Scale biz_state collects with dedicated workers and non-blocking UI poll.

Add PG claim/NE mutex, persist pool, and biz_state_worker replicas; fix double-SSH and row-count bugs; stop 32m collectNow while-loop from freezing page switches.

Co-authored-by: Cursor <cursoragent@cursor.com>

scripts/start_netx.ps1

@@ -34,6 +34,7 @@ $errFile = Join-Path $runDir "netx.err.log"
 $workerPidFile = Join-Path $runDir "worker.pid"
 $workerLogFile = Join-Path $runDir "worker.out.log"
 $workerErrFile = Join-Path $runDir "worker.err.log"
+$bizStateWorkerPidDir = Join-Path $runDir "biz_state_workers"
 $webPidFile = Join-Path $runDir "web.pid"
 $webLogFile = Join-Path $runDir "web.out.log"
 $webErrFile = Join-Path $runDir "web.err.log"
@@ -187,6 +188,41 @@ function Start-NetxWorker {
     }
 }
 
+function Start-BizStateWorkers {
+    if ($InlineSchedulers) {
+        return
+    }
+    $replicas = 2
+    if ($env:NETX_BIZ_STATE_WORKER_REPLICAS) {
+        try { $replicas = [int]$env:NETX_BIZ_STATE_WORKER_REPLICAS } catch { $replicas = 2 }
+    }
+    if ($replicas -lt 1) { $replicas = 1 }
+    if (-not (Test-Path $bizStateWorkerPidDir)) {
+        New-Item -ItemType Directory -Path $bizStateWorkerPidDir | Out-Null
+    }
+    Write-Host "==> Starting biz_state workers (replicas=$replicas)"
+    for ($i = 0; $i -lt $replicas; $i++) {
+        $outLog = Reset-LogFile -Path (Join-Path $bizStateWorkerPidDir "worker$i.out.log")
+        $errLog = Reset-LogFile -Path (Join-Path $bizStateWorkerPidDir "worker$i.err.log")
+        $pidPath = Join-Path $bizStateWorkerPidDir "worker$i.pid"
+        $proc = Start-Process -FilePath $pythonExe `
+            -ArgumentList @("-m", "netx_api.biz_state_worker") `
+            -WorkingDirectory $projectRoot `
+            -WindowStyle Hidden `
+            -RedirectStandardOutput $outLog `
+            -RedirectStandardError $errLog `
+            -PassThru
+        Set-Content -Path $pidPath -Value "$($proc.Id)"
+        Write-Host "biz_state_worker[$i] PID=$($proc.Id)"
+        Start-Sleep -Milliseconds 400
+        if ($proc.HasExited) {
+            Write-Host "[ERR] biz_state_worker[$i] exited immediately." -ForegroundColor Red
+            Show-LogTail -Path $errLog
+            exit 1
+        }
+    }
+}
+
 if ($Background) {
     # Truncate logs so a failed start is not confused with an old run.
     Set-Content -Path $logFile -Value "" -Encoding utf8
@@ -226,6 +262,7 @@ if ($Background) {
     }
     Write-Host "==> netx API ready: http://${BindHost}:${Port}/health" -ForegroundColor Green
     Start-NetxWorker
+    Start-BizStateWorkers
     if ($WithWeb) {
         Write-Host "==> Starting Vite dev server in background"
         $webRoot = Join-Path $projectRoot "web"
@@ -283,5 +320,6 @@ if ($WithWeb) {
 }
 
 Start-NetxWorker
+Start-BizStateWorkers
 Write-Host "==> Starting netx API in foreground"
 & $pythonExe -m netx_api.main

scripts/start_netx.sh

@@ -239,6 +239,21 @@ if [[ "${INLINE_SCHEDULERS}" != "1" ]]; then
   echo "PID        = ${WORKER_PID}"
   echo "Log        = ${WORKER_LOG_FILE}"
   echo "Err        = ${WORKER_ERR_FILE}"
+
+  REPLICAS="${NETX_BIZ_STATE_WORKER_REPLICAS:-2}"
+  if [[ "${REPLICAS}" -lt 1 ]]; then REPLICAS=1; fi
+  BIZ_DIR="${RUN_DIR}/biz_state_workers"
+  mkdir -p "${BIZ_DIR}"
+  echo "==> Starting biz_state workers (replicas=${REPLICAS})"
+  for ((i=0; i<REPLICAS; i++)); do
+    (
+      cd "${ROOT_DIR}"
+      nohup "${PYTHON_CMD}" -m netx_api.biz_state_worker \
+        >"${BIZ_DIR}/worker${i}.out.log" 2>"${BIZ_DIR}/worker${i}.err.log" &
+      echo $! > "${BIZ_DIR}/worker${i}.pid"
+    )
+    echo "biz_state_worker[${i}] PID=$(cat "${BIZ_DIR}/worker${i}.pid")"
+  done
 fi
 
