fix(workers): Fix false memory backoff and add backing-off color coding

- Fix memory calculation to use max-old-space-size (1500MB) instead of
  V8's dynamic heapTotal. This prevents false 95%+ readings when idle.
- Add yellow color for backing-off workers in pod visualization
- Update legend and tooltips with backing-off status
- Remove pool toggle from TasksDashboard (moved to Workers page)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

backend/src/tasks/task-worker.ts

@@ -84,6 +84,20 @@ const MAX_CONCURRENT_TASKS = parseInt(process.env.MAX_CONCURRENT_TASKS || '3');
 // Default 85% - gives headroom before OOM
 const MEMORY_BACKOFF_THRESHOLD = parseFloat(process.env.MEMORY_BACKOFF_THRESHOLD || '0.85');
 
+// Parse max heap size from NODE_OPTIONS (--max-old-space-size=1500)
+// This is used as the denominator for memory percentage calculation
+// V8's heapTotal is dynamic and stays small when idle, causing false high percentages
+function getMaxHeapSizeMb(): number {
+  const nodeOptions = process.env.NODE_OPTIONS || '';
+  const match = nodeOptions.match(/--max-old-space-size=(\d+)/);
+  if (match) {
+    return parseInt(match[1], 10);
+  }
+  // Fallback: use 512MB if not specified
+  return 512;
+}
+const MAX_HEAP_SIZE_MB = getMaxHeapSizeMb();
+
 // When CPU usage exceeds this threshold (as decimal 0.0-1.0), stop claiming new tasks
 // Default 90% - allows some burst capacity
 const CPU_BACKOFF_THRESHOLD = parseFloat(process.env.CPU_BACKOFF_THRESHOLD || '0.90');
@@ -186,12 +200,16 @@ export class TaskWorker {
 
   /**
    * Get current resource usage
+   * Memory percentage is calculated against MAX_HEAP_SIZE_MB (from --max-old-space-size)
+   * NOT against V8's dynamic heapTotal which stays small when idle
    */
   private getResourceStats(): ResourceStats {
     const memUsage = process.memoryUsage();
     const heapUsedMb = memUsage.heapUsed / 1024 / 1024;
-    const heapTotalMb = memUsage.heapTotal / 1024 / 1024;
+    // Use MAX_HEAP_SIZE_MB as ceiling, not dynamic heapTotal
-    const memoryPercent = heapUsedMb / heapTotalMb;
+    // V8's heapTotal stays small when idle (e.g., 36MB) causing false 95%+ readings
+    // With --max-old-space-size=1500, we should calculate against 1500MB
+    const memoryPercent = heapUsedMb / MAX_HEAP_SIZE_MB;
 
     // Calculate CPU usage since last check
     const cpuUsage = process.cpuUsage();
@@ -212,7 +230,7 @@ export class TaskWorker {
     return {
       memoryPercent,
       memoryMb: Math.round(heapUsedMb),
-      memoryTotalMb: Math.round(heapTotalMb),
+      memoryTotalMb: MAX_HEAP_SIZE_MB, // Use max-old-space-size, not dynamic heapTotal
       cpuPercent: Math.min(100, cpuPercent), // Cap at 100%
       isBackingOff: this.isBackingOff,
       backoffReason: this.backoffReason,

cannaiq/src/pages/TasksDashboard.tsx

@@ -16,7 +16,6 @@ import {
   ChevronRight,
   Gauge,
   Users,
-  Play,
   Square,
   Plus,
   X,
@@ -451,7 +450,6 @@ export default function TasksDashboard() {
   const [loading, setLoading] = useState(true);
   const [error, setError] = useState<string | null>(null);
   const [poolPaused, setPoolPaused] = useState(false);
-  const [poolLoading, setPoolLoading] = useState(false);
   const [showCreateModal, setShowCreateModal] = useState(false);
 
   // Pagination
@@ -490,23 +488,6 @@ export default function TasksDashboard() {
     }
   };
 
-  const togglePool = async () => {
-    setPoolLoading(true);
-    try {
-      if (poolPaused) {
-        await api.resumeTaskPool();
-        setPoolPaused(false);
-      } else {
-        await api.pauseTaskPool();
-        setPoolPaused(true);
-      }
-    } catch (err: any) {
-      setError(err.message || 'Failed to toggle pool');
-    } finally {
-      setPoolLoading(false);
-    }
-  };
-
   const handleDeleteTask = async (taskId: number) => {
     if (!confirm('Delete this task?')) return;
     try {
@@ -579,28 +560,13 @@ export default function TasksDashboard() {
                 <Plus className="w-4 h-4" />
                 Create Task
               </button>
-              {/* Pool Toggle */}
+              {/* Pool status indicator */}
-              <button
+              {poolPaused && (
-                onClick={togglePool}
+                <span className="inline-flex items-center gap-1.5 px-3 py-1.5 rounded-full text-sm font-medium bg-yellow-100 text-yellow-800">
-                disabled={poolLoading}
+                  <Square className="w-4 h-4" />
-                className={`flex items-center gap-2 px-4 py-2 rounded-lg font-medium transition-colors ${
+                  Pool Paused
-                  poolPaused
+                </span>
-                    ? 'bg-emerald-100 text-emerald-700 hover:bg-emerald-200'
+              )}
-                    : 'bg-red-100 text-red-700 hover:bg-red-200'
-                }`}
-              >
-                {poolPaused ? (
-                  <>
-                    <Play className={`w-5 h-5 ${poolLoading ? 'animate-pulse' : ''}`} />
-                    Resume Pool
-                  </>
-                ) : (
-                  <>
-                    <Square className={`w-5 h-5 ${poolLoading ? 'animate-pulse' : ''}`} />
-                    Pause Pool
-                  </>
-                )}
-              </button>
               <span className="text-sm text-gray-400">Auto-refreshes every 15s</span>
             </div>
           </div>

