feat: task queue, UI overhaul, burn tests, install-to-RAM

- Task queue: all SAT/audit jobs enqueue and run one-at-a-time;
  tasks persist past page navigation; new Tasks page with cancel/priority/log stream
- UI: consolidate nav (Validate, Burn, Tasks, Tools); Audit becomes modal;
  Dashboard hardware summary badges + split metrics charts (load/temp/power);
  Tools page consolidates network, services, install, support bundle
- AMD GPU: acceptance test and stress burn cards; GPU presence API greys
  out irrelevant SAT cards automatically
- Burn tests: Memory Stress (stress-ng --vm), SAT Stress (stressapptest)
- Install to RAM: copies squashfs to /dev/shm, re-associates loop devices
  via LOOP_CHANGE_FD ioctl so live media can be ejected
- Charts: relative time axis (0 = now, negative left)
- memtester: LimitMEMLOCK=infinity in bee-web.service; empty output → UNSUPPORTED
- SAT overlay applied dynamically on every /audit.json serve
- MIME panic guard for LiveCD ramdisk I/O errors
- ISO: add memtest86+, stressapptest packages; memtest86+ GRUB entry;
  disable screensaver/DPMS in bee-openbox-session
- Unknown SAT status severity = 1 (does not override OK)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
2026-03-28 21:15:11 +03:00
parent 911745e4da
commit 0a98ed8ae9
22 changed files with 1964 additions and 326 deletions

View File

@@ -9,7 +9,6 @@ import (
"net/http"
"os/exec"
"path/filepath"
"strings"
"sync/atomic"
"time"
@@ -110,39 +109,37 @@ func runCmdJob(j *jobState, cmd *exec.Cmd) {
// ── Audit ─────────────────────────────────────────────────────────────────────
func (h *handler) handleAPIAuditRun(w http.ResponseWriter, r *http.Request) {
func (h *handler) handleAPIAuditRun(w http.ResponseWriter, _ *http.Request) {
if h.opts.App == nil {
writeError(w, http.StatusServiceUnavailable, "app not configured")
return
}
id := newJobID("audit")
j := globalJobs.create(id)
go func() {
j.append("Running audit...")
result, err := h.opts.App.RunAuditNow(h.opts.RuntimeMode)
if err != nil {
j.append("ERROR: " + err.Error())
j.finish(err.Error())
return
}
for _, line := range strings.Split(result.Body, "\n") {
if line != "" {
j.append(line)
}
}
j.finish("")
}()
writeJSON(w, map[string]string{"job_id": id})
t := &Task{
ID: newJobID("audit"),
Name: "Audit",
Target: "audit",
Status: TaskPending,
CreatedAt: time.Now(),
}
globalQueue.enqueue(t)
writeJSON(w, map[string]string{"task_id": t.ID, "job_id": t.ID})
}
func (h *handler) handleAPIAuditStream(w http.ResponseWriter, r *http.Request) {
id := r.URL.Query().Get("job_id")
j, ok := globalJobs.get(id)
if !ok {
http.Error(w, "job not found", http.StatusNotFound)
if id == "" {
id = r.URL.Query().Get("task_id")
}
// Try task queue first, then legacy job manager
if j, ok := globalQueue.findJob(id); ok {
streamJob(w, r, j)
return
}
streamJob(w, r, j)
if j, ok := globalJobs.get(id); ok {
streamJob(w, r, j)
return
}
http.Error(w, "job not found", http.StatusNotFound)
}
// ── SAT ───────────────────────────────────────────────────────────────────────
@@ -153,96 +150,87 @@ func (h *handler) handleAPISATRun(target string) http.HandlerFunc {
writeError(w, http.StatusServiceUnavailable, "app not configured")
return
}
id := newJobID("sat-" + target)
j := globalJobs.create(id)
ctx, cancel := context.WithCancel(context.Background())
j.cancel = cancel
go func() {
defer cancel()
j.append(fmt.Sprintf("Starting %s acceptance test...", target))
var (
archive string
err error
)
var body struct {
Duration int `json:"duration"`
DiagLevel int `json:"diag_level"`
GPUIndices []int `json:"gpu_indices"`
}
body.DiagLevel = 1
if r.ContentLength > 0 {
_ = json.NewDecoder(r.Body).Decode(&body)
}
// Parse optional parameters
var body struct {
Duration int `json:"duration"`
DiagLevel int `json:"diag_level"`
GPUIndices []int `json:"gpu_indices"`
}
body.DiagLevel = 1
if r.ContentLength > 0 {
_ = json.NewDecoder(r.Body).Decode(&body)
}
switch target {
case "nvidia":
if len(body.GPUIndices) > 0 || body.DiagLevel > 0 {
result, e := h.opts.App.RunNvidiaAcceptancePackWithOptions(
ctx, "", body.DiagLevel, body.GPUIndices,
)
if e != nil {
err = e
} else {
archive = result.Body
}
} else {
archive, err = h.opts.App.RunNvidiaAcceptancePack("")
}
case "memory":
archive, err = h.opts.App.RunMemoryAcceptancePack("")
case "storage":
archive, err = h.opts.App.RunStorageAcceptancePack("")
case "cpu":
dur := body.Duration
if dur <= 0 {
dur = 60
}
archive, err = h.opts.App.RunCPUAcceptancePack("", dur)
}
if err != nil {
if ctx.Err() != nil {
j.append("Aborted.")
j.finish("aborted")
} else {
j.append("ERROR: " + err.Error())
j.finish(err.Error())
}
return
}
j.append(fmt.Sprintf("Archive written: %s", archive))
j.finish("")
}()
writeJSON(w, map[string]string{"job_id": id})
name := taskNames[target]
if name == "" {
name = target
}
t := &Task{
ID: newJobID("sat-" + target),
Name: name,
Target: target,
Status: TaskPending,
CreatedAt: time.Now(),
params: taskParams{
Duration: body.Duration,
DiagLevel: body.DiagLevel,
GPUIndices: body.GPUIndices,
},
}
globalQueue.enqueue(t)
writeJSON(w, map[string]string{"task_id": t.ID, "job_id": t.ID})
}
}
func (h *handler) handleAPISATStream(w http.ResponseWriter, r *http.Request) {
id := r.URL.Query().Get("job_id")
j, ok := globalJobs.get(id)
if !ok {
http.Error(w, "job not found", http.StatusNotFound)
if id == "" {
id = r.URL.Query().Get("task_id")
}
if j, ok := globalQueue.findJob(id); ok {
streamJob(w, r, j)
return
}
streamJob(w, r, j)
if j, ok := globalJobs.get(id); ok {
streamJob(w, r, j)
return
}
http.Error(w, "job not found", http.StatusNotFound)
}
func (h *handler) handleAPISATAbort(w http.ResponseWriter, r *http.Request) {
id := r.URL.Query().Get("job_id")
j, ok := globalJobs.get(id)
if !ok {
http.Error(w, "job not found", http.StatusNotFound)
if id == "" {
id = r.URL.Query().Get("task_id")
}
if t, ok := globalQueue.findByID(id); ok {
globalQueue.mu.Lock()
switch t.Status {
case TaskPending:
t.Status = TaskCancelled
now := time.Now()
t.DoneAt = &now
case TaskRunning:
if t.job != nil {
t.job.abort()
}
t.Status = TaskCancelled
now := time.Now()
t.DoneAt = &now
}
globalQueue.mu.Unlock()
writeJSON(w, map[string]string{"status": "aborted"})
return
}
if j.abort() {
writeJSON(w, map[string]string{"status": "aborted"})
} else {
writeJSON(w, map[string]string{"status": "not_running"})
if j, ok := globalJobs.get(id); ok {
if j.abort() {
writeJSON(w, map[string]string{"status": "aborted"})
} else {
writeJSON(w, map[string]string{"status": "not_running"})
}
return
}
http.Error(w, "job not found", http.StatusNotFound)
}
// ── Services ──────────────────────────────────────────────────────────────────
@@ -401,6 +389,51 @@ func (h *handler) handleAPIExportBundle(w http.ResponseWriter, r *http.Request)
})
}
// ── GPU presence ──────────────────────────────────────────────────────────────
func (h *handler) handleAPIGPUPresence(w http.ResponseWriter, r *http.Request) {
if h.opts.App == nil {
writeError(w, http.StatusServiceUnavailable, "app not configured")
return
}
gp := h.opts.App.DetectGPUPresence()
w.Header().Set("Content-Type", "application/json")
_ = json.NewEncoder(w).Encode(map[string]bool{
"nvidia": gp.Nvidia,
"amd": gp.AMD,
})
}
// ── System ────────────────────────────────────────────────────────────────────
func (h *handler) handleAPIRAMStatus(w http.ResponseWriter, r *http.Request) {
if h.opts.App == nil {
writeError(w, http.StatusServiceUnavailable, "app not configured")
return
}
inRAM := h.opts.App.IsLiveMediaInRAM()
w.Header().Set("Content-Type", "application/json")
_ = json.NewEncoder(w).Encode(map[string]bool{"in_ram": inRAM})
}
func (h *handler) handleAPIInstallToRAM(w http.ResponseWriter, r *http.Request) {
if h.opts.App == nil {
writeError(w, http.StatusServiceUnavailable, "app not configured")
return
}
t := &Task{
ID: newJobID("install-to-ram"),
Name: "Install to RAM",
Target: "install-to-ram",
Priority: 10,
Status: TaskPending,
CreatedAt: time.Now(),
}
globalQueue.enqueue(t)
w.Header().Set("Content-Type", "application/json")
_ = json.NewEncoder(w).Encode(map[string]string{"task_id": t.ID})
}
// ── Tools ─────────────────────────────────────────────────────────────────────
var standardTools = []string{
@@ -507,7 +540,7 @@ func (h *handler) handleAPIInstallRun(w http.ResponseWriter, r *http.Request) {
h.installMu.Unlock()
logFile := platform.InstallLogPath(req.Device)
go runCmdJob(j, exec.CommandContext(r.Context(), "bee-install", req.Device, logFile))
go runCmdJob(j, exec.CommandContext(context.Background(), "bee-install", req.Device, logFile))
w.WriteHeader(http.StatusNoContent)
}
@@ -589,3 +622,95 @@ func (h *handler) handleAPIMetricsStream(w http.ResponseWriter, r *http.Request)
}
}
}
// ── Network toggle ────────────────────────────────────────────────────────────
const netRollbackTimeout = 60 * time.Second
func (h *handler) handleAPINetworkToggle(w http.ResponseWriter, r *http.Request) {
if h.opts.App == nil {
writeError(w, http.StatusServiceUnavailable, "app not configured")
return
}
var req struct {
Iface string `json:"iface"`
}
if err := json.NewDecoder(r.Body).Decode(&req); err != nil || req.Iface == "" {
writeError(w, http.StatusBadRequest, "iface is required")
return
}
wasUp, err := h.opts.App.GetInterfaceState(req.Iface)
if err != nil {
writeError(w, http.StatusInternalServerError, err.Error())
return
}
if err := h.opts.App.SetInterfaceState(req.Iface, !wasUp); err != nil {
writeError(w, http.StatusInternalServerError, err.Error())
return
}
// Cancel any existing pending change (rollback it first).