 if [[ "${API_ONLY}" != "1" ]]; then

scripts/stop_netx.ps1

@@ -13,6 +13,7 @@ $runDir = Join-Path $PSScriptRoot ".run"
 $pidFile = Join-Path $runDir "netx.pid"
 $workerPidFile = Join-Path $runDir "worker.pid"
 $webPidFile = Join-Path $runDir "web.pid"
+$bizStateWorkerPidDir = Join-Path $runDir "biz_state_workers"
 
 function Stop-OnePid {
     param([int]$ProcId, [string]$Label)
@@ -62,13 +63,19 @@ function Get-ListenPids {
 function Stop-NetxByCommandLine {
     # Orphan workers often have no worker.pid but still hold worker.out.log.
     $hits = @(Get-CimInstance Win32_Process -ErrorAction SilentlyContinue |
-        Where-Object { $_.CommandLine -match 'netx_api\.(main|worker)' })
+        Where-Object { $_.CommandLine -match 'netx_api\.(main|worker|biz_state_worker)' })
     if ($hits.Count -eq 0) {
-        Write-Host "[INFO] No netx_api.main/worker process by command line"
+        Write-Host "[INFO] No netx_api.main/worker/biz_state_worker process by command line"
         return
     }
     foreach ($p in $hits) {
-        $kind = if ($p.CommandLine -match 'netx_api\.worker') { "worker(cmd)" } else { "api(cmd)" }
+        $kind = if ($p.CommandLine -match 'biz_state_worker') {
+            "biz_state_worker(cmd)"
+        } elseif ($p.CommandLine -match 'netx_api\.worker') {
+            "worker(cmd)"
+        } else {
+            "api(cmd)"
+        }
         Stop-OnePid -ProcId ([int]$p.ProcessId) -Label $kind
     }
 }
@@ -99,6 +106,18 @@ if (Test-Path $workerPidFile) {
     Write-Host "[INFO] No worker PID file"
 }
 
+if (Test-Path $bizStateWorkerPidDir) {
+    Get-ChildItem -Path $bizStateWorkerPidDir -Filter "*.pid" -ErrorAction SilentlyContinue | ForEach-Object {
+        $t = (Get-Content -Path $_.FullName -ErrorAction SilentlyContinue | Select-Object -First 1)
+        $id = 0
+        [void][int]::TryParse("$t", [ref]$id)
+        if ($id -gt 0) {
+            Stop-OnePid -ProcId $id -Label "biz_state_worker"
+        }
+        Remove-Item -Path $_.FullName -Force -ErrorAction SilentlyContinue
+    }
+}
+
 if (Test-Path $webPidFile) {
     $webPidText = (Get-Content -Path $webPidFile -ErrorAction SilentlyContinue | Select-Object -First 1)
     $webProcId = 0

scripts/stop_netx.sh

@@ -102,6 +102,16 @@ else
   echo "[INFO] No worker PID file: ${WORKER_PID_FILE}"
 fi
 
+BIZ_DIR="${RUN_DIR}/biz_state_workers"
+if [[ -d "${BIZ_DIR}" ]]; then
+  for f in "${BIZ_DIR}"/*.pid; do
+    [[ -f "${f}" ]] || continue
+    BPID="$(head -n 1 "${f}" | tr -d '[:space:]' || true)"
+    kill_pid "${BPID}" "biz_state_worker"
+    rm -f "${f}" || true
+  done
+fi
+
 if [[ -f "${WEB_PID_FILE}" ]]; then
   WEB_PID="$(head -n 1 "${WEB_PID_FILE}" | tr -d '[:space:]' || true)"
   kill_pid "${WEB_PID}" "web"