Merge pull request #1280 from pikasTech/fix/1278-playwright-memory-red

fix: 增强 Web 哨兵浏览器卡死报红

config/hwlab-node-lanes.yaml

@@ -210,6 +210,14 @@ lanes:
             domEvaluateTimeoutRedWindowMs: 30000
             screenshotTimeoutRedCount: 2
             pageErrorRedCount: 2
+            browserProcessSampleIntervalMs: 1000
+            browserTotalRssRedMb: 800
+            browserProcessRssRedMb: 600
+            browserRssGrowthRedMb: 300
+            browserRssGrowthWindowMs: 30000
+            playwrightResponsivenessRedMs: 5000
+            playwrightResponsivenessTimeoutRedCount: 2
+            cdpMetricsTimeoutRedCount: 2
             uncommandedStateChangeCommandWindowMs: 10000
             scrollJumpCommandWindowMs: 8000
             scrollJumpFromY: 250
@@ -603,6 +611,14 @@ lanes:
             domEvaluateTimeoutRedWindowMs: 30000
             screenshotTimeoutRedCount: 2
             pageErrorRedCount: 2
+            browserProcessSampleIntervalMs: 1000
+            browserTotalRssRedMb: 800
+            browserProcessRssRedMb: 600
+            browserRssGrowthRedMb: 300
+            browserRssGrowthWindowMs: 30000
+            playwrightResponsivenessRedMs: 5000
+            playwrightResponsivenessTimeoutRedCount: 2
+            cdpMetricsTimeoutRedCount: 2
             uncommandedStateChangeCommandWindowMs: 10000
             scrollJumpCommandWindowMs: 8000
             scrollJumpFromY: 250
@@ -926,6 +942,14 @@ lanes:
             domEvaluateTimeoutRedWindowMs: 30000
             screenshotTimeoutRedCount: 2
             pageErrorRedCount: 2
+            browserProcessSampleIntervalMs: 1000
+            browserTotalRssRedMb: 800
+            browserProcessRssRedMb: 600
+            browserRssGrowthRedMb: 300
+            browserRssGrowthWindowMs: 30000
+            playwrightResponsivenessRedMs: 5000
+            playwrightResponsivenessTimeoutRedCount: 2
+            cdpMetricsTimeoutRedCount: 2
             uncommandedStateChangeCommandWindowMs: 10000
             scrollJumpCommandWindowMs: 8000
             scrollJumpFromY: 250

scripts/src/hwlab-node-lanes.ts

@@ -184,6 +184,14 @@ export interface HwlabRuntimeWebProbeAlertThresholdsSpec {
   readonly domEvaluateTimeoutRedWindowMs: number;
   readonly screenshotTimeoutRedCount: number;
   readonly pageErrorRedCount: number;
+  readonly browserProcessSampleIntervalMs: number;
+  readonly browserTotalRssRedMb: number;
+  readonly browserProcessRssRedMb: number;
+  readonly browserRssGrowthRedMb: number;
+  readonly browserRssGrowthWindowMs: number;
+  readonly playwrightResponsivenessRedMs: number;
+  readonly playwrightResponsivenessTimeoutRedCount: number;
+  readonly cdpMetricsTimeoutRedCount: number;
   readonly uncommandedStateChangeCommandWindowMs: number;
   readonly scrollJumpCommandWindowMs: number;
   readonly scrollJumpFromY: number;
@@ -1184,6 +1192,14 @@ function webProbeAlertThresholdsConfig(value: unknown, path: string): HwlabRunti
     domEvaluateTimeoutRedWindowMs: positiveNumberField(raw, "domEvaluateTimeoutRedWindowMs", path),
     screenshotTimeoutRedCount: positiveNumberField(raw, "screenshotTimeoutRedCount", path),
     pageErrorRedCount: positiveNumberField(raw, "pageErrorRedCount", path),
+    browserProcessSampleIntervalMs: positiveNumberField(raw, "browserProcessSampleIntervalMs", path),
+    browserTotalRssRedMb: positiveNumberField(raw, "browserTotalRssRedMb", path),
+    browserProcessRssRedMb: positiveNumberField(raw, "browserProcessRssRedMb", path),
+    browserRssGrowthRedMb: positiveNumberField(raw, "browserRssGrowthRedMb", path),
+    browserRssGrowthWindowMs: positiveNumberField(raw, "browserRssGrowthWindowMs", path),
+    playwrightResponsivenessRedMs: positiveNumberField(raw, "playwrightResponsivenessRedMs", path),
+    playwrightResponsivenessTimeoutRedCount: positiveNumberField(raw, "playwrightResponsivenessTimeoutRedCount", path),
+    cdpMetricsTimeoutRedCount: positiveNumberField(raw, "cdpMetricsTimeoutRedCount", path),
     uncommandedStateChangeCommandWindowMs: positiveNumberField(raw, "uncommandedStateChangeCommandWindowMs", path),
     scrollJumpCommandWindowMs: positiveNumberField(raw, "scrollJumpCommandWindowMs", path),
     scrollJumpFromY: positiveNumberField(raw, "scrollJumpFromY", path),