h.pendingNetMu.Lock()
if h.pendingNet != nil {
prev := h.pendingNet
prev.mu.Lock()
prev.timer.Stop()
_ = h.opts.App.SetInterfaceState(prev.iface, prev.wasUp)
prev.mu.Unlock()
}
pnc := &pendingNetChange{iface: req.Iface, wasUp: wasUp}
pnc.timer = time.AfterFunc(netRollbackTimeout, func() {
_ = h.opts.App.SetInterfaceState(req.Iface, wasUp)
h.pendingNetMu.Lock()
if h.pendingNet == pnc {
h.pendingNet = nil
}
h.pendingNetMu.Unlock()
})
h.pendingNet = pnc
h.pendingNetMu.Unlock()
newState := "up"
if wasUp {
newState = "down"
}
writeJSON(w, map[string]any{
"iface": req.Iface,
"new_state": newState,
"rollback_in": int(netRollbackTimeout.Seconds()),
})
}
func (h *handler) handleAPINetworkConfirm(w http.ResponseWriter, _ *http.Request) {
h.pendingNetMu.Lock()
pnc := h.pendingNet
h.pendingNet = nil
h.pendingNetMu.Unlock()
if pnc != nil {
pnc.mu.Lock()
pnc.timer.Stop()
pnc.mu.Unlock()
}
writeJSON(w, map[string]string{"status": "confirmed"})
}
func (h *handler) handleAPINetworkRollback(w http.ResponseWriter, _ *http.Request) {
h.pendingNetMu.Lock()
pnc := h.pendingNet
h.pendingNet = nil
h.pendingNetMu.Unlock()
if pnc == nil {
writeError(w, http.StatusConflict, "no pending network change")
return
}
pnc.mu.Lock()
pnc.timer.Stop()
pnc.mu.Unlock()
if h.opts.App != nil {
_ = h.opts.App.SetInterfaceState(pnc.iface, pnc.wasUp)
}
writeJSON(w, map[string]string{"status": "rolled back"})
}

View File

@@ -84,17 +84,13 @@ tbody tr:hover td{background:rgba(0,0,0,.03)}
}
func layoutNav(active string) string {
items := []struct{ id, label, href string }{
{"dashboard", "Dashboard", "/"},
{"viewer", "Audit Snapshot", "/viewer"},
{"metrics", "Metrics", "/metrics"},
{"tests", "Acceptance Tests", "/tests"},
{"burn-in", "Burn-in", "/burn-in"},
{"network", "Network", "/network"},
{"services", "Services", "/services"},
{"export", "Export", "/export"},
{"tools", "Tools", "/tools"},
{"install", "Install to Disk", "/install"},
items := []struct{ id, label, href, onclick string }{
{"dashboard", "Dashboard", "/", ""},
{"audit", "Audit", "#", "openAuditModal();return false;"},
{"validate", "Validate", "/validate", ""},
{"burn", "Burn", "/burn", ""},
{"tasks", "Tasks", "/tasks", ""},
{"tools", "Tools", "/tools", ""},
}
var b strings.Builder
b.WriteString(`<aside class="sidebar">`)
@@ -105,8 +101,13 @@ func layoutNav(active string) string {
if item.id == active {
cls += " active"
}
b.WriteString(fmt.Sprintf(`<a class="%s" href="%s">%s</a>`,
cls, item.href, item.label))
if item.onclick != "" {
b.WriteString(fmt.Sprintf(`<a class="%s" href="%s" onclick="%s">%s</a>`,
cls, item.href, item.onclick, item.label))
} else {
b.WriteString(fmt.Sprintf(`<a class="%s" href="%s">%s</a>`,
cls, item.href, item.label))
}
}
b.WriteString(`</nav></aside>`)
return b.String()
@@ -120,18 +121,35 @@ func renderPage(page string, opts HandlerOptions) string {
pageID = "dashboard"
title = "Dashboard"
body = renderDashboard(opts)
case "validate":
pageID = "validate"
title = "Validate"
body = renderValidate()
case "burn":
pageID = "burn"
title = "Burn"
body = renderBurn()
case "tasks":
pageID = "tasks"
title = "Tasks"
body = renderTasks()
case "tools":
pageID = "tools"
title = "Tools"
body = renderTools()
// Legacy routes kept accessible but not in nav
case "metrics":
pageID = "metrics"
title = "Live Metrics"
body = renderMetrics()
case "tests":
pageID = "tests"
pageID = "validate"
title = "Acceptance Tests"
body = renderTests()
body = renderValidate()
case "burn-in":
pageID = "burn-in"
pageID = "burn"
title = "Burn-in Tests"
body = renderBurnIn()
body = renderBurn()
case "network":
pageID = "network"
title = "Network"
@@ -144,10 +162,6 @@ func renderPage(page string, opts HandlerOptions) string {
pageID = "export"
title = "Export"
body = renderExport(opts.ExportDir)
case "tools":
pageID = "tools"
title = "Tools"
body = renderTools()
case "install":
pageID = "install"
title = "Install to Disk"
@@ -162,48 +176,158 @@ func renderPage(page string, opts HandlerOptions) string {
layoutNav(pageID) +
`<div class="main"><div class="topbar"><h1>` + html.EscapeString(title) + `</h1></div><div class="content">` +
body +
`</div></div></body></html>`
`</div></div>` +
renderAuditModal() +
`</body></html>`
}
// ── Dashboard ─────────────────────────────────────────────────────────────────
func renderDashboard(opts HandlerOptions) string {
var b strings.Builder
b.WriteString(`<div class="grid2">`)
// Left: health summary
b.WriteString(`<div>`)
b.WriteString(renderHardwareSummaryCard(opts))
b.WriteString(renderHealthCard(opts))
b.WriteString(`</div>`)
// Right: quick actions
b.WriteString(`<div>`)
b.WriteString(`<div class="card"><div class="card-head">Quick Actions</div><div class="card-body">`)
b.WriteString(`<a class="btn btn-primary" href="/export/support.tar.gz" style="display:block;margin-bottom:10px">⬇ Download Support Bundle</a>`)
b.WriteString(`<a class="btn btn-secondary" href="/audit.json" style="display:block;margin-bottom:10px" target="_blank">📄 Open audit.json</a>`)
b.WriteString(`<a class="btn btn-secondary" href="/export/" style="display:block">📁 Browse Export Files</a>`)
b.WriteString(`<div style="margin-top:14px"><button class="btn btn-secondary" onclick="runAudit()">▶ Re-run Audit</button></div>`)
b.WriteString(`</div></div>`)
b.WriteString(`</div>`)
b.WriteString(`</div>`)
// Audit run output div
b.WriteString(`<div id="audit-output" style="display:none" class="card"><div class="card-head">Audit Output</div><div class="card-body"><div id="audit-terminal" class="terminal"></div></div></div>`)
b.WriteString(`<script>
function runAudit() {
document.getElementById('audit-output').style.display='block';
const term = document.getElementById('audit-terminal');
term.textContent = 'Starting audit...\n';
fetch('/api/audit/run', {method:'POST'})
.then(r => r.json())
.then(d => {
const es = new EventSource('/api/audit/stream?job_id=' + d.job_id);
es.onmessage = e => { term.textContent += e.data + '\n'; term.scrollTop = term.scrollHeight; };
es.addEventListener('done', e => { es.close(); term.textContent += (e.data ? '\\nERROR: ' + e.data : '\\nDone.') + '\n'; location.reload(); });
});
}
</script>`)
b.WriteString(renderMetrics())
return b.String()
}
func renderHardwareSummaryCard(opts HandlerOptions) string {
data, err := loadSnapshot(opts.AuditPath)
if err != nil {
return `<div class="card"><div class="card-head">Hardware Summary</div><div class="card-body"><span class="badge badge-unknown">No audit data</span></div></div>`
}
// Parse just enough fields for the summary banner
var snap struct {
Summary struct {
CPU struct{ Model string }
Memory struct{ TotalGB float64 }
Storage []struct{ Device, Model, Size string }
GPUs []struct{ Model string }
PSUs []struct{ Model string }
}
Network struct {
Interfaces []struct {
Name string
IPv4 []string
State string
}
}
}
// Try to extract top-level fields loosely
var raw map[string]json.RawMessage
if err := json.Unmarshal(data, &raw); err != nil {
return `<div class="card"><div class="card-head">Hardware Summary</div><div class="card-body"><span class="badge badge-err">Parse error</span></div></div>`
}
_ = snap
// Also load runtime-health for badges
type componentHealth struct {
FailCount int `json:"fail_count"`
WarnCount int `json:"warn_count"`
}
type healthSummary struct {
CPU componentHealth `json:"cpu"`
Memory componentHealth `json:"memory"`
Storage componentHealth `json:"storage"`
GPU componentHealth `json:"gpu"`
PSU componentHealth `json:"psu"`
Network componentHealth `json:"network"`
}
var health struct {
HardwareHealth healthSummary `json:"hardware_health"`
}
if hdata, herr := loadSnapshot(filepath.Join(opts.ExportDir, "runtime-health.json")); herr == nil {
_ = json.Unmarshal(hdata, &health)
}
badge := func(h componentHealth) string {
if h.FailCount > 0 {
return `<span class="badge badge-err">FAIL</span>`
}
if h.WarnCount > 0 {
return `<span class="badge badge-warn">WARN</span>`
}
return `<span class="badge badge-ok">OK</span>`
}
// Extract readable strings from raw JSON
getString := func(key string) string {
v, ok := raw[key]
if !ok {
return ""
}
var s string
if err := json.Unmarshal(v, &s); err == nil {
return s
}
return ""
}
cpuModel := getString("cpu_model")
memStr := getString("memory_summary")
gpuSummary := getString("gpu_summary")
var b strings.Builder
b.WriteString(`<div class="card"><div class="card-head">Hardware Summary</div><div class="card-body">`)
b.WriteString(`<table style="width:auto">`)
writeRow := func(label, value, badgeHTML string) {
b.WriteString(fmt.Sprintf(`<tr><td style="padding:6px 14px 6px 0;font-weight:700;white-space:nowrap">%s</td><td style="padding:6px 0">%s</td><td style="padding:6px 0 6px 12px">%s</td></tr>`,
html.EscapeString(label), html.EscapeString(value), badgeHTML))
}
if cpuModel != "" {
writeRow("CPU", cpuModel, badge(health.HardwareHealth.CPU))
} else {
writeRow("CPU", "—", badge(health.HardwareHealth.CPU))
}
if memStr != "" {
writeRow("Memory", memStr, badge(health.HardwareHealth.Memory))
} else {
writeRow("Memory", "—", badge(health.HardwareHealth.Memory))
}
if gpuSummary != "" {
writeRow("GPU", gpuSummary, badge(health.HardwareHealth.GPU))
} else {
writeRow("GPU", "—", badge(health.HardwareHealth.GPU))
}
writeRow("Storage", "—", badge(health.HardwareHealth.Storage))
writeRow("PSU", "—", badge(health.HardwareHealth.PSU))
b.WriteString(`</table>`)
b.WriteString(`</div></div>`)
return b.String()
}
func renderAuditModal() string {
return `<div id="audit-modal-overlay" style="display:none;position:fixed;inset:0;background:rgba(0,0,0,.5);z-index:100;align-items:center;justify-content:center">
<div style="background:#fff;border-radius:6px;padding:24px;min-width:480px;max-width:700px;position:relative">
<div style="font-weight:700;font-size:16px;margin-bottom:16px">Audit</div>
<div style="margin-bottom:12px;display:flex;gap:8px">
<button class="btn btn-primary" onclick="auditModalRun()">&#9654; Re-run Audit</button>
<a class="btn btn-secondary" href="/audit.json" download>&#8595; Download</a>
<a class="btn btn-secondary" href="/viewer" target="_blank">Open Viewer</a>
</div>
<div id="audit-modal-terminal" class="terminal" style="display:none;max-height:300px"></div>
<button class="btn btn-secondary btn-sm" onclick="closeAuditModal()" style="position:absolute;top:12px;right:12px">&#10005;</button>
</div>
</div>
<script>
function openAuditModal() {
document.getElementById('audit-modal-overlay').style.display='flex';
}
function closeAuditModal() {
document.getElementById('audit-modal-overlay').style.display='none';
}
function auditModalRun() {
const term = document.getElementById('audit-modal-terminal');
term.style.display='block'; term.textContent='Starting...\n';
fetch('/api/audit/run',{method:'POST'}).then(r=>r.json()).then(d=>{
const es=new EventSource('/api/tasks/'+d.task_id+'/stream');
es.onmessage=e=>{term.textContent+=e.data+'\n';term.scrollTop=term.scrollHeight;};
es.addEventListener('done',e=>{es.close();term.textContent+=(e.data?'\nERROR: '+e.data:'\nDone.')+'\n';});
});
}
</script>`
}
func renderHealthCard(opts HandlerOptions) string {
data, err := loadSnapshot(filepath.Join(opts.ExportDir, "runtime-health.json"))
if err != nil {
@@ -239,12 +363,26 @@ func renderHealthCard(opts HandlerOptions) string {
// ── Metrics ───────────────────────────────────────────────────────────────────
func renderMetrics() string {
return `<p style="color:var(--muted);font-size:13px;margin-bottom:16px">Live metrics — updated every 2 seconds. Charts use go-analyze/charts (grafana theme).</p>
return `<p style="color:var(--muted);font-size:13px;margin-bottom:16px">Live metrics — updated every 2 seconds.</p>
<div class="card" style="margin-bottom:16px">
<div class="card-head">Server</div>
<div class="card-head">Server — Load</div>
<div class="card-body" style="padding:8px">
<img id="chart-server" src="/api/metrics/chart/server.svg" style="width:100%;display:block;border-radius:6px" alt="Server metrics">
<img id="chart-server-load" src="/api/metrics/chart/server-load.svg" style="width:100%;display:block;border-radius:6px" alt="CPU/Mem load">
</div>
</div>
<div class="card" style="margin-bottom:16px">
<div class="card-head">Server — Temperature</div>
<div class="card-body" style="padding:8px">
<img id="chart-server-temp" src="/api/metrics/chart/server-temp.svg" style="width:100%;display:block;border-radius:6px" alt="CPU temperature">
</div>
</div>
<div class="card" style="margin-bottom:16px">
<div class="card-head">Server — Power</div>
<div class="card-body" style="padding:8px">
<img id="chart-server-power" src="/api/metrics/chart/server-power.svg" style="width:100%;display:block;border-radius:6px" alt="System power">
<div id="sys-table" style="margin-top:8px;font-size:12px"></div>
</div>
</div>
@@ -256,12 +394,16 @@ let knownGPUs = [];
function refreshCharts() {
const t = '?t=' + Date.now();
const srv = document.getElementById('chart-server');
if (srv) srv.src = srv.src.split('?')[0] + t;
knownGPUs.forEach(idx => {
const el = document.getElementById('chart-gpu-' + idx);
['chart-server-load','chart-server-temp','chart-server-power'].forEach(id => {
const el = document.getElementById(id);
if (el) el.src = el.src.split('?')[0] + t;
});
knownGPUs.forEach(idx => {
['load','temp','power'].forEach(kind => {
const el = document.getElementById('chart-gpu-' + idx + '-' + kind);
if (el) el.src = el.src.split('?')[0] + t;
});
});
}
setInterval(refreshCharts, 2000);
@@ -276,10 +418,19 @@ es.addEventListener('metrics', e => {
const div = document.createElement('div');
div.className = 'card';
div.style.marginBottom = '16px';
div.innerHTML = '<div class="card-head">GPU ' + g.index + '</div>' +
div.innerHTML =
'<div class="card-head">GPU ' + g.index + ' — Load</div>' +
'<div class="card-body" style="padding:8px">' +
'<img id="chart-gpu-' + g.index + '" src="/api/metrics/chart/gpu/' + g.index + '.svg" style="width:100%;display:block;border-radius:6px" alt="GPU ' + g.index + '">' +
'<div id="gpu-table-' + g.index + '" style="margin-top:8px;font-size:12px"></div>' +
'<img id="chart-gpu-' + g.index + '-load" src="/api/metrics/chart/gpu/' + g.index + '-load.svg" style="width:100%;display:block;border-radius:6px" alt="GPU ' + g.index + ' load">' +
'</div>' +
'<div class="card-head">GPU ' + g.index + ' — Temperature</div>' +
'<div class="card-body" style="padding:8px">' +
'<img id="chart-gpu-' + g.index + '-temp" src="/api/metrics/chart/gpu/' + g.index + '-temp.svg" style="width:100%;display:block;border-radius:6px" alt="GPU ' + g.index + ' temp">' +
'</div>' +
'<div class="card-head">GPU ' + g.index + ' — Power</div>' +
'<div class="card-body" style="padding:8px">' +
'<img id="chart-gpu-' + g.index + '-power" src="/api/metrics/chart/gpu/' + g.index + '-power.svg" style="width:100%;display:block;border-radius:6px" alt="GPU ' + g.index + ' power">' +
'<div id="gpu-table-' + g.index + '" style="margin-top:8px;font-size:12px"></div>' +
'</div>';
document.getElementById('gpu-charts').appendChild(div);
});
@@ -309,15 +460,27 @@ es.onerror = () => {};
</script>`
}
// ── Acceptance Tests ──────────────────────────────────────────────────────────
// ── Validate (Acceptance Tests) ───────────────────────────────────────────────
func renderTests() string {
return `<p style="color:var(--muted);font-size:13px;margin-bottom:16px">Run hardware acceptance tests and view results.</p>
<div class="grid2">
func renderValidate() string {
return `<div class="alert alert-info" style="margin-bottom:16px"><strong>Non-destructive:</strong> Validate tests collect diagnostics only. They do not write to disks, do not run sustained load, and do not increment hardware wear counters.</div>
<p style="color:var(--muted);font-size:13px;margin-bottom:16px">Tasks continue in the background — view progress in <a href="/tasks">Tasks</a>.</p>
<div class="card" style="margin-bottom:16px">
<div class="card-head">Run All Tests</div>
<div class="card-body" style="display:flex;align-items:center;gap:12px;flex-wrap:wrap">
<div class="form-row" style="margin:0"><label style="margin-right:6px">Cycles</label><input type="number" id="sat-cycles" value="1" min="1" max="100" style="width:70px;display:inline-block"></div>
<button class="btn btn-primary" onclick="runAllSAT()">&#9654; Run All</button>
<span id="sat-all-status" style="font-size:12px;color:var(--muted)"></span>
</div>
</div>
<div class="grid3">
` + renderSATCard("nvidia", "NVIDIA GPU", `<div class="form-row"><label>Diag Level</label><select id="sat-nvidia-level"><option value="1">Level 1 — Quick</option><option value="2">Level 2 — Standard</option><option value="3">Level 3 — Extended</option><option value="4">Level 4 — Full</option></select></div>`) +
renderSATCard("memory", "Memory", "") +
renderSATCard("storage", "Storage", "") +
renderSATCard("cpu", "CPU", `<div class="form-row"><label>Duration (seconds)</label><input type="number" id="sat-cpu-dur" value="60" min="10"></div>`) +
renderSATCard("amd", "AMD GPU", "") +
`</div>
<div id="sat-output" style="display:none;margin-top:16px" class="card">
<div class="card-head">Test Output <span id="sat-title"></span></div>
@@ -326,82 +489,175 @@ func renderTests() string {
<script>
let satES = null;
function runSAT(target) {
if (satES) satES.close();
if (satES) { satES.close(); satES = null; }
const body = {};
if (target === 'nvidia') body.diag_level = parseInt(document.getElementById('sat-nvidia-level').value)||1;
if (target === 'cpu') body.duration = parseInt(document.getElementById('sat-cpu-dur').value)||60;
document.getElementById('sat-output').style.display='block';
document.getElementById('sat-title').textContent = '— ' + target;
const term = document.getElementById('sat-terminal');
term.textContent = 'Starting ' + target + ' test...\n';
fetch('/api/sat/'+target+'/run', {method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify(body)})
term.textContent = 'Enqueuing ' + target + ' test...\n';
return fetch('/api/sat/'+target+'/run', {method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify(body)})