cannaiq/src/pages/WorkersDashboard.tsx

@@ -369,8 +369,10 @@ function PodVisualization({
 
           const isBusy = worker.current_task_id !== null;
           const isDecommissioning = worker.decommission_requested;
-          const workerColor = isDecommissioning ? 'bg-orange-500' : isBusy ? 'bg-blue-500' : 'bg-emerald-500';
+          const isBackingOff = worker.metadata?.is_backing_off;
-          const workerBorder = isDecommissioning ? 'border-orange-300' : isBusy ? 'border-blue-300' : 'border-emerald-300';
+          // Color priority: decommissioning > backing off > busy > idle
+          const workerColor = isDecommissioning ? 'bg-orange-500' : isBackingOff ? 'bg-yellow-500' : isBusy ? 'bg-blue-500' : 'bg-emerald-500';
+          const workerBorder = isDecommissioning ? 'border-orange-300' : isBackingOff ? 'border-yellow-300' : isBusy ? 'border-blue-300' : 'border-emerald-300';
 
           // Line from center to worker
           const lineLength = radius - 10;
@@ -381,7 +383,7 @@ function PodVisualization({
             <div key={worker.id}>
               {/* Connection line */}
               <div
-                className={`absolute w-0.5 ${isDecommissioning ? 'bg-orange-300' : isBusy ? 'bg-blue-300' : 'bg-emerald-300'}`}
+                className={`absolute w-0.5 ${isDecommissioning ? 'bg-orange-300' : isBackingOff ? 'bg-yellow-300' : isBusy ? 'bg-blue-300' : 'bg-emerald-300'}`}
                 style={{
                   height: `${lineLength}px`,
                   left: '50%',
@@ -398,7 +400,7 @@ function PodVisualization({
                   top: '50%',
                   transform: `translate(-50%, -50%) translate(${x}px, ${y}px)`,
                 }}
-                title={`${worker.friendly_name}\nStatus: ${isDecommissioning ? 'Stopping after current task' : isBusy ? `Working on task #${worker.current_task_id}` : 'Idle - waiting for tasks'}\nMemory: ${worker.metadata?.memory_mb || 0} MB\nCPU: ${formatCpuTime(worker.metadata?.cpu_user_ms || 0)} user, ${formatCpuTime(worker.metadata?.cpu_system_ms || 0)} sys\nCompleted: ${worker.tasks_completed} | Failed: ${worker.tasks_failed}\nLast heartbeat: ${new Date(worker.last_heartbeat_at).toLocaleTimeString()}`}
+                title={`${worker.friendly_name}\nStatus: ${isDecommissioning ? 'Stopping after current task' : isBackingOff ? `Backing off: ${worker.metadata?.backoff_reason || 'resource pressure'}` : isBusy ? `Working on task #${worker.current_task_id}` : 'Ready - waiting for tasks'}\nMemory: ${worker.metadata?.memory_mb || 0} MB (${worker.metadata?.memory_percent || 0}%)\nCPU: ${formatCpuTime(worker.metadata?.cpu_user_ms || 0)} user, ${formatCpuTime(worker.metadata?.cpu_system_ms || 0)} sys\nCompleted: ${worker.tasks_completed} | Failed: ${worker.tasks_failed}\nLast heartbeat: ${new Date(worker.last_heartbeat_at).toLocaleTimeString()}`}
               >
                 {index + 1}
               </div>
@@ -700,11 +702,11 @@ export function WorkersDashboard() {
                   Worker Pods ({Array.from(groupWorkersByPod(workers)).length} pods, {activeWorkers.length} workers)
                 </h3>
                 <p className="text-xs text-gray-500 mt-0.5">
-                  <span className="inline-flex items-center gap-1"><span className="w-2 h-2 rounded-full bg-emerald-500"></span> idle</span>
+                  <span className="inline-flex items-center gap-1"><span className="w-2 h-2 rounded-full bg-emerald-500"></span> ready</span>
                   <span className="mx-2">|</span>
                   <span className="inline-flex items-center gap-1"><span className="w-2 h-2 rounded-full bg-blue-500"></span> busy</span>
                   <span className="mx-2">|</span>
-                  <span className="inline-flex items-center gap-1"><span className="w-2 h-2 rounded-full bg-yellow-500"></span> mixed</span>
+                  <span className="inline-flex items-center gap-1"><span className="w-2 h-2 rounded-full bg-yellow-500"></span> backing off</span>
                   <span className="mx-2">|</span>
                   <span className="inline-flex items-center gap-1"><span className="w-2 h-2 rounded-full bg-orange-500"></span> stopping</span>
                 </p>

k8s/scraper-worker.yaml

@@ -40,12 +40,16 @@ spec:
               valueFrom:
                 fieldRef:
                   fieldPath: metadata.name
+            - name: API_BASE_URL
+              value: "http://scraper"
+            - name: NODE_OPTIONS
+              value: "--max-old-space-size=1500"
           resources:
             requests:
-              memory: "256Mi"
+              memory: "1Gi"
               cpu: "100m"
             limits:
-              memory: "512Mi"
+              memory: "2Gi"
               cpu: "500m"
           livenessProbe:
             exec: