / GPU 0`,
`NVIDIA H100 PCIe / GPU 1`,
`GPU #0 — NVIDIA H100 PCIe`,
`GPU #1 — NVIDIA H100 PCIe`,
`#1`,
wantTime,
`1176.25`,

30
audit/internal/webui/tasks.go
View File

@@ -115,10 +115,11 @@ 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"`
StressMode bool `json:"stress_mode,omitempty"`
GPUIndices []int `json:"gpu_indices,omitempty"`
ExcludeGPUIndices []int `json:"exclude_gpu_indices,omitempty"`
SizeMB int `json:"size_mb,omitempty"`
Passes int `json:"passes,omitempty"`
Loader string `json:"loader,omitempty"`
BurnProfile string `json:"burn_profile,omitempty"`
BenchmarkProfile string `json:"benchmark_profile,omitempty"`
@@ -215,11 +216,11 @@ var globalQueue = &taskQueue{trigger: make(chan struct{}, 1)}
const maxTaskHistory = 50
var (
runMemoryAcceptancePackCtx = func(a *app.App, ctx context.Context, baseDir string, logFunc func(string)) (string, error) {
return a.RunMemoryAcceptancePackCtx(ctx, baseDir, logFunc)
runMemoryAcceptancePackCtx = func(a *app.App, ctx context.Context, baseDir string, sizeMB, passes int, logFunc func(string)) (string, error) {
return a.RunMemoryAcceptancePackCtx(ctx, baseDir, sizeMB, passes, logFunc)
}
runStorageAcceptancePackCtx = func(a *app.App, ctx context.Context, baseDir string, logFunc func(string)) (string, error) {
return a.RunStorageAcceptancePackCtx(ctx, baseDir, logFunc)
runStorageAcceptancePackCtx = func(a *app.App, ctx context.Context, baseDir string, extended bool, logFunc func(string)) (string, error) {
return a.RunStorageAcceptancePackCtx(ctx, baseDir, extended, logFunc)
}
runCPUAcceptancePackCtx = func(a *app.App, ctx context.Context, baseDir string, durationSec int, logFunc func(string)) (string, error) {
return a.RunCPUAcceptancePackCtx(ctx, baseDir, durationSec, logFunc)
@@ -552,7 +553,10 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
err = fmt.Errorf("app not configured")
break
}
diagLevel := t.params.DiagLevel
diagLevel := 2
if t.params.StressMode {
diagLevel = 3
}
if len(t.params.GPUIndices) > 0 || diagLevel > 0 {
result, e := a.RunNvidiaAcceptancePackWithOptions(
ctx, "", diagLevel, t.params.GPUIndices, j.append,
@@ -658,13 +662,17 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
err = fmt.Errorf("app not configured")
break
}
archive, err = runMemoryAcceptancePackCtx(a, ctx, "", j.append)
sizeMB, passes := 256, 1
if t.params.StressMode {
sizeMB, passes = 1024, 3
}
archive, err = runMemoryAcceptancePackCtx(a, ctx, "", sizeMB, passes, j.append)
case "storage":
if a == nil {
err = fmt.Errorf("app not configured")
break
}
archive, err = runStorageAcceptancePackCtx(a, ctx, "", j.append)
archive, err = runStorageAcceptancePackCtx(a, ctx, "", t.params.StressMode, j.append)
case "cpu":
if a == nil {
err = fmt.Errorf("app not configured")
@@ -675,7 +683,11 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
dur = resolveBurnPreset(t.params.BurnProfile).DurationSec
}
if dur <= 0 {
dur = 60
if t.params.StressMode {
dur = 1800
} else {
dur = 60
}
}
j.append(fmt.Sprintf("CPU stress duration: %ds", dur))
archive, err = runCPUAcceptancePackCtx(a, ctx, "", dur, j.append)

2
audit/internal/webui/tasks_test.go
View File

@@ -422,7 +422,7 @@ func TestWriteTaskReportArtifactsIncludesBenchmarkResultsForTask(t *testing.T) {
for _, needle := range []string{
`Benchmark Results`,
`Composite score for this benchmark task.`,
`NVIDIA H100 PCIe / GPU 0`,
`GPU #0 — NVIDIA H100 PCIe`,
`1176.25`,
} {
if !strings.Contains(html, needle) {