.then(r => r.json())
.then(d => {
satES = new EventSource('/api/sat/stream?job_id='+d.job_id);
term.textContent += 'Task ' + d.task_id + ' queued. Streaming log...\n';
satES = new EventSource('/api/tasks/'+d.task_id+'/stream');
satES.onmessage = e => { term.textContent += e.data+'\n'; term.scrollTop=term.scrollHeight; };
satES.addEventListener('done', e => { satES.close(); term.textContent += (e.data ? '\nERROR: '+e.data : '\nCompleted.')+'\n'; });
satES.addEventListener('done', e => { satES.close(); satES=null; term.textContent += (e.data ? '\nERROR: '+e.data : '\nCompleted.')+'\n'; });
});
}
function runAllSAT() {
const cycles = Math.max(1, parseInt(document.getElementById('sat-cycles').value)||1);
const targets = ['nvidia','memory','storage','cpu','amd'];
const total = targets.length * cycles;
let enqueued = 0;
const status = document.getElementById('sat-all-status');
status.textContent = 'Enqueuing...';
const enqueueNext = (cycle, idx) => {
if (cycle >= cycles) { status.textContent = 'Enqueued '+total+' tasks.'; return; }
if (idx >= targets.length) { enqueueNext(cycle+1, 0); return; }
const target = targets[idx];
const btn = document.getElementById('sat-btn-' + target);
if (btn && btn.disabled) { enqueueNext(cycle, idx+1); return; }
const body = {};
if (target === 'nvidia') body.diag_level = parseInt(document.getElementById('sat-nvidia-level').value)||1;
if (target === 'cpu') body.duration = parseInt(document.getElementById('sat-cpu-dur').value)||60;
fetch('/api/sat/'+target+'/run', {method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify(body)})
.then(r=>r.json()).then(()=>{
enqueued++;
status.textContent = 'Enqueued '+enqueued+'/'+total+'...';
enqueueNext(cycle, idx+1);
});
};
enqueueNext(0, 0);
}
</script>
<script>
fetch('/api/gpu/presence').then(r=>r.json()).then(gp => {
if (!gp.nvidia) disableSATCard('nvidia', 'No NVIDIA GPU detected');
if (!gp.amd) disableSATCard('amd', 'No AMD GPU detected');
});
function disableSATCard(id, reason) {
const btn = document.getElementById('sat-btn-' + id);
if (!btn) return;
btn.disabled = true;
btn.title = reason;
btn.style.opacity = '0.4';
const card = btn.closest('.card');
if (card) {
let note = card.querySelector('.sat-unavail');
if (!note) {
note = document.createElement('p');
note.className = 'sat-unavail';
note.style.cssText = 'color:var(--muted);font-size:12px;margin-top:6px';
btn.parentNode.insertBefore(note, btn.nextSibling);
}
note.textContent = reason;
}
}
</script>`
}
func renderSATCard(id, label, extra string) string {
return fmt.Sprintf(`<div class="card"><div class="card-head">%s</div><div class="card-body">%s<button class="btn btn-primary" onclick="runSAT('%s')">▶ Run Test</button></div></div>`,
label, extra, id)
return fmt.Sprintf(`<div class="card"><div class="card-head">%s</div><div class="card-body">%s<button id="sat-btn-%s" class="btn btn-primary" onclick="runSAT('%s')">▶ Run Test</button></div></div>`,
label, extra, id, id)
}
// ── Burn-in ───────────────────────────────────────────────────────────────────
// ── Burn ──────────────────────────────────────────────────────────────────────
func renderBurnIn() string {
return `<p style="color:var(--muted);font-size:13px;margin-bottom:16px">Long-running GPU and system stress tests. Check <a href="/metrics" style="color:var(--accent)">Metrics</a> page for live telemetry.</p>
<div class="grid2">
<div class="card"><div class="card-head">GPU Platform Stress</div><div class="card-body">
func renderBurn() string {
return `<div class="alert alert-warn" style="margin-bottom:16px"><strong>&#9888; Warning:</strong> Stress tests on this page run hardware at maximum load. Repeated or prolonged use may reduce hardware lifespan (storage endurance, GPU wear). Use only when necessary.</div>
<p style="color:var(--muted);font-size:13px;margin-bottom:16px">Tasks continue in the background — view progress in <a href="/tasks">Tasks</a>.</p>
<div class="grid3">
<div class="card"><div class="card-head">NVIDIA GPU Stress</div><div class="card-body">
<div class="form-row"><label>Duration</label><select id="bi-dur"><option value="600">10 minutes</option><option value="3600">1 hour</option><option value="28800">8 hours</option><option value="86400">24 hours</option></select></div>
<button class="btn btn-primary" onclick="runBurnIn('nvidia')"> Start GPU Stress</button>
<button id="sat-btn-nvidia" class="btn btn-primary" onclick="runBurnIn('nvidia')">&#9654; Start NVIDIA Stress</button>
</div></div>
<div class="card"><div class="card-head">CPU Stress</div><div class="card-body">
<div class="form-row"><label>Duration (seconds)</label><input type="number" id="bi-cpu-dur" value="300" min="60"></div>
<button class="btn btn-primary" onclick="runBurnIn('cpu')"> Start CPU Stress</button>
<button class="btn btn-primary" onclick="runBurnIn('cpu')">&#9654; Start CPU Stress</button>
</div></div>
<div class="card"><div class="card-head">AMD GPU Stress</div><div class="card-body">
<p style="color:var(--muted);font-size:12px;margin-bottom:8px">Requires ROCm tools (rocm-bandwidth-test). Missing tools reported as UNSUPPORTED.</p>
<button id="sat-btn-amd-stress" class="btn btn-primary" onclick="runBurnIn('amd-stress')">&#9654; Start AMD Stress</button>
</div></div>
<div class="card"><div class="card-head">Memory Stress</div><div class="card-body">
<p style="color:var(--muted);font-size:12px;margin-bottom:8px">stress-ng --vm writes and verifies memory patterns across all of RAM. Env: <code>BEE_VM_STRESS_SECONDS</code> (default 300), <code>BEE_VM_STRESS_SIZE_MB</code> (default 80%).</p>
<button class="btn btn-primary" onclick="runBurnIn('memory-stress')">&#9654; Start Memory Stress</button>
</div></div>
<div class="card"><div class="card-head">SAT Stress (stressapptest)</div><div class="card-body">
<p style="color:var(--muted);font-size:12px;margin-bottom:8px">Google stressapptest saturates CPU, memory and cache buses simultaneously. Env: <code>BEE_SAT_STRESS_SECONDS</code> (default 300), <code>BEE_SAT_STRESS_MB</code> (default auto).</p>
<button class="btn btn-primary" onclick="runBurnIn('sat-stress')">&#9654; Start SAT Stress</button>
</div></div>
</div>
<div id="bi-output" style="display:none;margin-top:16px" class="card">
<div class="card-head">Output</div>
<div class="card-head">Output <span id="bi-title"></span></div>
<div class="card-body"><div id="bi-terminal" class="terminal"></div></div>
</div>
<script>
let biES = null;
function runBurnIn(target) {
if (biES) biES.close();
if (biES) { biES.close(); biES = null; }
const body = {};
if (target === 'nvidia') body.duration = parseInt(document.getElementById('bi-dur').value)||600;
if (target === 'cpu') body.duration = parseInt(document.getElementById('bi-cpu-dur').value)||300;
document.getElementById('bi-output').style.display='block';
document.getElementById('bi-title').textContent = '— ' + target;
const term = document.getElementById('bi-terminal');
term.textContent = 'Starting ' + target + ' burn-in...\n';
term.textContent = 'Enqueuing ' + target + ' stress...\n';
fetch('/api/sat/'+target+'/run', {method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify(body)})
.then(r => r.json())
.then(d => {
biES = new EventSource('/api/sat/stream?job_id='+d.job_id);
term.textContent += 'Task ' + d.task_id + ' queued.\n';
biES = new EventSource('/api/tasks/'+d.task_id+'/stream');
biES.onmessage = e => { term.textContent += e.data+'\n'; term.scrollTop=term.scrollHeight; };
biES.addEventListener('done', e => { biES.close(); term.textContent += (e.data ? '\nERROR: '+e.data : '\nCompleted.')+'\n'; });
biES.addEventListener('done', e => { biES.close(); biES=null; term.textContent += (e.data ? '\nERROR: '+e.data : '\nCompleted.')+'\n'; });
});
}
</script>
<script>
fetch('/api/gpu/presence').then(r=>r.json()).then(gp => {
if (!gp.nvidia) disableSATCard('nvidia', 'No NVIDIA GPU detected');
if (!gp.amd) disableSATCard('amd-stress', 'No AMD GPU detected');
});
function disableSATCard(id, reason) {
const btn = document.getElementById('sat-btn-' + id);
if (!btn) return;
btn.disabled = true;
btn.title = reason;
btn.style.opacity = '0.4';
const card = btn.closest('.card');
if (card) {
let note = card.querySelector('.sat-unavail');
if (!note) {
note = document.createElement('p');
note.className = 'sat-unavail';
note.style.cssText = 'color:var(--muted);font-size:12px;margin-top:6px';
btn.parentNode.insertBefore(note, btn.nextSibling);
}
note.textContent = reason;
}
}
</script>`
}
// ── Network ───────────────────────────────────────────────────────────────────
func renderNetwork() string {
return `<div class="card"><div class="card-head">Network Interfaces</div><div class="card-body">
// renderNetworkInline returns the network UI without a wrapping card (for embedding in Tools).
func renderNetworkInline() string {
return `<div id="net-pending" style="display:none" class="alert alert-warn">
<strong>&#9888; Network change applied.</strong> Reverting in <span id="net-countdown">60</span>s unless confirmed.
<button class="btn btn-primary btn-sm" style="margin-left:8px" onclick="confirmNetChange()">Confirm</button>
<button class="btn btn-secondary btn-sm" style="margin-left:4px" onclick="rollbackNetChange()">Rollback</button>
</div>
<div id="iface-table"><p style="color:var(--muted);font-size:13px">Loading...</p></div>
</div></div>
<div class="grid2">
<div class="card"><div class="card-head">DHCP</div><div class="card-body">
<div class="grid2" style="margin-top:16px">
<div><div style="font-weight:700;font-size:13px;margin-bottom:8px">DHCP</div>
<div class="form-row"><label>Interface (leave empty for all)</label><input type="text" id="dhcp-iface" placeholder="eth0"></div>
<button class="btn btn-primary" onclick="runDHCP()"> Run DHCP</button>
<button class="btn btn-primary" onclick="runDHCP()">&#9654; Run DHCP</button>
<div id="dhcp-out" style="margin-top:10px;font-size:12px;color:var(--ok-fg)"></div>
</div></div>
<div class="card"><div class="card-head">Static IPv4</div><div class="card-body">
</div>
<div><div style="font-weight:700;font-size:13px;margin-bottom:8px">Static IPv4</div>
<div class="form-row"><label>Interface</label><input type="text" id="st-iface" placeholder="eth0"></div>
<div class="form-row"><label>Address</label><input type="text" id="st-addr" placeholder="192.168.1.100"></div>
<div class="form-row"><label>Prefix length</label><input type="text" id="st-prefix" placeholder="24"></div>
@@ -409,19 +665,52 @@ func renderNetwork() string {
<div class="form-row"><label>DNS (comma-separated)</label><input type="text" id="st-dns" placeholder="8.8.8.8,8.8.4.4"></div>
<button class="btn btn-primary" onclick="setStatic()">Apply Static IP</button>
<div id="static-out" style="margin-top:10px;font-size:12px;color:var(--ok-fg)"></div>
</div></div>
</div>
</div>
<script>
var _netCountdownTimer = null;
function loadNetwork() {
fetch('/api/network').then(r=>r.json()).then(d => {
const rows = (d.interfaces||[]).map(i =>
'<tr><td>'+i.Name+'</td><td><span class="badge '+(i.State==='up'?'badge-ok':'badge-warn')+'">'+i.State+'</span></td><td>'+(i.IPv4||[]).join(', ')+'</td></tr>'
'<tr><td>'+i.Name+'</td>' +
'<td style="cursor:pointer" onclick="toggleIface(\''+i.Name+'\',\''+i.State+'\')" title="Click to toggle"><span class="badge '+(i.State==='up'?'badge-ok':'badge-warn')+'">'+i.State+'</span></td>' +
'<td>'+(i.IPv4||[]).join(', ')+'</td></tr>'
).join('');
document.getElementById('iface-table').innerHTML =
'<table><tr><th>Interface</th><th>State</th><th>Addresses</th></tr>'+rows+'</table>' +
'<table><tr><th>Interface</th><th>State (click to toggle)</th><th>Addresses</th></tr>'+rows+'</table>' +
(d.default_route ? '<p style="font-size:12px;color:var(--muted);margin-top:8px">Default route: '+d.default_route+'</p>' : '');
});
}
function toggleIface(iface, currentState) {
fetch('/api/network/toggle',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({iface:iface})})
.then(r=>r.json()).then(d => {
if (d.error) { alert('Error: '+d.error); return; }
loadNetwork();
showNetPending(d.rollback_in || 60);
});
}
function showNetPending(secs) {
const el = document.getElementById('net-pending');
el.style.display = 'block';
if (_netCountdownTimer) clearInterval(_netCountdownTimer);
let remaining = secs;
document.getElementById('net-countdown').textContent = remaining;
_netCountdownTimer = setInterval(function() {
remaining--;
document.getElementById('net-countdown').textContent = remaining;
if (remaining <= 0) { clearInterval(_netCountdownTimer); _netCountdownTimer=null; el.style.display='none'; loadNetwork(); }
}, 1000);
}
function confirmNetChange() {
if (_netCountdownTimer) { clearInterval(_netCountdownTimer); _netCountdownTimer=null; }
document.getElementById('net-pending').style.display='none';
fetch('/api/network/confirm',{method:'POST'});
}
function rollbackNetChange() {
if (_netCountdownTimer) { clearInterval(_netCountdownTimer); _netCountdownTimer=null; }
document.getElementById('net-pending').style.display='none';
fetch('/api/network/rollback',{method:'POST'}).then(()=>loadNetwork());
}
function runDHCP() {
const iface = document.getElementById('dhcp-iface').value.trim();
fetch('/api/network/dhcp',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({interface:iface||'all'})})
@@ -447,13 +736,17 @@ loadNetwork();
</script>`
}
func renderNetwork() string {
return `<div class="card"><div class="card-head">Network Interfaces</div><div class="card-body">` +
renderNetworkInline() +
`</div></div>`
}
// ── Services ──────────────────────────────────────────────────────────────────
func renderServices() string {
return `<div class="card"><div class="card-head">Bee Services <button class="btn btn-sm btn-secondary" onclick="loadServices()" style="margin-left:auto">↻ Refresh</button></div>
<div class="card-body">
func renderServicesInline() string {
return `<div style="display:flex;justify-content:flex-end;margin-bottom:8px"><button class="btn btn-sm btn-secondary" onclick="loadServices()">&#8635; Refresh</button></div>
<div id="svc-table"><p style="color:var(--muted);font-size:13px">Loading...</p></div>
</div></div>
<div id="svc-out" style="display:none;margin-top:8px" class="card">
<div class="card-head">Output</div>
<div class="card-body" style="padding:10px"><div id="svc-terminal" class="terminal" style="max-height:150px"></div></div>
@@ -497,6 +790,12 @@ loadServices();
</script>`
}
func renderServices() string {
return `<div class="card"><div class="card-head">Bee Services</div><div class="card-body">` +
renderServicesInline() +
`</div></div>`
}
// ── Export ────────────────────────────────────────────────────────────────────
func renderExport(exportDir string) string {
@@ -546,14 +845,56 @@ func listExportFiles(exportDir string) ([]string, error) {
// ── Tools ─────────────────────────────────────────────────────────────────────
func renderTools() string {
return `<div class="card"><div class="card-head">Tool Check <button class="btn btn-sm btn-secondary" onclick="checkTools()" style="margin-left:auto">↻ Check</button></div>
<div class="card-body"><div id="tools-table"><p style="color:var(--muted);font-size:13px">Click Check to verify installed tools.</p></div></div></div>
return `<div class="card" style="margin-bottom:16px">
<div class="card-head">Install to RAM</div>
<div class="card-body">
<p id="ram-status-text" style="color:var(--muted);font-size:13px;margin-bottom:8px">Checking...</p>
<button id="ram-install-btn" class="btn btn-primary" onclick="installToRAM()" style="display:none">&#9654; Copy to RAM</button>
</div>
</div>
<script>
fetch('/api/system/ram-status').then(r=>r.json()).then(d=>{
const txt = document.getElementById('ram-status-text');
const btn = document.getElementById('ram-install-btn');
if (d.in_ram) {
txt.textContent = '✓ Running from RAM — installation media can be safely disconnected.';
txt.style.color = 'var(--ok, green)';
} else {
txt.textContent = 'Live media is mounted from installation device. Copy to RAM to allow media removal.';
btn.style.display = '';
}
});
function installToRAM() {
document.getElementById('ram-install-btn').disabled = true;
fetch('/api/system/install-to-ram', {method:'POST'}).then(r=>r.json()).then(d=>{
window.location.href = '/tasks#' + d.task_id;
});
}
</script>
<div class="card"><div class="card-head">Support Bundle</div><div class="card-body">
<p style="font-size:13px;color:var(--muted);margin-bottom:12px">Downloads a tar.gz archive of all audit files, SAT results, and logs.</p>
<a class="btn btn-primary" href="/export/support.tar.gz">&#8595; Download Support Bundle</a>
</div></div>
<div class="card"><div class="card-head">Tool Check <button class="btn btn-sm btn-secondary" onclick="checkTools()" style="margin-left:auto">&#8635; Check</button></div>
<div class="card-body"><div id="tools-table"><p style="color:var(--muted);font-size:13px">Checking...</p></div></div></div>
<div class="card"><div class="card-head">Network</div><div class="card-body">` +
renderNetworkInline() + `</div></div>
<div class="card"><div class="card-head">Services</div><div class="card-body">` +
renderServicesInline() + `</div></div>
<div class="card"><div class="card-head">Install to Disk</div><div class="card-body">` +
renderInstallInline() + `</div></div>
<script>
function checkTools() {
document.getElementById('tools-table').innerHTML = '<p style="color:var(--muted);font-size:13px">Checking...</p>';
fetch('/api/tools/check').then(r=>r.json()).then(tools => {
const rows = tools.map(t =>
'<tr><td>'+t.Name+'</td><td><span class="badge '+(t.OK ? 'badge-ok' : 'badge-err')+'">'+(t.OK ? ' '+t.Path : ' missing')+'</span></td></tr>'
'<tr><td>'+t.Name+'</td><td><span class="badge '+(t.OK ? 'badge-ok' : 'badge-err')+'">'+(t.OK ? '&#10003; '+t.Path : '&#10007; missing')+'</span></td></tr>'
).join('');
document.getElementById('tools-table').innerHTML =
'<table><tr><th>Tool</th><th>Status</th></tr>'+rows+'</table>';
@@ -565,11 +906,8 @@ checkTools();
// ── Install to Disk ──────────────────────────────────────────────────────────
func renderInstall() string {
func renderInstallInline() string {
return `
<div class="card">
<div class="card-head">Install Live System to Disk</div>
<div class="card-body">
<div class="alert alert-warn" style="margin-bottom:16px">
<strong>Warning:</strong> Installing will <strong>completely erase</strong> the selected
disk and write the live system onto it. All existing data on the target disk will be lost.