scripts/src/hwlab-node-web-observe-analyzer-source.ts

@@ -43,6 +43,7 @@ const promptNetworkRows = applyAnalysisFocus(filterRowsByTimeWindow(sourceNetwor
 const consoleEvents = applyAnalysisFocus(filterRowsByTimeWindow(await readJsonl(dataFile("console.jsonl"), { tail: smallJsonlTailLimit }), sampleWindow), analysisFocus);
 const errors = applyAnalysisFocus(filterRowsByTimeWindow(await readJsonl(dataFile("errors.jsonl"), { tail: smallJsonlTailLimit }), sampleWindow), analysisFocus);
 const artifacts = applyAnalysisFocus(filterRowsByTimeWindow(await readJsonl(dataFile("artifacts.jsonl"), { tail: smallJsonlTailLimit }), sampleWindow), analysisFocus);
+const browserProcessRows = applyAnalysisFocus(filterRowsByTimeWindow(await readJsonl(dataFile("browser-process.jsonl"), { tail: smallJsonlTailLimit }), sampleWindow), analysisFocus);
 const manifest = await readJson(path.join(stateDir, "manifest.json"));
 const heartbeat = await readJson(path.join(stateDir, "heartbeat.json"));
 const commandState = await readCommandState(stateDir);
@@ -55,6 +56,7 @@ const pagePerformance = buildPagePerformanceReport(samples, manifest);
 const promptNetwork = buildPromptNetworkReport(control, promptNetworkRows);
 const runtimeAlerts = buildRuntimeAlerts(samples, control, network, consoleEvents, errors);
 const apiDomLag = buildApiDomLagReport(samples, network);
+const browserProcess = buildBrowserProcessReport(browserProcessRows);
 const projectManagement = buildProjectManagementReport(samples, control, network, pagePerformance, projectManagementConfig);
 const runnerErrors = errors.slice(-8).map((item) => {
   const details = item.error?.details && typeof item.error.details === "object" ? item.error.details : {};
@@ -105,9 +107,9 @@ const runnerErrors = errors.slice(-8).map((item) => {
 });
 const commandFailures = summarizeCommandFailures(control);
 const toolFindings = buildToolFindings({ manifest, heartbeat, commandState });
-const findings = [...toolFindings, ...buildProjectManagementFindings(projectManagement), ...buildFindings(samples, control, network, errors, sampleMetrics, promptNetwork, runtimeAlerts, pagePerformance, pageProvenance, commandFailures, manifest, apiDomLag)];
+const findings = [...toolFindings, ...buildProjectManagementFindings(projectManagement), ...buildFindings(samples, control, network, errors, sampleMetrics, promptNetwork, runtimeAlerts, pagePerformance, pageProvenance, commandFailures, manifest, apiDomLag, browserProcess)];
 if (jsonlReadIssues.length > 0) findings.unshift({ id: "jsonl-read-issues", severity: "red", summary: "observer analyzer hit JSONL read/parse issues", count: jsonlReadIssues.length, issues: jsonlReadIssues.slice(0, 20) });
-const recentWindow = buildRecentAnalysisWindow({ samples, control, network, consoleEvents, errors, manifest });
+const recentWindow = buildRecentAnalysisWindow({ samples, control, network, consoleEvents, errors, browserProcessRows, manifest });
 const commandTimeline = control.filter((item) => item.phase === "completed" || item.phase === "failed").map((item) => ({ ts: item.ts, phase: item.phase, commandId: item.commandId, type: item.type, input: item.input, afterUrl: item.afterUrl }));
 const report = {
   ok: findings.filter((item) => item.severity === "red").length === 0,
@@ -118,7 +120,7 @@ const report = {
   alertThresholds,
   manifest: compactManifest(manifest),
   heartbeat: compactHeartbeat(heartbeat),
-  counts: { samples: samples.length, control: control.length, network: network.length, console: consoleEvents.length, errors: errors.length, artifacts: artifacts.length },
+  counts: { samples: samples.length, control: control.length, network: network.length, console: consoleEvents.length, errors: errors.length, artifacts: artifacts.length, browserProcess: browserProcessRows.length },
   commandTimeline,
   transitions,
   sampleMetrics,
@@ -128,6 +130,7 @@ const report = {
   promptNetwork,
   runtimeAlerts,
   apiDomLag,
+  browserProcess,
   runnerErrors,
   commandFailures,
   commandState,
@@ -157,6 +160,7 @@ console.log(JSON.stringify({
     projectManagement: projectManagement.summary,
     runtimeAlerts: runtimeAlerts.summary,
     apiDomLag: apiDomLag.summary,
+    browserProcess: browserProcess.summary,
     findingCount: findings.length,
     redFindingCount: findings.filter((item) => item.severity === "red").length,
     redFindings: prioritizeFindings(findings.filter((item) => item.severity === "red")).slice(0, 12).map((item) => ({ kind: item.id ?? item.kind ?? item.code, severity: item.severity, count: item.count ?? item.sampleCount ?? null, summary: String(item.summary ?? item.message ?? "").slice(0, 180) })),
@@ -177,6 +181,7 @@ console.log(JSON.stringify({
   promptNetwork: recentWindow.promptNetwork.summary,
   runtimeAlerts: recentWindow.runtimeAlerts.summary,
   apiDomLag: compactApiDomLagForOutput(apiDomLag),
+  browserProcess: recentWindow.browserProcess.summary,
   runnerErrors,
   commandFailures: commandFailures.slice(-8),
   commandState,
@@ -561,6 +566,14 @@ function parseAlertThresholds(value) {
     domEvaluateTimeoutRedWindowMs: requiredPositiveThreshold(raw, "domEvaluateTimeoutRedWindowMs"),
     screenshotTimeoutRedCount: requiredPositiveThreshold(raw, "screenshotTimeoutRedCount"),
     pageErrorRedCount: requiredPositiveThreshold(raw, "pageErrorRedCount"),
+    browserProcessSampleIntervalMs: requiredPositiveThreshold(raw, "browserProcessSampleIntervalMs"),
+    browserTotalRssRedMb: requiredPositiveThreshold(raw, "browserTotalRssRedMb"),
+    browserProcessRssRedMb: requiredPositiveThreshold(raw, "browserProcessRssRedMb"),
+    browserRssGrowthRedMb: requiredPositiveThreshold(raw, "browserRssGrowthRedMb"),
+    browserRssGrowthWindowMs: requiredPositiveThreshold(raw, "browserRssGrowthWindowMs"),
+    playwrightResponsivenessRedMs: requiredPositiveThreshold(raw, "playwrightResponsivenessRedMs"),
+    playwrightResponsivenessTimeoutRedCount: requiredPositiveThreshold(raw, "playwrightResponsivenessTimeoutRedCount"),
+    cdpMetricsTimeoutRedCount: requiredPositiveThreshold(raw, "cdpMetricsTimeoutRedCount"),
     uncommandedStateChangeCommandWindowMs: requiredPositiveThreshold(raw, "uncommandedStateChangeCommandWindowMs"),
     scrollJumpCommandWindowMs: requiredPositiveThreshold(raw, "scrollJumpCommandWindowMs"),
     scrollJumpFromY: requiredPositiveThreshold(raw, "scrollJumpFromY"),
@@ -2725,11 +2738,12 @@ function promptCommandHasAuthoritativeSubmitSideEffect(control, promptRound) {
     || Number(sideEffect.messageCountDelta ?? 0) > 0;
 }
 
-function buildFindings(samples, control, network, errors, sampleMetrics, promptNetwork, runtimeAlerts, pagePerformance, pageProvenance, commandFailures = [], manifest = {}, apiDomLag = null) {
+function buildFindings(samples, control, network, errors, sampleMetrics, promptNetwork, runtimeAlerts, pagePerformance, pageProvenance, commandFailures = [], manifest = {}, apiDomLag = null, browserProcess = null) {
   const findings = [];
   const effectiveApiDomLag = apiDomLag || buildApiDomLagReport(samples, network);
   if (commandFailures.length > 0) findings.push({ id: "observer-command-failed", severity: "red", summary: "observer control commands failed; analyze must surface command failure instead of hiding it in command artifacts", count: commandFailures.length, commands: commandFailures.slice(0, 20) });
   findings.push(...buildFrontendFreezeFindings(errors, control));
+  findings.push(...buildBrowserProcessFindings(browserProcess));
   findings.push(...buildControlledNavigationRootCauseFindings(control, manifest));
   findings.push(...buildSessionInvariantFindings(control, manifest));
   const commandTimes = control
@@ -3092,6 +3106,314 @@ function buildFrontendFreezeFindings(errors, control) {
   return findings;
 }
 
