WebSphere环境下iText 5.5.6处理超长文本卡顿耗时问题排查

环境与问题概述 #

• 应用技术栈：iText 5.5.6、Java 1.8.0_171 64位
• 部署环境：WebSphere Application Server ND 9.0.5.13、Windows Server 2016 Standard 1607 64位
• 核心功能：实现HTML转PDF
• 问题表现：当输入3万字符的超长单词（如连续的"aaaaaaaa..."）时，代码在document.add(myPdfPTable)行出现卡顿，无异常抛出；相同场景在Tomcat服务器下无此问题

相关代码 #

生成PDF核心方法 #

public Document generatePDF(String fontPath,String html) throws Exception {

        com.itextpdf.text.Document document = new com.itextpdf.text.Document(PageSize.A4, 20, 20, 100, 50);

        ByteArrayOutputStream baos = new ByteArrayOutputStream();
        pdfWriter = PdfWriter.getInstance(document, baos);

        pdfWriter.setPageEvent(ppaHeaderFooter);
        pdfWriter.setViewerPreferences(PdfWriter.PageModeUseOC);
        pdfWriter.setPdfVersion(PdfWriter.VERSION_1_5);
                    
        document.open();
        document.addAuthor(authorName);
        document.addCreator(authorName);
        document.addSubject(filename);
        document.addCreationDate();
        document.addTitle(filename);
        
        Font arabicFont = FontFactory.getFont(fontPath, BaseFont.IDENTITY_H, 14);
        FontFactory.register(fontPath);
        

        Font bold = FontFactory.getFont(fontPath, BaseFont.IDENTITY_H, 14f, Font.BOLD);
        Font regularfont = FontFactory.getFont(fontPath, BaseFont.IDENTITY_H, 10);
        List<Element>  pdfElements = getElementsFromHtml(html, fontPath);

        PdfPTable contentTable = new PdfPTable(1);
        contentTable.setRunDirection(PdfWriter.RUN_DIRECTION_RTL);
        contentTable.setSplitLate(false);
        contentTable.setWidthPercentage(90);
        contentTable.setExtendLastRow(true); 

        

        for (Element element : pdfElements) {
            PdfPCell cell = new PdfPCell();
            cell.setHorizontalAlignment(Element.ALIGN_CENTER);
            cell.setVerticalAlignment(Element.ALIGN_CENTER);
            cell.setBorder(Rectangle.NO_BORDER);
            cell.setArabicOptions(ColumnText.DIGITS_EN2AN);
            cell.addElement(element);
            contentTable.addCell(cell);
        }

        
        document.add(contentTable);
        return document;
        
}

HTML转PDF元素解析方法 #

public ArrayList<Element> getElementsFromHtml(final String html, String fontPath) throws IOException {

        final ElementList elements = new ElementList();

        StringBuilder stylefile = new StringBuilder();

        stylefile.append("body {font-size: 14pt;font-family:MyCustomFont;line-height: 1.5;dir:rtl}");
        stylefile.append("table{padding: 0px;border-spacing: 0px;font-size: 14pt;}");
        stylefile.append("td{height:30px;}");

        

        CSSResolver cssResolver = new StyleAttrCSSResolver();
        CssFile cssFile = XMLWorkerHelper.getCSS(new ByteArrayInputStream(stylefile.toString().getBytes()));
        cssResolver.addCss(cssFile);

        XMLWorkerFontProvider fontProvider = new XMLWorkerFontProvider(XMLWorkerFontProvider.DONTLOOKFORFONTS);
        fontProvider.register(fontPath, "MyCustomFont");
        CssAppliers cssAppliers = new CssAppliersImpl(fontProvider);
        HtmlPipelineContext htmlContext = new HtmlPipelineContext(cssAppliers);
        htmlContext.setTagFactory(Tags.getHtmlTagProcessorFactory());

        final ElementHandlerPipeline end = new ElementHandlerPipeline(elements, null);
        final HtmlPipeline htmlPipeline = new HtmlPipeline(htmlContext, end);
        final CssResolverPipeline cssPipeline = new CssResolverPipeline(cssResolver, htmlPipeline);

        
        final XMLWorker worker = new XMLWorker(cssPipeline, true);
        final XMLParser p = new XMLParser(worker, Charset.forName("UTF-8"));

        final String resolvedHtml = "<body>" + html + "</body>";
        p.parse(new ByteArrayInputStream(resolvedHtml.getBytes(Charset.forName("UTF-8"))), Charset.forName("UTF-8"));

        return new ArrayList<Element>(elements);
}

UPDATE1：线程挂起栈信息 #

启用WebSphere挂起线程检测后，定位到线程挂起的调用栈：

ThreadMonitor W   WSVR0605W: Thread "WebContainer : 4" (00003bed) has been active for 672628 milliseconds and may be hung.  There is/are 1 thread(s) in total in the server that may be hung.
    at com.ibm.tivoli.itcam.gdc.data.GDCInstanceDataManager.resetProvisional(GDCInstanceDataManager.java:489)
    at com.ibm.tivoli.itcam.gdc.GDCMethodProbePoint.customProbeEndEvent(GDCMethodProbePoint.java:1691)
    at com.ibm.tivoli.itcam.toolkit.ai.boot.callbacks.generic.CustomProbeCallback.customProbeAfterCallbackInstance(CustomProbeCallback.java:450)
    at com.ibm.tivoli.itcam.toolkit.ai.boot.callbacks.generic.CustomProbeCallback.customProbeAfterCallback(CustomProbeCallback.java:409)
    at com.itextpdf.text.pdf.PdfFont.width(PdfFont.java:138)
    at com.itextpdf.text.pdf.PdfChunk.getCharWidth(PdfChunk.java:961)
    at com.itextpdf.text.pdf.BidiLine.processLine(BidiLine.java:383)
    at com.itextpdf.text.pdf.ColumnText.go(ColumnText.java:1068)
    at com.itextpdf.text.pdf.ColumnText.go(ColumnText.java:990)
    at com.itextpdf.text.pdf.ColumnText.goComposite(ColumnText.java:1523)
    at com.itextpdf.text.pdf.ColumnText.go(ColumnText.java:996)
    at com.itextpdf.text.pdf.ColumnText.go(ColumnText.java:990)
    at com.itextpdf.text.pdf.PdfPCell.getMaxHeight(PdfPCell.java:1044)
    at com.itextpdf.text.pdf.PdfPTable.getFittingRows(PdfPTable.java:2141)
    at com.itextpdf.text.pdf.ColumnText.goComposite(ColumnText.java:1753)
    at com.itextpdf.text.pdf.ColumnText.go(ColumnText.java:996)
    at com.itextpdf.text.pdf.ColumnText.go(ColumnText.java:990)
    at com.itextpdf.text.pdf.ColumnText.go(ColumnText.java:978)
    at com.itextpdf.text.pdf.PdfDocument.addPTable(PdfDocument.java:2673)
    at com.itextpdf.text.pdf.PdfDocument.add(PdfDocument.java:750)
    at com.itextpdf.text.Document.add(Document.java:278)

UPDATE2：性能差异对比 #

• WebSphere环境：最终PDF可生成，但耗时约15分钟（文件大小350KB，共280页）
• Tomcat 9+Java 8环境：相同请求仅需30秒，问题与WebSphere环境强相关

UPDATE3：JVM配置信息 #

JVM参数 #

-Dcom.ibm.jsse2.overrideDefaultProtocol=TLSv12 
-Dcom.ibm.jsse2.overrideDefaultTLS=true 
-DFileNet.WSI.AutoDetectLTPAToken=true 
-Dcom.filenet.authentication.token.userid=sso:ltpa 
-Dam.wasserver=server1 
-Dam.wasnode=MyNode01 
-Dam.wasprofile=AppSrv01 
-Dam.wascell=MyNode01Cell
-Dcom.ibm.tivoli.itcam.ai.runtimebuilder.inputs=${ITCAMDCHOME}/runtime/AppSrv01.MyNode01Cell.MyNode01.server1.DCManualInput.txt  
-Dws.bundle.metadata=${ITCAMDCHOME}/runtime/wsBundleMetaData 
-Djava.security.policy=${ITCAMDCHOME}/itcamdc/etc/datacollector.policy 
-Xbootclasspath/p:${ITCAMDCHOME}/toolkit/lib/bcm-bootstrap.jar 
-Dsun.rmi.transport.connectionTimeout=300000 
-Dsun.rmi.dgc.server.gcInterval=3600000 
-Dsun.rmi.dgc.client.gcInterval=3600000-verbosegc -Xgc:allocationSamplingGranularity=10000 
-Xhealthcenter:transport=jrmp,level=inprocess 
-agentlib:am_ibm_16=${WAS_SERVER_NAME} 
-Djava.net.preferIPv6Addresses=false 
-Djava.net.preferIPv4Stack=true 
-Dcom.ibm.cacheLocalHost=true 
-Djaxws.payload.highFidelity=true 
-Djava.security.auth.login.config=C:\FileNet\AE\config\jaas.conf.WebSphere -agentlib:getClasses

JVM自定义属性 #

![JVM自定义属性截图]

内容的提问来源于stack exchange，提问作者Mahmoud Saleh