@@ -767,6 +1105,107 @@ installRefreshDisks();
`
}
func renderInstall() string {
return `<div class="card"><div class="card-head">Install Live System to Disk</div><div class="card-body">` +
renderInstallInline() +
`</div></div>`
}
// ── Tasks ─────────────────────────────────────────────────────────────────────
func renderTasks() string {
return `<div style="display:flex;align-items:center;gap:12px;margin-bottom:16px">
<button class="btn btn-danger btn-sm" onclick="cancelAll()">Cancel All</button>
<span style="font-size:12px;color:var(--muted)">Tasks run one at a time. Logs persist after navigation.</span>
</div>
<div class="card">
<div id="tasks-table"><p style="color:var(--muted);font-size:13px;padding:16px">Loading...</p></div>
</div>
<div id="task-log-section" style="display:none;margin-top:16px" class="card">
<div class="card-head">Logs — <span id="task-log-title"></span>
<button class="btn btn-sm btn-secondary" onclick="closeTaskLog()" style="margin-left:auto">&#10005;</button>
</div>
<div class="card-body"><div id="task-log-terminal" class="terminal" style="max-height:500px"></div></div>
</div>
<script>
var _taskLogES = null;
var _taskRefreshTimer = null;
function loadTasks() {
fetch('/api/tasks').then(r=>r.json()).then(tasks => {
if (!tasks || tasks.length === 0) {
document.getElementById('tasks-table').innerHTML = '<p style="color:var(--muted);font-size:13px;padding:16px">No tasks.</p>';
return;
}
const rows = tasks.map(t => {
const dur = t.started_at ? formatDur(t.started_at, t.done_at) : '';
const statusClass = {running:'badge-ok',pending:'badge-unknown',done:'badge-ok',failed:'badge-err',cancelled:'badge-unknown'}[t.status]||'badge-unknown';
const statusLabel = {running:'&#9654; running',pending:'pending',done:'&#10003; done',failed:'&#10007; failed',cancelled:'cancelled'}[t.status]||t.status;
let actions = '<button class="btn btn-sm btn-secondary" onclick="viewLog(\''+t.id+'\',\''+escHtml(t.name)+'\')">Logs</button>';
if (t.status === 'running' || t.status === 'pending') {
actions += ' <button class="btn btn-sm btn-danger" onclick="cancelTask(\''+t.id+'\')">Cancel</button>';
}
if (t.status === 'pending') {
actions += ' <button class="btn btn-sm btn-secondary" onclick="setPriority(\''+t.id+'\',1)" title="Increase priority">&#8679;</button>';
actions += ' <button class="btn btn-sm btn-secondary" onclick="setPriority(\''+t.id+'\',-1)" title="Decrease priority">&#8681;</button>';
}
return '<tr><td>'+escHtml(t.name)+'</td>' +
'<td><span class="badge '+statusClass+'">'+statusLabel+'</span></td>' +
'<td style="font-size:12px;color:var(--muted)">'+fmtTime(t.created_at)+'</td>' +
'<td style="font-size:12px;color:var(--muted)">'+dur+'</td>' +
'<td>'+t.priority+'</td>' +
'<td>'+actions+'</td></tr>';
}).join('');
document.getElementById('tasks-table').innerHTML =
'<table><tr><th>Name</th><th>Status</th><th>Created</th><th>Duration</th><th>Priority</th><th>Actions</th></tr>'+rows+'</table>';
});
}
function escHtml(s) { return (s||'').replace(/&/g,'&amp;').replace(/</g,'&lt;').replace(/>/g,'&gt;').replace(/"/g,'&quot;'); }
function fmtTime(s) { if (!s) return ''; try { return new Date(s).toLocaleTimeString(); } catch(e){ return s; } }
function formatDur(start, end) {
try {
const s = new Date(start), e = end ? new Date(end) : new Date();
const sec = Math.round((e-s)/1000);
if (sec < 60) return sec+'s';
const m = Math.floor(sec/60), ss = sec%60;
return m+'m '+ss+'s';
} catch(e){ return ''; }
}
function cancelTask(id) {
fetch('/api/tasks/'+id+'/cancel',{method:'POST'}).then(()=>loadTasks());
}
function cancelAll() {
fetch('/api/tasks/cancel-all',{method:'POST'}).then(()=>loadTasks());
}
function setPriority(id, delta) {
fetch('/api/tasks/'+id+'/priority',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({delta:delta})})
.then(()=>loadTasks());
}
function viewLog(id, name) {
if (_taskLogES) { _taskLogES.close(); _taskLogES = null; }
document.getElementById('task-log-section').style.display = '';
document.getElementById('task-log-title').textContent = name;
const term = document.getElementById('task-log-terminal');
term.textContent = 'Connecting...\n';
_taskLogES = new EventSource('/api/tasks/'+id+'/stream');
_taskLogES.onmessage = e => { term.textContent += e.data+'\n'; term.scrollTop=term.scrollHeight; };
_taskLogES.addEventListener('done', e => {
_taskLogES.close(); _taskLogES=null;
term.textContent += (e.data ? '\nERROR: '+e.data : '\nDone.')+'\n';
});
}
function closeTaskLog() {
if (_taskLogES) { _taskLogES.close(); _taskLogES=null; }
document.getElementById('task-log-section').style.display='none';
}
loadTasks();
_taskRefreshTimer = setInterval(loadTasks, 2000);
</script>`
}
func renderExportIndex(exportDir string) (string, error) {
entries, err := listExportFiles(exportDir)
if err != nil {

View File

@@ -4,6 +4,7 @@ import (
"encoding/json"
"errors"
"fmt"
"mime"
"net/http"
"os"
"path/filepath"
@@ -20,6 +21,17 @@ import (
const defaultTitle = "Bee Hardware Audit"
func init() {
// On some LiveCD ramdisk environments, /usr/share/mime/globs2 exists but
// causes an I/O error mid-read. Go's mime package panics (not errors) in
// that case, crashing the first HTTP goroutine that serves a static file.
// Pre-trigger initialization here with recover so subsequent calls are safe.
func() {
defer func() { recover() }() //nolint:errcheck
mime.TypeByExtension(".gz")
}()
}
// HandlerOptions configures the web UI handler.
type HandlerOptions struct {
Title string
@@ -31,14 +43,14 @@ type HandlerOptions struct {
// metricsRing holds a rolling window of live metric samples.
type metricsRing struct {
mu sync.Mutex
vals []float64
labels []string
size int
mu sync.Mutex
vals []float64
times []time.Time
size int
}
func newMetricsRing(size int) *metricsRing {
return &metricsRing{size: size, vals: make([]float64, 0, size), labels: make([]string, 0, size)}
return &metricsRing{size: size, vals: make([]float64, 0, size), times: make([]time.Time, 0, size)}
}
func (r *metricsRing) push(v float64) {
@@ -46,20 +58,40 @@ func (r *metricsRing) push(v float64) {
defer r.mu.Unlock()
if len(r.vals) >= r.size {
r.vals = r.vals[1:]
r.labels = r.labels[1:]
r.times = r.times[1:]
}
r.vals = append(r.vals, v)
r.labels = append(r.labels, time.Now().Format("15:04"))
r.times = append(r.times, time.Now())
}
func (r *metricsRing) snapshot() ([]float64, []string) {
r.mu.Lock()
defer r.mu.Unlock()
v := make([]float64, len(r.vals))
l := make([]string, len(r.labels))
copy(v, r.vals)
copy(l, r.labels)
return v, l
now := time.Now()
labels := make([]string, len(r.times))
for i, t := range r.times {
labels[i] = relAgeLabel(now.Sub(t))
}
return v, labels
}
func relAgeLabel(age time.Duration) string {
if age <= 0 {
return "0"
}
if age < time.Hour {
m := int(age.Minutes())
if m == 0 {
return "-<1m"
}
return fmt.Sprintf("-%dm", m)
}
if age < 24*time.Hour {
return fmt.Sprintf("-%dh", int(age.Hours()))
}
return fmt.Sprintf("-%dd", int(age.Hours()/24))
}
// gpuRings holds per-GPU ring buffers.
@@ -70,6 +102,14 @@ type gpuRings struct {
Power *metricsRing
}
// pendingNetChange tracks a network state change awaiting confirmation.
type pendingNetChange struct {
iface string
wasUp bool
timer *time.Timer
mu sync.Mutex
}
// handler is the HTTP handler for the web UI.
type handler struct {
opts HandlerOptions
@@ -87,6 +127,9 @@ type handler struct {
// install job (at most one at a time)
installJob *jobState
installMu sync.Mutex
// pending network change (rollback on timeout)
pendingNet *pendingNetChange
pendingNetMu sync.Mutex
}
// NewHandler creates the HTTP mux with all routes.
@@ -108,6 +151,7 @@ func NewHandler(opts HandlerOptions) http.Handler {
ringMemLoad: newMetricsRing(120),
ringPower: newMetricsRing(120),
}
globalQueue.startWorker(&opts)
mux := http.NewServeMux()
// ── Infrastructure ──────────────────────────────────────────────────────
@@ -131,9 +175,20 @@ func NewHandler(opts HandlerOptions) http.Handler {
mux.HandleFunc("POST /api/sat/memory/run", h.handleAPISATRun("memory"))
mux.HandleFunc("POST /api/sat/storage/run", h.handleAPISATRun("storage"))
mux.HandleFunc("POST /api/sat/cpu/run", h.handleAPISATRun("cpu"))
mux.HandleFunc("POST /api/sat/amd/run", h.handleAPISATRun("amd"))
mux.HandleFunc("POST /api/sat/amd-stress/run", h.handleAPISATRun("amd-stress"))
mux.HandleFunc("POST /api/sat/memory-stress/run", h.handleAPISATRun("memory-stress"))
mux.HandleFunc("POST /api/sat/sat-stress/run", h.handleAPISATRun("sat-stress"))
mux.HandleFunc("GET /api/sat/stream", h.handleAPISATStream)
mux.HandleFunc("POST /api/sat/abort", h.handleAPISATAbort)
// Tasks
mux.HandleFunc("GET /api/tasks", h.handleAPITasksList)
mux.HandleFunc("POST /api/tasks/cancel-all", h.handleAPITasksCancelAll)
mux.HandleFunc("POST /api/tasks/{id}/cancel", h.handleAPITasksCancel)
mux.HandleFunc("POST /api/tasks/{id}/priority", h.handleAPITasksPriority)
mux.HandleFunc("GET /api/tasks/{id}/stream", h.handleAPITasksStream)
// Services
mux.HandleFunc("GET /api/services", h.handleAPIServicesList)
mux.HandleFunc("POST /api/services/action", h.handleAPIServicesAction)
@@ -142,6 +197,9 @@ func NewHandler(opts HandlerOptions) http.Handler {
mux.HandleFunc("GET /api/network", h.handleAPINetworkStatus)
mux.HandleFunc("POST /api/network/dhcp", h.handleAPINetworkDHCP)
mux.HandleFunc("POST /api/network/static", h.handleAPINetworkStatic)
mux.HandleFunc("POST /api/network/toggle", h.handleAPINetworkToggle)
mux.HandleFunc("POST /api/network/confirm", h.handleAPINetworkConfirm)
mux.HandleFunc("POST /api/network/rollback", h.handleAPINetworkRollback)
// Export
mux.HandleFunc("GET /api/export/list", h.handleAPIExportList)
@@ -150,6 +208,13 @@ func NewHandler(opts HandlerOptions) http.Handler {
// Tools
mux.HandleFunc("GET /api/tools/check", h.handleAPIToolsCheck)
// GPU presence
mux.HandleFunc("GET /api/gpu/presence", h.handleAPIGPUPresence)
// System
mux.HandleFunc("GET /api/system/ram-status", h.handleAPIRAMStatus)
mux.HandleFunc("POST /api/system/install-to-ram", h.handleAPIInstallToRAM)
// Preflight
mux.HandleFunc("GET /api/preflight", h.handleAPIPreflight)
@@ -197,6 +262,11 @@ func (h *handler) handleAuditJSON(w http.ResponseWriter, r *http.Request) {
http.Error(w, fmt.Sprintf("read audit snapshot: %v", err), http.StatusInternalServerError)
return
}
// Re-apply SAT overlay on every request so that SAT results run after the
// last audit always appear in the downloaded JSON without needing a re-audit.
if overlaid, err := app.ApplySATOverlay(data); err == nil {
data = overlaid
}
w.Header().Set("Cache-Control", "no-store")
w.Header().Set("Content-Type", "application/json; charset=utf-8")
_, _ = w.Write(data)
@@ -240,9 +310,33 @@ func (h *handler) handleExportFile(w http.ResponseWriter, r *http.Request) {
http.Error(w, "invalid path", http.StatusBadRequest)
return
}
// Set Content-Type explicitly to avoid mime.TypeByExtension which panics on
// LiveCD environments where /usr/share/mime/globs2 has an I/O read error.
w.Header().Set("Content-Type", mimeByExt(filepath.Ext(clean)))
http.ServeFile(w, r, filepath.Join(h.opts.ExportDir, clean))
}
// mimeByExt returns a Content-Type for known extensions, falling back to
// application/octet-stream. Used to avoid calling mime.TypeByExtension.
func mimeByExt(ext string) string {
switch strings.ToLower(ext) {
case ".json":
return "application/json"
case ".gz":
return "application/gzip"
case ".tar":
return "application/x-tar"
case ".log", ".txt":
return "text/plain; charset=utf-8"
case ".html":
return "text/html; charset=utf-8"
case ".svg":
return "image/svg+xml"
default:
return "application/octet-stream"
}
}
func (h *handler) handleExportIndex(w http.ResponseWriter, r *http.Request) {
body, err := renderExportIndex(h.opts.ExportDir)
if err != nil {
@@ -274,18 +368,35 @@ func (h *handler) handleMetricsChartSVG(w http.ResponseWriter, r *http.Request)
var names []string
var labels []string
var title string
var yMin, yMax *float64 // nil = auto; for load charts fixed 0-100
switch {
case path == "server":
title = "Server"
vCPUTemp, l := h.ringCPUTemp.snapshot()
vCPULoad, _ := h.ringCPULoad.snapshot()
// ── Server sub-charts ─────────────────────────────────────────────────
case path == "server-load":
title = "CPU / Memory Load"
vCPULoad, l := h.ringCPULoad.snapshot()
vMemLoad, _ := h.ringMemLoad.snapshot()
vPower, _ := h.ringPower.snapshot()
labels = l
datasets = [][]float64{vCPUTemp, vCPULoad, vMemLoad, vPower}
names = []string{"CPU Temp °C", "CPU Load %", "Mem Load %", "Power W"}
datasets = [][]float64{vCPULoad, vMemLoad}
names = []string{"CPU Load %", "Mem Load %"}
yMin = floatPtr(0)
yMax = floatPtr(100)
case path == "server-temp":
title = "CPU Temperature"
vCPUTemp, l := h.ringCPUTemp.snapshot()
labels = l
datasets = [][]float64{vCPUTemp}
names = []string{"CPU Temp °C"}
yMin = floatPtr(0)
yMax = autoMax120(vCPUTemp)
case path == "server-power":
title = "Power & Fans"
vPower, l := h.ringPower.snapshot()
labels = l
datasets = [][]float64{vPower}
names = []string{"Power W"}
h.ringsMu.Lock()
for i, fr := range h.ringFans {
fv, _ := fr.snapshot()
@@ -297,11 +408,20 @@ func (h *handler) handleMetricsChartSVG(w http.ResponseWriter, r *http.Request)
names = append(names, name+" RPM")
}
h.ringsMu.Unlock()
yMin = floatPtr(0)
yMax = autoMax120(datasets...)
// ── GPU sub-charts ────────────────────────────────────────────────────
case strings.HasPrefix(path, "gpu/"):
idxStr := strings.TrimPrefix(path, "gpu/")
rest := strings.TrimPrefix(path, "gpu/")
// rest is either "{idx}-load", "{idx}-temp", "{idx}-power", or legacy "{idx}"
sub := ""
if i := strings.LastIndex(rest, "-"); i > 0 {
sub = rest[i+1:]
rest = rest[:i]
}
idx := 0
fmt.Sscanf(idxStr, "%d", &idx)
fmt.Sscanf(rest, "%d", &idx)
h.ringsMu.Lock()
var gr *gpuRings
if idx < len(h.gpuRings) {
@@ -312,21 +432,71 @@ func (h *handler) handleMetricsChartSVG(w http.ResponseWriter, r *http.Request)
http.NotFound(w, r)
return
}
vTemp, l := gr.Temp.snapshot()
vUtil, _ := gr.Util.snapshot()
vMemUtil, _ := gr.MemUtil.snapshot()
vPower, _ := gr.Power.snapshot()
labels = l
title = fmt.Sprintf("GPU %d", idx)
datasets = [][]float64{vTemp, vUtil, vMemUtil, vPower}
names = []string{"Temp °C", "Load %", "Mem %", "Power W"}
switch sub {
case "load":
vUtil, l := gr.Util.snapshot()
vMemUtil, _ := gr.MemUtil.snapshot()
labels = l
title = fmt.Sprintf("GPU %d Load", idx)
datasets = [][]float64{vUtil, vMemUtil}
names = []string{"Load %", "Mem %"}
yMin = floatPtr(0)
yMax = floatPtr(100)
case "temp":
vTemp, l := gr.Temp.snapshot()
labels = l
title = fmt.Sprintf("GPU %d Temperature", idx)
datasets = [][]float64{vTemp}
names = []string{"Temp °C"}
yMin = floatPtr(0)
yMax = autoMax120(vTemp)
default: // "power" or legacy (no sub)
vPower, l := gr.Power.snapshot()
labels = l
title = fmt.Sprintf("GPU %d Power", idx)
datasets = [][]float64{vPower}
names = []string{"Power W"}
yMin = floatPtr(0)
yMax = autoMax120(vPower)
}
default:
http.NotFound(w, r)
return
}
// Ensure all datasets same length as labels
buf, err := renderChartSVG(title, datasets, names, labels, yMin, yMax)
if err != nil {
http.Error(w, err.Error(), http.StatusInternalServerError)
return
}
w.Header().Set("Content-Type", "image/svg+xml")
w.Header().Set("Cache-Control", "no-store")
_, _ = w.Write(buf)
}
// floatPtr returns a pointer to a float64 value.
func floatPtr(v float64) *float64 { return &v }
// autoMax120 returns 0→max+20% Y-axis max across all datasets.
func autoMax120(datasets ...[]float64) *float64 {
max := 0.0
for _, ds := range datasets {
for _, v := range ds {
if v > max {
max = v
}
}
}
if max == 0 {
return nil // let library auto-scale
}
v := max * 1.2
return &v
}
// renderChartSVG renders a line chart SVG with a fixed Y-axis range.
func renderChartSVG(title string, datasets [][]float64, names []string, labels []string, yMin, yMax *float64) ([]byte, error) {
n := len(labels)
if n == 0 {
n = 1
@@ -337,31 +507,25 @@ func (h *handler) handleMetricsChartSVG(w http.ResponseWriter, r *http.Request)
datasets[i] = make([]float64, n)
}
}
sparse := sparseLabels(labels, 6)
opt := gocharts.NewLineChartOptionWithData(datasets)
opt.Title = gocharts.TitleOption{Text: title}
opt.XAxis.Labels = sparse
opt.Legend = gocharts.LegendOption{SeriesNames: names}
if yMin != nil || yMax != nil {
opt.YAxis = []gocharts.YAxisOption{{Min: yMin, Max: yMax}}
}
p := gocharts.NewPainter(gocharts.PainterOptions{
OutputFormat: gocharts.ChartOutputSVG,
Width: 1400,
Height: 280,
Height: 240,
}, gocharts.PainterThemeOption(gocharts.GetTheme("grafana")))
if err := p.LineChart(opt); err != nil {
http.Error(w, err.Error(), http.StatusInternalServerError)
return
return nil, err
}
buf, err := p.Bytes()
if err != nil {
http.Error(w, err.Error(), http.StatusInternalServerError)
return
}
w.Header().Set("Content-Type", "image/svg+xml")
w.Header().Set("Cache-Control", "no-store")
_, _ = w.Write(buf)
return p.Bytes()
}
func safeIdx(s []float64, i int) float64 {
@@ -392,6 +556,15 @@ func (h *handler) handlePage(w http.ResponseWriter, r *http.Request) {
if page == "" {
page = "dashboard"
}
// Redirect old routes to new names
switch page {
case "tests":
http.Redirect(w, r, "/validate", http.StatusMovedPermanently)
return
case "burn-in":
http.Redirect(w, r, "/burn", http.StatusMovedPermanently)
return
}
body := renderPage(page, h.opts)
w.Header().Set("Cache-Control", "no-store")
w.Header().Set("Content-Type", "text/html; charset=utf-8")

View File

@@ -9,7 +9,7 @@ import (
"testing"
)
func TestRootRendersShellWithIframe(t *testing.T) {
func TestRootRendersDashboard(t *testing.T) {
dir := t.TempDir()
path := filepath.Join(dir, "audit.json")
exportDir := filepath.Join(dir, "export")
@@ -31,11 +31,12 @@ func TestRootRendersShellWithIframe(t *testing.T) {
if first.Code != http.StatusOK {
t.Fatalf("first status=%d", first.Code)
}
if !strings.Contains(first.Body.String(), `iframe`) || !strings.Contains(first.Body.String(), `src="/viewer"`) {
t.Fatalf("first body missing iframe viewer: %s", first.Body.String())
// Dashboard should contain the audit modal (with viewer link) and hardware summary
if !strings.Contains(first.Body.String(), `openAuditModal`) {
t.Fatalf("first body missing audit modal trigger: %s", first.Body.String())
}
if !strings.Contains(first.Body.String(), "/export/support.tar.gz") {
t.Fatalf("first body missing support bundle link: %s", first.Body.String())
if !strings.Contains(first.Body.String(), `/viewer`) {
t.Fatalf("first body missing viewer link: %s", first.Body.String())
}
if got := first.Header().Get("Cache-Control"); got != "no-store" {
t.Fatalf("first cache-control=%q", got)
@@ -50,8 +51,8 @@ func TestRootRendersShellWithIframe(t *testing.T) {
if second.Code != http.StatusOK {
t.Fatalf("second status=%d", second.Code)
}
if !strings.Contains(second.Body.String(), `src="/viewer"`) {
t.Fatalf("second body missing iframe viewer: %s", second.Body.String())
if !strings.Contains(second.Body.String(), `Hardware Summary`) {
t.Fatalf("second body missing hardware summary: %s", second.Body.String())
}
}
@@ -103,8 +104,8 @@ func TestAuditJSONServesLatestSnapshot(t *testing.T) {
if rec.Code != http.StatusOK {
t.Fatalf("status=%d", rec.Code)
}
if got := strings.TrimSpace(rec.Body.String()); got != body {
t.Fatalf("body=%q want %q", got, body)
if !strings.Contains(rec.Body.String(), "SERIAL-API") {
t.Fatalf("body missing expected serial: %s", rec.Body.String())
}
if got := rec.Header().Get("Content-Type"); !strings.Contains(got, "application/json") {
t.Fatalf("content-type=%q", got)

View File

@@ -0,0 +1,420 @@
package webui
import (
"context"
"encoding/json"
"fmt"
"net/http"
"sort"
"sync"
"time"
)
// Task statuses.
const (
TaskPending = "pending"
TaskRunning = "running"
TaskDone = "done"
TaskFailed = "failed"
TaskCancelled = "cancelled"
)
// taskNames maps target → human-readable name.
var taskNames = map[string]string{
"nvidia": "NVIDIA SAT",
"memory": "Memory SAT",
"storage": "Storage SAT",
"cpu": "CPU SAT",
"amd": "AMD GPU SAT",
"amd-stress": "AMD GPU Burn-in",
"memory-stress": "Memory Burn-in",
"sat-stress": "SAT Stress (stressapptest)",
"audit": "Audit",
"install": "Install to Disk",
"install-to-ram": "Install to RAM",
}
// Task represents one unit of work in the queue.
type Task struct {
ID string `json:"id"`
Name string `json:"name"`
Target string `json:"target"`
Priority int `json:"priority"`
Status string `json:"status"`
CreatedAt time.Time `json:"created_at"`
StartedAt *time.Time `json:"started_at,omitempty"`
DoneAt *time.Time `json:"done_at,omitempty"`
ErrMsg string `json:"error,omitempty"`
// runtime fields (not serialised)
job *jobState
params taskParams
}
// taskParams holds optional parameters parsed from the run request.
type taskParams struct {
Duration int
DiagLevel int
GPUIndices []int
Device string // for install
}
// taskQueue manages a priority-ordered list of tasks and runs them one at a time.
type taskQueue struct {
mu sync.Mutex
tasks []*Task
trigger chan struct{}
opts *HandlerOptions // set by startWorker
}
var globalQueue = &taskQueue{trigger: make(chan struct{}, 1)}
const maxTaskHistory = 50
// enqueue adds a task to the queue and notifies the worker.
func (q *taskQueue) enqueue(t *Task) {
q.mu.Lock()
q.tasks = append(q.tasks, t)
q.prune()
q.mu.Unlock()
select {
case q.trigger <- struct{}{}:
default:
}
}
// prune removes oldest completed tasks beyond maxTaskHistory.
func (q *taskQueue) prune() {
var done []*Task
var active []*Task
for _, t := range q.tasks {
switch t.Status {
case TaskDone, TaskFailed, TaskCancelled:
done = append(done, t)
default:
active = append(active, t)
}
}
if len(done) > maxTaskHistory {
done = done[len(done)-maxTaskHistory:]
}
q.tasks = append(active, done...)
}
// nextPending returns the highest-priority pending task (nil if none).
func (q *taskQueue) nextPending() *Task {
var best *Task
for _, t := range q.tasks {
if t.Status != TaskPending {
continue
}
if best == nil || t.Priority > best.Priority ||
(t.Priority == best.Priority && t.CreatedAt.Before(best.CreatedAt)) {
best = t
}
}
return best
}
// findByID looks up a task by ID.
func (q *taskQueue) findByID(id string) (*Task, bool) {
q.mu.Lock()
defer q.mu.Unlock()
for _, t := range q.tasks {
if t.ID == id {
return t, true
}
}
return nil, false
}
// findJob returns the jobState for a task ID (for SSE streaming compatibility).
func (q *taskQueue) findJob(id string) (*jobState, bool) {
t, ok := q.findByID(id)
if !ok || t.job == nil {
return nil, false
}
return t.job, true
}
// snapshot returns a copy of all tasks sorted for display (running first, then pending by priority, then done by doneAt desc).
func (q *taskQueue) snapshot() []Task {
q.mu.Lock()
defer q.mu.Unlock()
out := make([]Task, len(q.tasks))
for i, t := range q.tasks {
out[i] = *t
}
sort.SliceStable(out, func(i, j int) bool {
si := statusOrder(out[i].Status)
sj := statusOrder(out[j].Status)
if si != sj {
return si < sj
}
if out[i].Priority != out[j].Priority {
return out[i].Priority > out[j].Priority
}
return out[i].CreatedAt.Before(out[j].CreatedAt)
})
return out
}
func statusOrder(s string) int {
switch s {
case TaskRunning:
return 0
case TaskPending:
return 1
default:
return 2
}
}
// startWorker launches the queue runner goroutine.
func (q *taskQueue) startWorker(opts *HandlerOptions) {
q.opts = opts
go q.worker()
}
func (q *taskQueue) worker() {
for {
<-q.trigger
for {
q.mu.Lock()
t := q.nextPending()
if t == nil {
q.mu.Unlock()
break
}
now := time.Now()
t.Status = TaskRunning
t.StartedAt = &now
j := &jobState{}
ctx, cancel := context.WithCancel(context.Background())
j.cancel = cancel
t.job = j
q.mu.Unlock()
q.runTask(t, j, ctx)
q.mu.Lock()
now2 := time.Now()
t.DoneAt = &now2
if t.Status == TaskRunning { // not cancelled externally
if j.err != "" {
t.Status = TaskFailed
t.ErrMsg = j.err
} else {
t.Status = TaskDone
}
}
q.prune()
q.mu.Unlock()
}
}
}
// runTask executes the work for a task, writing output to j.
func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
if q.opts == nil || q.opts.App == nil {
j.append("ERROR: app not configured")
j.finish("app not configured")
return
}
a := q.opts.App
j.append(fmt.Sprintf("Starting %s...", t.Name))
var (
archive string
err error
)
switch t.Target {
case "nvidia":
if len(t.params.GPUIndices) > 0 || t.params.DiagLevel > 0 {
result, e := a.RunNvidiaAcceptancePackWithOptions(
ctx, "", t.params.DiagLevel, t.params.GPUIndices, j.append,
)
if e != nil {
err = e
} else {
archive = result.Body
}
} else {
archive, err = a.RunNvidiaAcceptancePack("", j.append)
}
case "memory":
archive, err = a.RunMemoryAcceptancePack("", j.append)
case "storage":
archive, err = a.RunStorageAcceptancePack("", j.append)
case "cpu":
dur := t.params.Duration
if dur <= 0 {
dur = 60
}
archive, err = a.RunCPUAcceptancePack("", dur, j.append)
case "amd":
archive, err = a.RunAMDAcceptancePack("", j.append)
case "amd-stress":
archive, err = a.RunAMDStressPack("", j.append)
case "memory-stress":
archive, err = a.RunMemoryStressPack("", j.append)
case "sat-stress":
archive, err = a.RunSATStressPack("", j.append)
case "audit":
result, e := a.RunAuditNow(q.opts.RuntimeMode)
if e != nil {
err = e
} else {
for _, line := range splitLines(result.Body) {
j.append(line)
}
}
case "install-to-ram":
err = a.RunInstallToRAM(j.append)
default:
j.append("ERROR: unknown target: " + t.Target)
j.finish("unknown target")
return
}
if err != nil {
if ctx.Err() != nil {
j.append("Aborted.")
j.finish("aborted")
} else {
j.append("ERROR: " + err.Error())
j.finish(err.Error())
}
return
}
if archive != "" {
j.append("Archive: " + archive)
}
j.finish("")
}
func splitLines(s string) []string {
var out []string
for _, l := range splitNL(s) {
if l != "" {
out = append(out, l)
}
}
return out
}
func splitNL(s string) []string {
var out []string
start := 0
for i, c := range s {
if c == '\n' {
out = append(out, s[start:i])
start = i + 1
}
}
out = append(out, s[start:])
return out
}
// ── HTTP handlers ─────────────────────────────────────────────────────────────
func (h *handler) handleAPITasksList(w http.ResponseWriter, _ *http.Request) {
tasks := globalQueue.snapshot()
writeJSON(w, tasks)
}
func (h *handler) handleAPITasksCancel(w http.ResponseWriter, r *http.Request) {
id := r.PathValue("id")
t, ok := globalQueue.findByID(id)
if !ok {
writeError(w, http.StatusNotFound, "task not found")
return
}
globalQueue.mu.Lock()
defer globalQueue.mu.Unlock()
switch t.Status {
case TaskPending:
t.Status = TaskCancelled
now := time.Now()
t.DoneAt = &now
writeJSON(w, map[string]string{"status": "cancelled"})
case TaskRunning:
if t.job != nil {
t.job.abort()
}
t.Status = TaskCancelled
now := time.Now()
t.DoneAt = &now
writeJSON(w, map[string]string{"status": "cancelled"})
default:
writeError(w, http.StatusConflict, "task is not running or pending")
}
}
func (h *handler) handleAPITasksPriority(w http.ResponseWriter, r *http.Request) {
id := r.PathValue("id")
t, ok := globalQueue.findByID(id)
if !ok {
writeError(w, http.StatusNotFound, "task not found")
return
}
var req struct {
Delta int `json:"delta"`
}
if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
writeError(w, http.StatusBadRequest, "invalid body")
return
}
globalQueue.mu.Lock()
defer globalQueue.mu.Unlock()
if t.Status != TaskPending {
writeError(w, http.StatusConflict, "only pending tasks can be reprioritised")
return
}
t.Priority += req.Delta
writeJSON(w, map[string]int{"priority": t.Priority})
}
func (h *handler) handleAPITasksCancelAll(w http.ResponseWriter, _ *http.Request) {
globalQueue.mu.Lock()
now := time.Now()
n := 0
for _, t := range globalQueue.tasks {
switch t.Status {
case TaskPending:
t.Status = TaskCancelled
t.DoneAt = &now
n++
case TaskRunning:
if t.job != nil {
t.job.abort()
}
t.Status = TaskCancelled
t.DoneAt = &now
n++
}
}
globalQueue.mu.Unlock()
writeJSON(w, map[string]int{"cancelled": n})
}
func (h *handler) handleAPITasksStream(w http.ResponseWriter, r *http.Request) {
id := r.PathValue("id")
// Wait up to 5s for the task to get a job (it may be pending)
deadline := time.Now().Add(5 * time.Second)
var j *jobState
for time.Now().Before(deadline) {
if jj, ok := globalQueue.findJob(id); ok {
j = jj
break
}
time.Sleep(200 * time.Millisecond)
}
if j == nil {
http.Error(w, "task not found or not yet started", http.StatusNotFound)
return
}
streamJob(w, r, j)
}