+function buildBrowserProcessReport(rows) {
+  const samples = (Array.isArray(rows) ? rows : [])
+    .filter((item) => item && (item.type === "browser-process-sample" || item.process || item.pages))
+    .map((item) => ({ ...item, tsMs: Date.parse(String(item.ts || "")) }))
+    .filter((item) => Number.isFinite(item.tsMs))
+    .sort((a, b) => a.tsMs - b.tsMs);
+  const memorySamples = samples.map((item) => {
+    const process = item.process && typeof item.process === "object" ? item.process : {};
+    const growth = item.growth && typeof item.growth === "object" ? item.growth : {};
+    const totalRssBytes = browserMetricNumber(process.totalRssBytes);
+    const maxProcessRssBytes = browserMetricNumber(process.maxProcessRssBytes);
+    return {
+      ts: item.ts,
+      tsMs: item.tsMs,
+      seq: item.seq ?? null,
+      sampleSeq: item.sampleSeq ?? null,
+      commandId: item.commandId ?? null,
+      processCount: browserMetricNumber(process.chromiumProcessCount),
+      totalRssBytes,
+      totalRssMb: bytesToMb(totalRssBytes),
+      maxProcessRssBytes,
+      maxProcessRssMb: bytesToMb(maxProcessRssBytes),
+      maxProcess: compactBrowserProcessRow(process.maxProcess),
+      totalRssGrowthBytes: browserMetricNumber(growth.totalRssGrowthBytes),
+      totalRssGrowthMb: bytesToMb(growth.totalRssGrowthBytes),
+      maxProcessRssGrowthBytes: browserMetricNumber(growth.maxProcessRssGrowthBytes),
+      maxProcessRssGrowthMb: bytesToMb(growth.maxProcessRssGrowthBytes),
+      roles: process.roles && typeof process.roles === "object" ? process.roles : {},
+      valuesRedacted: true,
+    };
+  });
+  const computedGrowth = computeBrowserProcessGrowth(memorySamples, alertThresholds.browserRssGrowthWindowMs);
+  const pageEvents = [];
+  const cdpTimeoutEvents = [];
+  for (const sample of samples) {
+    for (const page of Array.isArray(sample.pages) ? sample.pages : []) {
+      const responsiveness = page?.responsiveness && typeof page.responsiveness === "object" ? page.responsiveness : {};
+      const cdp = page?.cdp && typeof page.cdp === "object" ? page.cdp : {};
+      const calls = Array.isArray(cdp.calls) ? cdp.calls : [];
+      const event = {
+        ts: sample.ts,
+        tsMs: sample.tsMs,
+        seq: sample.seq ?? null,
+        sampleSeq: sample.sampleSeq ?? null,
+        commandId: sample.commandId ?? null,
+        pageRole: page?.pageRole ?? null,
+        pageId: page?.pageId ?? null,
+        pageEpoch: page?.pageEpoch ?? null,
+        url: safeReportUrl(page?.url),
+        timeoutMs: browserMetricNumber(page?.timeoutMs),
+        responsivenessOk: responsiveness.ok === true,
+        responsivenessTimeout: responsiveness.timeout === true,
+        responsivenessLatencyMs: browserMetricNumber(responsiveness.latencyMs),
+        cdpTimeoutCount: browserMetricNumber(cdp.timeoutCount),
+        cdpErrorCount: browserMetricNumber(cdp.errorCount),
+        heapUsedMb: page?.heapUsage ? bytesToMb(page.heapUsage.usedSize) : null,
+        heapTotalMb: page?.heapUsage ? bytesToMb(page.heapUsage.totalSize) : null,
+        domNodes: page?.domCounters ? browserMetricNumber(page.domCounters.nodes) : null,
+        valuesRedacted: true,
+      };
+      pageEvents.push(event);
+      for (const call of calls) {
+        if (call?.timeout !== true || call?.method === "Runtime.evaluate") continue;
+        cdpTimeoutEvents.push({
+          ...event,
+          method: call.method ?? null,
+          latencyMs: browserMetricNumber(call.latencyMs),
+          errorMessage: limitText(call.error?.message ?? "", 180),
+          valuesRedacted: true,
+        });
+      }
+      const callTimeoutCount = calls.filter((call) => call?.timeout === true && call?.method !== "Runtime.evaluate").length;
+      if (Number(event.cdpTimeoutCount || 0) > callTimeoutCount && calls.length === 0) {
+        cdpTimeoutEvents.push({ ...event, method: "cdp-session", latencyMs: event.responsivenessLatencyMs, errorMessage: "CDP session or metric collection timed out", valuesRedacted: true });
+      }
+    }
+  }
+  const responsivenessEvents = pageEvents.filter((item) => item.responsivenessTimeout === true || Number(item.responsivenessLatencyMs ?? 0) >= alertThresholds.playwrightResponsivenessRedMs);
+  const maxTotal = maxByNumber(memorySamples, (item) => item.totalRssBytes);
+  const maxProcess = maxByNumber(memorySamples, (item) => item.maxProcessRssBytes);
+  const maxGrowth = maxByNumber(computedGrowth, (item) => item.totalRssGrowthBytes);
+  const maxProcessGrowth = maxByNumber(computedGrowth, (item) => item.maxProcessRssGrowthBytes);
+  return {
+    summary: {
+      sampleCount: samples.length,
+      memorySampleCount: memorySamples.length,
+      pageMetricCount: pageEvents.length,
+      cdpMetricsTimeoutCount: cdpTimeoutEvents.length,
+      responsivenessSlowCount: responsivenessEvents.length,
+      responsivenessTimeoutCount: responsivenessEvents.filter((item) => item.responsivenessTimeout === true).length,
+      maxTotalRssMb: maxTotal ? bytesToMb(maxTotal.totalRssBytes) : null,
+      maxProcessRssMb: maxProcess ? bytesToMb(maxProcess.maxProcessRssBytes) : null,
+      maxTotalRssGrowthMb: maxGrowth ? bytesToMb(maxGrowth.totalRssGrowthBytes) : null,
+      maxProcessRssGrowthMb: maxProcessGrowth ? bytesToMb(maxProcessGrowth.maxProcessRssGrowthBytes) : null,
+      totalRssRedMb: alertThresholds.browserTotalRssRedMb,
+      processRssRedMb: alertThresholds.browserProcessRssRedMb,
+      growthRedMb: alertThresholds.browserRssGrowthRedMb,
+      growthWindowMs: alertThresholds.browserRssGrowthWindowMs,
+      responsivenessRedMs: alertThresholds.playwrightResponsivenessRedMs,
+      valuesRedacted: true,
+    },
+    memorySamples: memorySamples.slice(-60),
+    growthSamples: computedGrowth.slice(-60),
+    maxTotalRssSample: maxTotal,
+    maxProcessRssSample: maxProcess,
+    maxGrowthSample: maxGrowth,
+    maxProcessGrowthSample: maxProcessGrowth,
+    responsivenessEvents: responsivenessEvents.slice(-80),
+    cdpTimeoutEvents: cdpTimeoutEvents.slice(-80),
+    latestPageEvents: pageEvents.slice(-20),
+    valuesRedacted: true,
+  };
+}
+
+function buildBrowserProcessFindings(report) {
+  const summary = report?.summary || {};
+  if (!summary || Number(summary.sampleCount ?? 0) <= 0) return [];
+  const findings = [];
+  const maxTotalRssMb = Number(summary.maxTotalRssMb ?? 0);
+  const maxProcessRssMb = Number(summary.maxProcessRssMb ?? 0);
+  if (maxTotalRssMb >= alertThresholds.browserTotalRssRedMb || maxProcessRssMb >= alertThresholds.browserProcessRssRedMb) {
+    findings.push({
+      id: "frontend-browser-memory-rss-red",
+      severity: "red",
+      summary: "Chromium RSS exceeded YAML red threshold during observation; browser memory growth is freeze evidence and must not be hidden by page refresh/fallback",
+      maxTotalRssMb,
+      maxProcessRssMb,
+      totalRssRedMb: alertThresholds.browserTotalRssRedMb,
+      processRssRedMb: alertThresholds.browserProcessRssRedMb,
+      maxTotalRssSample: report.maxTotalRssSample,
+      maxProcessRssSample: report.maxProcessRssSample,
+      rootCause: "frontend_browser_process_memory_pressure",
+      rootCauseStatus: "confirmed-from-runner-process-rss",
+      rootCauseConfidence: "high",
+      fallbackAllowed: false,
+      valuesRedacted: true,
+    });
+  }
+  const maxTotalRssGrowthMb = Number(summary.maxTotalRssGrowthMb ?? 0);
+  const maxProcessRssGrowthMb = Number(summary.maxProcessRssGrowthMb ?? 0);
+  if (maxTotalRssGrowthMb >= alertThresholds.browserRssGrowthRedMb || maxProcessRssGrowthMb >= alertThresholds.browserRssGrowthRedMb) {
+    findings.push({
+      id: "frontend-browser-memory-growth-red",
+      severity: "red",
+      summary: "Chromium RSS grew beyond YAML window budget; this matches the reported freeze pattern of browser memory rapidly climbing",
+      maxTotalRssGrowthMb,
+      maxProcessRssGrowthMb,
+      growthRedMb: alertThresholds.browserRssGrowthRedMb,
+      windowMs: alertThresholds.browserRssGrowthWindowMs,
+      maxGrowthSample: report.maxGrowthSample,
+      maxProcessGrowthSample: report.maxProcessGrowthSample,
+      rootCause: "frontend_browser_process_memory_leak_or_unbounded_render_growth",
+      rootCauseStatus: "confirmed-from-runner-process-rss-growth",
+      rootCauseConfidence: "high",
+      fallbackAllowed: false,
+      valuesRedacted: true,
+    });
+  }
+  const responsivenessEvents = Array.isArray(report.responsivenessEvents) ? report.responsivenessEvents : [];
+  const severeLatencyEvents = responsivenessEvents.filter((item) => item.responsivenessTimeout !== true && Number(item.responsivenessLatencyMs ?? 0) >= alertThresholds.playwrightResponsivenessRedMs);
+  const responsivenessTimeoutBurst = firstBurst(
+    responsivenessEvents.filter((item) => item.responsivenessTimeout === true),
+    alertThresholds.playwrightResponsivenessTimeoutRedCount,
+    alertThresholds.domEvaluateTimeoutRedWindowMs,
+  );
+  if (severeLatencyEvents.length > 0 || responsivenessTimeoutBurst) {
+    const events = severeLatencyEvents.length > 0 ? severeLatencyEvents : responsivenessTimeoutBurst;
+    findings.push({
+      id: "frontend-playwright-responsiveness-red",
+      severity: "red",
+      summary: "Playwright/CDP responsiveness probe exceeded YAML budget; treat the browser page as frozen instead of refreshing or falling back",
+      count: events.length,
+      latencyRedMs: alertThresholds.playwrightResponsivenessRedMs,
+      timeoutThresholdCount: alertThresholds.playwrightResponsivenessTimeoutRedCount,
+      windowMs: alertThresholds.domEvaluateTimeoutRedWindowMs,
+      events: events.slice(0, 20).map(browserProcessEventRef),
+      rootCause: "frontend_browser_page_unresponsive_to_playwright",
+      rootCauseStatus: "confirmed-from-cdp-runtime-evaluate",
+      rootCauseConfidence: "high",
+      fallbackAllowed: false,
+      valuesRedacted: true,
+    });
+  }
+  const cdpTimeoutEvents = Array.isArray(report.cdpTimeoutEvents) ? report.cdpTimeoutEvents : [];
+  const cdpTimeoutBurst = firstBurst(cdpTimeoutEvents, alertThresholds.cdpMetricsTimeoutRedCount, alertThresholds.domEvaluateTimeoutRedWindowMs);
+  if (cdpTimeoutBurst) {
+    findings.push({
+      id: "frontend-cdp-metrics-timeout-red",
+      severity: "red",
+      summary: "CDP browser metrics timed out repeatedly; the sentinel has direct browser-side evidence of a hung page/runtime",
+      count: cdpTimeoutBurst.length,
+      thresholdCount: alertThresholds.cdpMetricsTimeoutRedCount,
+      windowMs: alertThresholds.domEvaluateTimeoutRedWindowMs,
+      events: cdpTimeoutBurst.slice(0, 20).map(browserProcessEventRef),
+      rootCause: "frontend_browser_cdp_metrics_unresponsive",
+      rootCauseStatus: "confirmed-from-cdp-metrics-timeouts",
+      rootCauseConfidence: "high",
+      fallbackAllowed: false,
+      valuesRedacted: true,
+    });
+  }
+  return findings;
+}
+
+function computeBrowserProcessGrowth(samples, windowMs) {
+  const budgetMs = Math.max(1000, Number(windowMs || 0));
+  const sorted = (samples || []).filter((item) => Number.isFinite(item.tsMs)).sort((a, b) => a.tsMs - b.tsMs);
+  return sorted.map((item, index) => {
+    const candidates = sorted.slice(0, index + 1).filter((candidate) => item.tsMs - candidate.tsMs <= budgetMs);
+    const totalBaseline = minByNumber(candidates, (candidate) => candidate.totalRssBytes);
+    const processBaseline = minByNumber(candidates, (candidate) => candidate.maxProcessRssBytes);
+    const totalGrowth = Number(item.totalRssBytes || 0) - Number(totalBaseline?.totalRssBytes || 0);
+    const processGrowth = Number(item.maxProcessRssBytes || 0) - Number(processBaseline?.maxProcessRssBytes || 0);
+    return {
+      ...item,
+      windowMs: budgetMs,
+      totalBaselineAt: totalBaseline?.ts ?? null,
+      processBaselineAt: processBaseline?.ts ?? null,
+      totalRssGrowthBytes: totalGrowth,
+      totalRssGrowthMb: bytesToMb(totalGrowth),
+      maxProcessRssGrowthBytes: processGrowth,
+      maxProcessRssGrowthMb: bytesToMb(processGrowth),
+      valuesRedacted: true,
+    };
+  });
+}
+
+function browserProcessEventRef(item) {
+  return {
+    ts: item?.ts ?? null,
+    seq: item?.seq ?? null,
+    sampleSeq: item?.sampleSeq ?? null,
+    commandId: item?.commandId ?? null,
+    pageRole: item?.pageRole ?? null,
+    pageId: item?.pageId ?? null,
+    pageEpoch: item?.pageEpoch ?? null,
+    timeoutMs: item?.timeoutMs ?? null,
+    responsivenessLatencyMs: item?.responsivenessLatencyMs ?? item?.latencyMs ?? null,
+    responsivenessTimeout: item?.responsivenessTimeout === true,
+    cdpTimeoutCount: item?.cdpTimeoutCount ?? null,
+    method: item?.method ?? null,
+    errorMessage: item?.errorMessage ?? null,
+    valuesRedacted: true,
+  };
+}
+
+function compactBrowserProcessRow(item) {
+  if (!item || typeof item !== "object") return null;
+  return {
+    pid: item.pid ?? null,
+    role: item.role ?? null,
+    name: item.name ?? null,
+    rssMb: bytesToMb(item.rssBytes),
+    vmSizeMb: bytesToMb(item.vmSizeBytes),
+    commandHash: item.commandHash ?? null,
+    valuesRedacted: true,
+  };
+}
+
+function browserMetricNumber(value) {
+  const numeric = Number(value);
+  return Number.isFinite(numeric) ? numeric : null;
+}
+
+function bytesToMb(value) {
+  const numeric = Number(value);
+  if (!Number.isFinite(numeric)) return null;
+  return Number((numeric / 1024 / 1024).toFixed(1));
+}
+
+function maxByNumber(items, getter) {
+  let selected = null;
+  let selectedValue = Number.NEGATIVE_INFINITY;
+  for (const item of Array.isArray(items) ? items : []) {
+    const value = Number(getter(item));
+    if (!Number.isFinite(value)) continue;
+    if (!selected || value > selectedValue) {
+      selected = item;
+      selectedValue = value;
+    }
+  }
+  return selected;
+}
+
+function minByNumber(items, getter) {
+  let selected = null;
+  let selectedValue = Number.POSITIVE_INFINITY;
+  for (const item of Array.isArray(items) ? items : []) {
+    const value = Number(getter(item));
+    if (!Number.isFinite(value)) continue;
+    if (!selected || value < selectedValue) {
+      selected = item;
+      selectedValue = value;
+    }
+  }
+  return selected;
+}
+
+function safeReportUrl(value) {
+  if (!value) return null;
+  try {
+    const url = new URL(String(value));
+    return url.origin + url.pathname;
+  } catch {
+    return limitText(String(value), 200);
+  }
+}
+
 function frontendFreezeErrorEvent(item, promptTimes) {
   const details = objectValue(item?.error?.details);
   const message = String(item?.error?.message ?? item?.message ?? item?.error ?? "");
@@ -3220,7 +3542,7 @@ function maxPresentNumber(values) {
   return numbers.length > 0 ? Math.max(...numbers) : null;
 }
 
-function buildRecentAnalysisWindow({ samples, control, network, consoleEvents, errors, manifest }) {
+function buildRecentAnalysisWindow({ samples, control, network, consoleEvents, errors, browserProcessRows, manifest }) {
   const latestSampleMs = latestTimestampMs(samples);
   const windowMs = 5 * 60 * 1000;
   const fromMs = Number.isFinite(latestSampleMs) ? latestSampleMs - windowMs : Number.NEGATIVE_INFINITY;
@@ -3234,13 +3556,15 @@ function buildRecentAnalysisWindow({ samples, control, network, consoleEvents, e
   const windowNetwork = network.filter(inWindow);
   const windowConsole = consoleEvents.filter(inWindow);
   const windowErrors = errors.filter(inWindow);
+  const windowBrowserProcessRows = (browserProcessRows || []).filter(inWindow);
   const sampleMetrics = buildSampleMetrics(windowSamples, control);
   const pageProvenance = buildPageProvenanceReport(windowSamples, windowControl, manifest);
   const pagePerformance = buildPagePerformanceReport(windowSamples, manifest);
   const promptNetwork = buildPromptNetworkReport(windowControl, windowNetwork);
   const runtimeAlerts = buildRuntimeAlerts(windowSamples, control, windowNetwork, windowConsole, windowErrors);
   const apiDomLag = buildApiDomLagReport(windowSamples, windowNetwork);
-  const findings = buildFindings(windowSamples, control, windowNetwork, windowErrors, sampleMetrics, promptNetwork, runtimeAlerts, pagePerformance, pageProvenance, [], {}, apiDomLag);
+  const browserProcess = buildBrowserProcessReport(windowBrowserProcessRows);
+  const findings = buildFindings(windowSamples, control, windowNetwork, windowErrors, sampleMetrics, promptNetwork, runtimeAlerts, pagePerformance, pageProvenance, [], {}, apiDomLag, browserProcess);
   return {
     summary: {
       name: "recent-5m",
@@ -3252,6 +3576,7 @@ function buildRecentAnalysisWindow({ samples, control, network, consoleEvents, e
       network: windowNetwork.length,
       console: windowConsole.length,
       errors: windowErrors.length,
+      browserProcess: windowBrowserProcessRows.length,
       valuesRedacted: true
     },
     sampleMetrics,
@@ -3260,6 +3585,7 @@ function buildRecentAnalysisWindow({ samples, control, network, consoleEvents, e
     promptNetwork,
     runtimeAlerts,
     apiDomLag,
+    browserProcess,
     findings,
     valuesRedacted: true
   };

scripts/src/hwlab-node-web-observe-analyzer-timing-source.ts

@@ -2227,6 +2227,14 @@ function renderMarkdown(report) {
   const alertSummary = report.runtimeAlerts?.summary || {};
   const apiDomLag = report.apiDomLag || {};
   const apiDomLagSummary = apiDomLag.summary || {};
+  const browserProcess = report.browserProcess || {};
+  const browserProcessSummary = browserProcess.summary || {};
+  const browserProcessLines = Array.isArray(browserProcess.memorySamples) && browserProcess.memorySamples.length > 0
+    ? browserProcess.memorySamples.slice(-40).map((item) => "- " + (item.ts || "-") + " seq=" + (item.seq ?? "-") + " totalRssMb=" + (item.totalRssMb ?? "-") + " maxProcessRssMb=" + (item.maxProcessRssMb ?? "-") + " totalGrowthMb=" + (item.totalRssGrowthMb ?? "-") + " processGrowthMb=" + (item.maxProcessRssGrowthMb ?? "-") + " processes=" + (item.processCount ?? "-")).join("\n")
+    : "- 无 browser-process RSS 样本。";
+  const browserResponsivenessLines = Array.isArray(browserProcess.responsivenessEvents) && browserProcess.responsivenessEvents.length > 0
+    ? browserProcess.responsivenessEvents.slice(-40).map((item) => "- " + (item.ts || "-") + " role=" + (item.pageRole || "-") + " pageId=" + (item.pageId || "-") + " latencyMs=" + (item.responsivenessLatencyMs ?? "-") + " timeout=" + String(item.responsivenessTimeout === true) + " cdpTimeouts=" + (item.cdpTimeoutCount ?? 0)).join("\n")
+    : "- 无 Playwright/CDP 响应性超阈值样本。";
   const projectManagement = report.projectManagement || {};
   const projectSummary = projectManagement.summary || {};
   const projectCommandLines = Array.isArray(projectManagement.commands) && projectManagement.commands.length > 0
@@ -2336,6 +2344,16 @@ function renderMarkdown(report) {
     + "- console: " + (report.counts.console ?? 0) + "\n"
     + "- errors: " + report.counts.errors + "\n\n"
     + "## Findings\n\n" + findingLines + "\n\n"
+    + "## Browser process\n\n"
+    + "- browserProcessSamples: " + (browserProcessSummary.sampleCount ?? 0) + "\n"
+    + "- pageMetricCount: " + (browserProcessSummary.pageMetricCount ?? 0) + "\n"
+    + "- maxTotalRssMb: " + (browserProcessSummary.maxTotalRssMb ?? "-") + " / red=" + (browserProcessSummary.totalRssRedMb ?? "-") + "\n"
+    + "- maxProcessRssMb: " + (browserProcessSummary.maxProcessRssMb ?? "-") + " / red=" + (browserProcessSummary.processRssRedMb ?? "-") + "\n"
+    + "- maxTotalRssGrowthMb: " + (browserProcessSummary.maxTotalRssGrowthMb ?? "-") + " / red=" + (browserProcessSummary.growthRedMb ?? "-") + " windowMs=" + (browserProcessSummary.growthWindowMs ?? "-") + "\n"
+    + "- responsivenessSlowCount: " + (browserProcessSummary.responsivenessSlowCount ?? 0) + " timeoutCount=" + (browserProcessSummary.responsivenessTimeoutCount ?? 0) + " redMs=" + (browserProcessSummary.responsivenessRedMs ?? "-") + "\n"
+    + "- cdpMetricsTimeoutCount: " + (browserProcessSummary.cdpMetricsTimeoutCount ?? 0) + "\n\n"
+    + "### Browser RSS samples\n\n" + browserProcessLines + "\n\n"
+    + "### Browser responsiveness samples\n\n" + browserResponsivenessLines + "\n\n"
     + "## Project management\n\n"
     + "- enabled: " + String(projectSummary.enabled === true) + "\n"
     + "- projectSampleCount: " + (projectSummary.projectSampleCount ?? 0) + "\n"

scripts/src/hwlab-node-web-observe-runner-source.ts

@@ -55,6 +55,7 @@ const files = {
   console: path.join(stateDir, "console.jsonl"),
   errors: path.join(stateDir, "errors.jsonl"),
   artifacts: path.join(stateDir, "artifacts.jsonl"),
+  browserProcess: path.join(stateDir, "browser-process.jsonl"),
 };
 
 let browser;
@@ -76,6 +77,9 @@ let controlPageEpoch = 0;
 let observerPageEpoch = 0;
 let currentPageProvenance = null;
 let heartbeatPulseTimer = null;
+let browserProcessMonitorStop = null;
+let browserProcessMonitorSeq = 0;
+const browserProcessHistory = [];
 const jsonlRotation = { stamp: compactFileTimestamp(startedAt), files: [] };
 
 try {
@@ -107,6 +111,7 @@ try {
       }
       return result ?? { ok: false, reason: "observer-not-ready", pageRole: "observer", pageId: observerPageId, valuesRedacted: true };
     });
+  browserProcessMonitorStop = startBrowserProcessMonitor();
   terminalStatus = "running";
   await writeManifest({ status: "running", auth: publicAuth(auth) });
   await writeHeartbeat({ status: "running" });
@@ -130,6 +135,7 @@ try {
   await writeManifest({ status: "failed", error: errorSummary(error) }).catch(() => {});
   process.exitCode = 2;
 } finally {
+  if (browserProcessMonitorStop) browserProcessMonitorStop();
   if (heartbeatPulseTimer) clearInterval(heartbeatPulseTimer);
   if (browser) await browser.close().catch(() => {});
 }
@@ -236,6 +242,339 @@ function startHeartbeatPulse() {
   return timer;
 }
 
+function startBrowserProcessMonitor() {
+  const intervalMs = Math.max(250, Math.floor(Number(alertThresholds.browserProcessSampleIntervalMs) || 1000));
+  let stopped = false;
+  let running = false;
+  const tick = async (reason) => {
+    if (stopped || running) return;
+    running = true;
+    try {
+      await collectBrowserProcessSample(reason || "interval");
+    } catch (error) {
+      await appendJsonl(files.errors, eventRecord("browser-process-monitor-error", { error: errorSummary(error), valuesRedacted: true })).catch(() => {});
+    } finally {
+      running = false;
+    }
+  };
+  void tick("startup");
+  const timer = setInterval(() => {
+    void tick("interval");
+  }, intervalMs);
+  if (timer && typeof timer.unref === "function") timer.unref();
+  return () => {
+    stopped = true;
+    clearInterval(timer);
+  };
+}
+
+async function collectBrowserProcessSample(reason) {
+  browserProcessMonitorSeq += 1;
+  const tsMs = Date.now();
+  const browserPid = browserProcessPid();
+  const processSummary = await collectChromiumProcessSummary(browserPid);
+  const growth = updateBrowserProcessHistory(tsMs, processSummary);
+  const pages = [];
+  if (page && !page.isClosed()) {
+    pages.push(await collectBrowserPageRuntimeMetrics(page, { pageRole: "control", targetPageId: pageId, pageEpoch: controlPageEpoch }).catch((error) => browserPageRuntimeMetricError("control", pageId, controlPageEpoch, error)));
+  }
+  if (observerPage && !observerPage.isClosed()) {
+    pages.push(await collectBrowserPageRuntimeMetrics(observerPage, { pageRole: "observer", targetPageId: observerPageId, pageEpoch: observerPageEpoch }).catch((error) => browserPageRuntimeMetricError("observer", observerPageId, observerPageEpoch, error)));
+  }
+  await appendJsonl(files.browserProcess, eventRecord("browser-process-sample", {
+    seq: browserProcessMonitorSeq,
+    reason,
+    monitorIntervalMs: Math.max(250, Math.floor(Number(alertThresholds.browserProcessSampleIntervalMs) || 1000)),
+    browserPid,
+    process: processSummary,
+    growth,
+    pages,
+    valuesRedacted: true,
+  }));
+}
+
+function browserProcessPid() {
+  try {
+    const childProcess = browser && typeof browser.process === "function" ? browser.process() : null;
+    const pid = Number(childProcess?.pid);
+    return Number.isFinite(pid) && pid > 0 ? Math.floor(pid) : null;
+  } catch {
+    return null;
+  }
+}
+
+async function collectChromiumProcessSummary(browserPid) {
+  const all = await readProcProcessTable();
+  const childrenByPpid = new Map();
+  for (const item of all) {
+    if (!Number.isFinite(item.ppid)) continue;
+    const list = childrenByPpid.get(item.ppid) || [];
+    list.push(item.pid);
+    childrenByPpid.set(item.ppid, list);
+  }
+  const roots = [process.pid, browserPid].filter((pid, index, items) => Number.isFinite(Number(pid)) && Number(pid) > 0 && items.indexOf(pid) === index).map((pid) => Number(pid));
+  const descendants = new Set(roots);
+  const queue = roots.slice();
+  while (queue.length > 0) {
+    const pid = queue.shift();
+    for (const child of childrenByPpid.get(pid) || []) {
+      if (descendants.has(child)) continue;
+      descendants.add(child);
+      queue.push(child);
+    }
+  }
+  const processes = all
+    .filter((item) => descendants.has(item.pid))
+    .map((item) => ({ ...item, role: classifyChromiumProcess(item) }))
+    .filter((item) => item.role)
+    .map((item) => ({
+      pid: item.pid,
+      ppid: item.ppid,
+      name: item.name || null,
+      role: item.role,
+      rssBytes: item.rssBytes,
+      vmSizeBytes: item.vmSizeBytes,
+      commandHash: item.cmdline ? sha256Text(item.cmdline) : null,
+      commandPreview: redactProcessCommandPreview(item.cmdline),
+      valuesRedacted: true,
+    }))
+    .sort((a, b) => Number(b.rssBytes || 0) - Number(a.rssBytes || 0));
+  const roles = {};
+  for (const item of processes) {
+    const role = item.role || "unknown";
+    const summary = roles[role] || { count: 0, rssBytes: 0, maxRssBytes: 0 };
+    summary.count += 1;
+    summary.rssBytes += Number(item.rssBytes || 0);
+    summary.maxRssBytes = Math.max(summary.maxRssBytes, Number(item.rssBytes || 0));
+    roles[role] = summary;
+  }
+  const totalRssBytes = processes.reduce((sum, item) => sum + Number(item.rssBytes || 0), 0);
+  const maxProcess = processes[0] || null;
+  return {
+    sampledRootPids: roots,
+    chromiumProcessCount: processes.length,
+    totalRssBytes,
+    totalRssMb: bytesToMb(totalRssBytes),
+    maxProcessRssBytes: maxProcess ? Number(maxProcess.rssBytes || 0) : 0,
+    maxProcessRssMb: maxProcess ? bytesToMb(maxProcess.rssBytes) : 0,
+    maxProcess,
+    roles,
+    processes: processes.slice(0, 20),
+    valuesRedacted: true,
+  };
+}
+
+async function readProcProcessTable() {
+  let entries = [];
+  try {
+    entries = await readdir("/proc");
+  } catch {
+    return [];
+  }
+  const numeric = entries.map((name) => Number(name)).filter((pid) => Number.isInteger(pid) && pid > 0);
+  const rows = await Promise.all(numeric.map((pid) => readOneProcProcess(pid).catch(() => null)));
+  return rows.filter(Boolean);
+}
+
+async function readOneProcProcess(pid) {
+  const [statText, statusText, cmdlineText] = await Promise.all([
+    readFile(path.join("/proc", String(pid), "stat"), "utf8").catch(() => ""),
+    readFile(path.join("/proc", String(pid), "status"), "utf8").catch(() => ""),
+    readFile(path.join("/proc", String(pid), "cmdline"), "utf8").catch(() => ""),
+  ]);
+  if (!statText && !statusText) return null;
+  const ppid = procPpidFromStat(statText);
+  const name = procStatusField(statusText, "Name") || null;
+  const rssKb = procStatusKb(statusText, "VmRSS");
+  const vmSizeKb = procStatusKb(statusText, "VmSize");
+  return {
+    pid,
+    ppid,
+    name,
+    cmdline: String(cmdlineText || "").replace(/\0/gu, " ").replace(/\s+/gu, " ").trim(),
+    rssBytes: Number.isFinite(rssKb) ? rssKb * 1024 : 0,
+    vmSizeBytes: Number.isFinite(vmSizeKb) ? vmSizeKb * 1024 : 0,
+  };
+}
+
+function procPpidFromStat(value) {
+  const text = String(value || "");
+  const end = text.lastIndexOf(")");
+  if (end < 0) return null;
+  const parts = text.slice(end + 1).trim().split(/\s+/u);
+  const ppid = Number(parts[1]);
+  return Number.isFinite(ppid) ? ppid : null;
+}
+
+function procStatusField(value, key) {
+  const prefix = String(key || "") + ":";
+  for (const line of String(value || "").split(/\n/u)) {
+    if (line.startsWith(prefix)) return line.slice(prefix.length).trim();
+  }
+  return null;
+}
+
+function procStatusKb(value, key) {
+  const raw = procStatusField(value, key);
+  const match = String(raw || "").match(/^(\d+)\s+kB$/u);
+  return match ? Number(match[1]) : null;
+}
+
+function classifyChromiumProcess(item) {
+  const cmdline = String(item?.cmdline || "");
+  const combined = (cmdline + " " + String(item?.name || "")).toLowerCase();
+  if (!/(?:chromium|chrome|headless_shell)/u.test(combined)) return null;
+  if (/--type=renderer\b/u.test(cmdline)) return "renderer";
+  if (/--type=gpu-process\b/u.test(cmdline)) return "gpu";
+  if (/--type=utility\b/u.test(cmdline)) return "utility";
+  if (/--type=zygote\b/u.test(cmdline)) return "zygote";
+  if (/--type=broker\b/u.test(cmdline)) return "broker";
+  if (/--type=/u.test(cmdline)) return "other";
+  return "browser";
+}
+
+function redactProcessCommandPreview(value) {
+  const text = String(value || "");
+  if (!text) return null;
+  const parts = text.split(/\s+/u).slice(0, 18).map((part) => {
+    if (/--(?:proxy|token|secret|password|cookie|auth|key)[^=]*=/iu.test(part)) return part.replace(/=.*/u, "=[redacted]");
+    return part.replace(/:\/\/[^/@\s]+@/u, "://[redacted]@");
+  });
+  return truncate(parts.join(" "), 260);
+}
+
+function updateBrowserProcessHistory(tsMs, processSummary) {
+  const windowMs = Math.max(1000, Number(alertThresholds.browserRssGrowthWindowMs) || 30000);
+  const totalRssBytes = Number(processSummary?.totalRssBytes || 0);
+  const maxProcessRssBytes = Number(processSummary?.maxProcessRssBytes || 0);
+  browserProcessHistory.push({ tsMs, totalRssBytes, maxProcessRssBytes });
+  const retentionMs = Math.max(windowMs * 3, 180000);
+  while (browserProcessHistory.length > 0 && tsMs - browserProcessHistory[0].tsMs > retentionMs) browserProcessHistory.shift();
+  const windowStartMs = tsMs - windowMs;
+  const candidates = browserProcessHistory.filter((item) => item.tsMs >= windowStartMs && item.tsMs <= tsMs);
+  const baseline = candidates[0] || browserProcessHistory[0] || null;
+  const totalRssGrowthBytes = baseline ? totalRssBytes - Number(baseline.totalRssBytes || 0) : 0;
+  const maxProcessRssGrowthBytes = baseline ? maxProcessRssBytes - Number(baseline.maxProcessRssBytes || 0) : 0;
+  return {
+    windowMs,
+    baselineAt: baseline ? new Date(baseline.tsMs).toISOString() : null,
+    baselineTotalRssBytes: baseline ? baseline.totalRssBytes : null,
+    baselineMaxProcessRssBytes: baseline ? baseline.maxProcessRssBytes : null,
+    totalRssGrowthBytes,
+    totalRssGrowthMb: bytesToMb(totalRssGrowthBytes),
+    maxProcessRssGrowthBytes,
+    maxProcessRssGrowthMb: bytesToMb(maxProcessRssGrowthBytes),
+    valuesRedacted: true,
+  };
+}
+
+async function collectBrowserPageRuntimeMetrics(targetPage, { pageRole, targetPageId, pageEpoch }) {
+  const timeoutMs = Math.max(1000, Math.floor(Number(alertThresholds.playwrightResponsivenessRedMs) || 5000));
+  const startedAtMs = Date.now();
+  let session = null;
+  const result = {
+    pageRole,
+    pageId: targetPageId,
+    pageEpoch: Number.isFinite(Number(pageEpoch)) ? Number(pageEpoch) : 0,
+    url: pageUrl(targetPage),
+    timeoutMs,
+    responsiveness: null,
+    cdp: { timeoutCount: 0, errorCount: 0, calls: [] },
+    valuesRedacted: true,
+  };
+  try {
+    session = await withHardTimeout(targetPage.context().newCDPSession(targetPage), timeoutMs, "newCDPSession exceeded " + timeoutMs + "ms");
+    const enable = await timedCdpSend(session, "Performance.enable", {}, timeoutMs);
+    if (enable.ok !== true) result.cdp.calls.push({ method: "Performance.enable", ...enable });
+    result.responsiveness = await timedCdpSend(session, "Runtime.evaluate", { expression: "1", returnByValue: true }, timeoutMs);
+    result.cdp.calls.push({ method: "Runtime.evaluate", ...result.responsiveness });
+    const performance = await timedCdpSend(session, "Performance.getMetrics", {}, timeoutMs);
+    result.cdp.calls.push({ method: "Performance.getMetrics", ...performance });
+    result.performance = performance.value;
+    const heap = await timedCdpSend(session, "Runtime.getHeapUsage", {}, timeoutMs);
+    result.cdp.calls.push({ method: "Runtime.getHeapUsage", ...heap });
+    result.heapUsage = heap.value;
+    const domCounters = await timedCdpSend(session, "Memory.getDOMCounters", {}, timeoutMs);
+    result.cdp.calls.push({ method: "Memory.getDOMCounters", ...domCounters });
+    result.domCounters = domCounters.value;
+    result.cdp.timeoutCount = result.cdp.calls.filter((item) => item.timeout === true).length;
+    result.cdp.errorCount = result.cdp.calls.filter((item) => item.ok !== true).length;
+  } catch (error) {
+    result.sessionError = errorSummary(error);
+    result.cdp.timeoutCount = isTimeoutErrorMessage(error?.message) ? 1 : 0;
+    result.cdp.errorCount = 1;
+  } finally {
+    result.latencyMs = Date.now() - startedAtMs;
+    if (session) await withHardTimeout(session.detach(), 1000, "CDP session detach exceeded 1000ms").catch(() => {});
+  }
+  return result;
+}
+
+function browserPageRuntimeMetricError(pageRole, targetPageId, pageEpoch, error) {
+  return {
+    pageRole,
+    pageId: targetPageId,
+    pageEpoch: Number.isFinite(Number(pageEpoch)) ? Number(pageEpoch) : 0,
+    url: null,
+    timeoutMs: Math.max(1000, Math.floor(Number(alertThresholds.playwrightResponsivenessRedMs) || 5000)),
+    responsiveness: { ok: false, timeout: isTimeoutErrorMessage(error?.message), error: errorSummary(error), valuesRedacted: true },
+    cdp: { timeoutCount: isTimeoutErrorMessage(error?.message) ? 1 : 0, errorCount: 1, calls: [] },
+    valuesRedacted: true,
+  };
+}
+
+async function timedCdpSend(session, method, params, timeoutMs) {
+  const startedAtMs = Date.now();
+  try {
+    const value = await withHardTimeout(session.send(method, params || {}), timeoutMs, method + " exceeded " + timeoutMs + "ms");
+    return { ok: true, timeout: false, latencyMs: Date.now() - startedAtMs, value: compactCdpValue(method, value), valuesRedacted: true };
+  } catch (error) {
+    return { ok: false, timeout: isTimeoutErrorMessage(error?.message), latencyMs: Date.now() - startedAtMs, error: errorSummary(error), valuesRedacted: true };
+  }
+}
+
+function compactCdpValue(method, value) {
+  if (!value || typeof value !== "object") return value ?? null;
+  if (method === "Performance.getMetrics") {
+    const wanted = new Set(["Timestamp", "Documents", "Frames", "JSEventListeners", "Nodes", "LayoutCount", "RecalcStyleCount", "LayoutDuration", "RecalcStyleDuration", "ScriptDuration", "TaskDuration", "JSHeapUsedSize", "JSHeapTotalSize"]);
+    const metrics = {};
+    for (const item of Array.isArray(value.metrics) ? value.metrics : []) {
+      if (wanted.has(item?.name)) metrics[item.name] = Number.isFinite(Number(item?.value)) ? Number(item.value) : null;
+    }
+    return { metricCount: Array.isArray(value.metrics) ? value.metrics.length : 0, metrics, valuesRedacted: true };
+  }
+  if (method === "Runtime.getHeapUsage") {
+    return {
+      usedSize: Number.isFinite(Number(value.usedSize)) ? Number(value.usedSize) : null,
+      totalSize: Number.isFinite(Number(value.totalSize)) ? Number(value.totalSize) : null,
+      embedderHeapUsedSize: Number.isFinite(Number(value.embedderHeapUsedSize)) ? Number(value.embedderHeapUsedSize) : null,
+      valuesRedacted: true,
+    };
+  }
+  if (method === "Memory.getDOMCounters") {
+    return {
+      documents: Number.isFinite(Number(value.documents)) ? Number(value.documents) : null,
+      nodes: Number.isFinite(Number(value.nodes)) ? Number(value.nodes) : null,
+      jsEventListeners: Number.isFinite(Number(value.jsEventListeners)) ? Number(value.jsEventListeners) : null,
+      valuesRedacted: true,
+    };
+  }
+  if (method === "Runtime.evaluate") {
+    return { resultType: value.result?.type ?? null, unserializableValue: value.result?.unserializableValue ?? null, exceptionDetails: value.exceptionDetails ? { text: truncate(value.exceptionDetails.text || "", 200), valuesRedacted: true } : null, valuesRedacted: true };
+  }
+  return { valuesRedacted: true };
+}
+
+function isTimeoutErrorMessage(value) {
+  return /timeout|timed\s*out|exceeded\s+\d+\s*ms/iu.test(String(value || ""));
+}
+
+function bytesToMb(value) {
+  const numeric = Number(value);
+  if (!Number.isFinite(numeric)) return null;
+  return Number((numeric / 1024 / 1024).toFixed(1));
+}
+
 function attachPassiveListeners(targetPage, pageRole = "control", targetPageId = pageId) {
   targetPage.on("request", (request) => {
     void appendJsonl(files.network, eventRecord("request", {
@@ -4713,6 +5052,14 @@ function parseAlertThresholds(value) {
     domEvaluateTimeoutRedWindowMs: requiredPositiveThreshold(raw, "domEvaluateTimeoutRedWindowMs"),
     screenshotTimeoutRedCount: requiredPositiveThreshold(raw, "screenshotTimeoutRedCount"),
     pageErrorRedCount: requiredPositiveThreshold(raw, "pageErrorRedCount"),
+    browserProcessSampleIntervalMs: requiredPositiveThreshold(raw, "browserProcessSampleIntervalMs"),
+    browserTotalRssRedMb: requiredPositiveThreshold(raw, "browserTotalRssRedMb"),
+    browserProcessRssRedMb: requiredPositiveThreshold(raw, "browserProcessRssRedMb"),
+    browserRssGrowthRedMb: requiredPositiveThreshold(raw, "browserRssGrowthRedMb"),
+    browserRssGrowthWindowMs: requiredPositiveThreshold(raw, "browserRssGrowthWindowMs"),
+    playwrightResponsivenessRedMs: requiredPositiveThreshold(raw, "playwrightResponsivenessRedMs"),
+    playwrightResponsivenessTimeoutRedCount: requiredPositiveThreshold(raw, "playwrightResponsivenessTimeoutRedCount"),
+    cdpMetricsTimeoutRedCount: requiredPositiveThreshold(raw, "cdpMetricsTimeoutRedCount"),
     uncommandedStateChangeCommandWindowMs: requiredPositiveThreshold(raw, "uncommandedStateChangeCommandWindowMs"),
     scrollJumpCommandWindowMs: requiredPositiveThreshold(raw, "scrollJumpCommandWindowMs"),
     scrollJumpFromY: requiredPositiveThreshold(raw, "scrollJumpFromY"),

scripts/src/hwlab-node/web-probe-observe.ts

@@ -2214,11 +2214,12 @@ export function recoverWebObserveAnalyzeFromArtifacts(options: NodeWebProbeObser
     "const srcMetrics = objectOrNull(source.sampleMetrics) || objectOrNull(recent.sampleMetrics) || {};",
     "const pagePerformance = objectOrNull(source.pagePerformance) || objectOrNull(recent.pagePerformance) || {};",
     "const runtimeAlerts = objectOrNull(source.runtimeAlerts) || objectOrNull(recent.runtimeAlerts) || {};",
+    "const browserProcess = objectOrNull(source.browserProcess) || objectOrNull(recent.browserProcess) || {};",
     "const promptNetwork = objectOrNull(source.promptNetwork) || objectOrNull(recent.promptNetwork) || {};",
     "const archiveSummary = objectOrNull(source.archiveSummary) || {};",
     "const archiveSampleMetrics = objectOrNull(archiveSummary.sampleMetrics) || objectOrNull(source.sampleMetrics?.summary) || objectOrNull(srcMetrics.summary) || {};",
     "const slowApis = arr(source.pagePerformanceSlowApi).length > 0 ? arr(source.pagePerformanceSlowApi) : arr(pagePerformance.sameOriginApiByPath).filter((item) => Number(item?.overBudgetCount ?? item?.overFiveSecondCount ?? 0) > 0);",
-    "const compact = { ok: source.ok === true, command: source.command ?? 'web-probe-observe analyze', stateDir: source.stateDir ?? stateDir, jsonlScope: source.jsonlScope ?? null, alertThresholds: source.alertThresholds ?? null, counts: source.counts ?? null, archiveSummary: { ...archiveSummary, sampleMetrics: archiveSampleMetrics, pagePerformance: objectOrNull(archiveSummary.pagePerformance) || objectOrNull(pagePerformance.summary) || {}, runtimeAlerts: objectOrNull(archiveSummary.runtimeAlerts) || objectOrNull(runtimeAlerts.summary) || {}, redFindings: arr(archiveSummary.redFindings).slice(0, 12).map(slimFinding) }, analysisWindow: source.analysisWindow ?? objectOrNull(recent.summary), sampleMetrics: compactMetrics(srcMetrics), pageProvenance: objectOrNull(source.pageProvenance?.summary) || source.pageProvenance ?? null, pagePerformance: objectOrNull(pagePerformance.summary) || pagePerformance, projectManagement: objectOrNull(source.projectManagement) || null, promptNetwork: objectOrNull(promptNetwork.summary) || promptNetwork, runtimeAlerts: objectOrNull(runtimeAlerts.summary) || runtimeAlerts, runnerErrors: arr(source.runnerErrors).slice(-8), commandFailures: arr(source.commandFailures).slice(-8), commandState: objectOrNull(source.commandState) || null, toolFindings: arr(source.toolFindings).slice(0, 8).map(slimFinding), httpErrorGroups: arr(source.httpErrorGroups ?? runtimeAlerts.networkHttpErrorsByPath).slice(0, 8).map(slimGroup), requestFailedGroups: arr(source.requestFailedGroups ?? runtimeAlerts.networkRequestFailedByPath).slice(0, 8).map(slimGroup), domDiagnosticGroups: arr(source.domDiagnosticGroups ?? runtimeAlerts.domDiagnosticsByText).slice(0, 5).map(slimGroup), domDiagnosticSamples: arr(source.domDiagnosticSamples ?? runtimeAlerts.domDiagnostics).slice(0, 8).map(slimGroup), consoleAlertGroups: arr(source.consoleAlertGroups ?? runtimeAlerts.consoleAlertsByPath).slice(0, 8).map(slimGroup), consoleAlertSamples: arr(source.consoleAlertSamples ?? runtimeAlerts.consoleAlerts).slice(0, 8).map(slimGroup), turnTimingRecentUpdateJumps: arr(source.turnTimingRecentUpdateJumps ?? srcMetrics.turnTimingRecentUpdateSawtoothJumps).slice(0, 8), turnTimingElapsedZeroResets: arr(source.turnTimingElapsedZeroResets ?? srcMetrics.turnTimingElapsedZeroResets).slice(0, 8), turnTimingTotalElapsedForwardJumps: arr(source.turnTimingTotalElapsedForwardJumps ?? srcMetrics.turnTimingTotalElapsedForwardJumps).slice(0, 8), pagePerformanceSlowApi: slowApis.slice(0, 8).map(slimSlowApi), archivePagePerformanceSlowApi: arr(source.archivePagePerformanceSlowApi).slice(0, 8).map(slimSlowApi), pagePerformancePartialApi: arr(source.pagePerformancePartialApi).slice(0, 8), pagePerformanceSseStreams: arr(source.pagePerformanceSseStreams).slice(0, 8), findings: arr(source.findings).slice(0, 12).map(slimFinding), archiveRedFindings: arr(source.archiveRedFindings ?? archiveSummary.redFindings).slice(0, 12).map(slimFinding), reportJsonPath: source.reportJsonPath ?? reportJsonPath, reportJsonSha256: source.reportJsonSha256 ?? sha256(reportJsonPath), reportMdPath: source.reportMdPath ?? reportMdPath, reportMdSha256: source.reportMdSha256 ?? sha256(reportMdPath), analyzer: { ...(objectOrNull(source.analyzer) || {}), recoveredFrom: 'analysis-artifact-after-transport-timeout', stdoutPath, stderrPath, stdoutBytes: statSize(stdoutPath), stderrBytes: statSize(stderrPath), reportJsonBytes: statSize(reportJsonPath), reportMdBytes: statSize(reportMdPath), transportExitCode: Number(transportExitRaw), transportTimedOut: transportTimedOutRaw === 'true', valuesRedacted: true }, valuesRedacted: true };",
+    "const compact = { ok: source.ok === true, command: source.command ?? 'web-probe-observe analyze', stateDir: source.stateDir ?? stateDir, jsonlScope: source.jsonlScope ?? null, alertThresholds: source.alertThresholds ?? null, counts: source.counts ?? null, archiveSummary: { ...archiveSummary, sampleMetrics: archiveSampleMetrics, pagePerformance: objectOrNull(archiveSummary.pagePerformance) || objectOrNull(pagePerformance.summary) || {}, runtimeAlerts: objectOrNull(archiveSummary.runtimeAlerts) || objectOrNull(runtimeAlerts.summary) || {}, browserProcess: objectOrNull(archiveSummary.browserProcess) || objectOrNull(browserProcess.summary) || {}, redFindings: arr(archiveSummary.redFindings).slice(0, 12).map(slimFinding) }, analysisWindow: source.analysisWindow ?? objectOrNull(recent.summary), sampleMetrics: compactMetrics(srcMetrics), pageProvenance: objectOrNull(source.pageProvenance?.summary) || source.pageProvenance ?? null, pagePerformance: objectOrNull(pagePerformance.summary) || pagePerformance, projectManagement: objectOrNull(source.projectManagement) || null, promptNetwork: objectOrNull(promptNetwork.summary) || promptNetwork, runtimeAlerts: objectOrNull(runtimeAlerts.summary) || runtimeAlerts, browserProcess: objectOrNull(browserProcess.summary) || browserProcess, runnerErrors: arr(source.runnerErrors).slice(-8), commandFailures: arr(source.commandFailures).slice(-8), commandState: objectOrNull(source.commandState) || null, toolFindings: arr(source.toolFindings).slice(0, 8).map(slimFinding), httpErrorGroups: arr(source.httpErrorGroups ?? runtimeAlerts.networkHttpErrorsByPath).slice(0, 8).map(slimGroup), requestFailedGroups: arr(source.requestFailedGroups ?? runtimeAlerts.networkRequestFailedByPath).slice(0, 8).map(slimGroup), domDiagnosticGroups: arr(source.domDiagnosticGroups ?? runtimeAlerts.domDiagnosticsByText).slice(0, 5).map(slimGroup), domDiagnosticSamples: arr(source.domDiagnosticSamples ?? runtimeAlerts.domDiagnostics).slice(0, 8).map(slimGroup), consoleAlertGroups: arr(source.consoleAlertGroups ?? runtimeAlerts.consoleAlertsByPath).slice(0, 8).map(slimGroup), consoleAlertSamples: arr(source.consoleAlertSamples ?? runtimeAlerts.consoleAlerts).slice(0, 8).map(slimGroup), turnTimingRecentUpdateJumps: arr(source.turnTimingRecentUpdateJumps ?? srcMetrics.turnTimingRecentUpdateSawtoothJumps).slice(0, 8), turnTimingElapsedZeroResets: arr(source.turnTimingElapsedZeroResets ?? srcMetrics.turnTimingElapsedZeroResets).slice(0, 8), turnTimingTotalElapsedForwardJumps: arr(source.turnTimingTotalElapsedForwardJumps ?? srcMetrics.turnTimingTotalElapsedForwardJumps).slice(0, 8), pagePerformanceSlowApi: slowApis.slice(0, 8).map(slimSlowApi), archivePagePerformanceSlowApi: arr(source.archivePagePerformanceSlowApi).slice(0, 8).map(slimSlowApi), pagePerformancePartialApi: arr(source.pagePerformancePartialApi).slice(0, 8), pagePerformanceSseStreams: arr(source.pagePerformanceSseStreams).slice(0, 8), findings: arr(source.findings).slice(0, 12).map(slimFinding), archiveRedFindings: arr(source.archiveRedFindings ?? archiveSummary.redFindings).slice(0, 12).map(slimFinding), reportJsonPath: source.reportJsonPath ?? reportJsonPath, reportJsonSha256: source.reportJsonSha256 ?? sha256(reportJsonPath), reportMdPath: source.reportMdPath ?? reportMdPath, reportMdSha256: source.reportMdSha256 ?? sha256(reportMdPath), analyzer: { ...(objectOrNull(source.analyzer) || {}), recoveredFrom: 'analysis-artifact-after-transport-timeout', stdoutPath, stderrPath, stdoutBytes: statSize(stdoutPath), stderrBytes: statSize(stderrPath), reportJsonBytes: statSize(reportJsonPath), reportMdBytes: statSize(reportMdPath), transportExitCode: Number(transportExitRaw), transportTimedOut: transportTimedOutRaw === 'true', valuesRedacted: true }, valuesRedacted: true };",
     "console.log(JSON.stringify(compact));",
     "UNIDESK_WEB_OBSERVE_RECOVER_ANALYZE_ARTIFACT",
   ].join("\n");