wM Integration Server Service Profiler, v1.1 (with patch #2)

Installation

• The Service Profiler tool directory is automatically created on 1^st time installation.
  Manual copy of the directory tree is no longer required.

Administration

• Profiling sessions can be named;
  A profiling session is started when the profiler is started and ends when the profiler is stopped.
  Each time you start the profiler, you can give it a diferent name. This name is part of the identification header of the snapshot with a timestamp appended.
  If the name is ommited, it defaults to the host name and port.
• Completely disabling the Service Profiler can be done from the tool's admin page;
  This operation still requires a restart to the Integration Server, for which a graphical hint is displayed.
• Clear generated export files.
  Whenever a snapshot is exported by generating a link for download, a file gets created file in the server's file system.
  These files are removed whenever the profiler package is (re)loaded. But now they can be removed directly from the profiler admin page.
  With this new version, it is also possible to export the snapshot without generating a file in the server's file system.

Snapshot XML File

• The XML contains the version of the snapshot structure as an attribute;
  If, and when, the snapshot XML structure changes, the version value will reflect this.
• The exported XML file is enriched with service metadata;
  This metadata allows loading the XML back into the tool's dashboard and still show the correct service type;
• A previouly exported XML file can be uploaded for viewing;
  The Load Snapshot allows loading previously exported snapshots (as XML) for viewing in the tool's dashboard.
  The snapshot can be from an Integration Server other than the one the profiler is running under.
• Export as content stream.
  With this new version it is possible to export the snapshot directly to your local drive without having to generate a file in the server's file system.
  Instead of creating a link to the generated file, the file browse dialog is open for you to choose the location where you want the file to be created, with the default given name or a name of your choice.

Dashboards

• Added the Maximum Thread Count (MTC) to the View Per Service and Code Coverage dashboards;
  This value indicates the maximum number of service threads used simultaneously during the profiling session.
  This value is part of the snapshot header and is included in the exported files.
• Browse Running Services with Java Stack Trace and Thread Info;
  This feature is only available if the Integration Server is running in a JVM of v1.5 or above.
  By default, the dashboard gets the snapshot without thread info, but this can be requested by selection an option in the dashboard and pressing the Refresh button.
  The thread info includes:
  • Thread state;
  • Block and suspend information;
  • Java Stack Trace;
  • Node cross-referencing.
    If the thread owning the block is a service thread and can be referenced in the current snapshot its ID is a link directly to that node in the tree.
  The icon representing the thread in the tree is also changed to reflect some of this information, which can be accessed by clicking the thread name link.
• On the Browse Snapshot and Browse Running Services dashboards, search a service in the tree using a Regular Expression over its name.
• Added the count of services of each package in the Code Coverage dashboard;
• Show the calling service name on service details when browsing the service call tree.
  It is only shown for False Roots.

Public API

• Added new public services to start & stop the profiler.
  • wia.pub.issprofiler.admin:startProfiler
  • wia.pub.issprofiler.admin:stopProfiler
  
  These avoid some potential issues with frozen snapshots and also having to use the UI services.

• Correct paging control of a large tree with small page setting;
  If a configuration of maximum nodes per page was defined as very small (20 to 50 nodes per page) a too great number of pages could be generated, being the most of them empty.
• False Root detection and reporting:
  • Not all False Roots were being detected;
  • Those False Root services were actually being detected were correctly flagged but were absent from the dashboard imagery.
    This means that the services were correctly flagged but the icon did not reflect that state.
    This affected the Browse Snapshot and Browse Running Services dashboards.
• On the Browse Running Services dashboard, on the Service Details panel of the currently running service, the Running Instance elapsed time was always shown as 0 (zero);
• On the Browse Snapshot and Browse Running Services dashboards, when navigation between pages, there was the possibility of ending up with a node selected in the tree and a blank details panel.
  Now, to keep the front-end consistent, if the details panel is blank, the selected node is automatically cleared (unselected).
• With Firefox 3.0.x (mainly, but not in exclusive), when using the browsing the tree of services, expanding the tree could make the graphical hints appear dissociated from the corresponding service icons;
• Corrected the sorting by percentage in the Code Coverage dashboard when packages without services are being listed.

None.

For upgrading from a previous version, it is recommended to completely uninstall the previous version and install the new version from scratch. The scope of the introduced changes reach the replacement of the WiaUtilities package, the entire set of libraries (both native and Java) and also the script code added to the server.(bat|sh) file.

The Service Profiler has been successfully tested using SUN, HP and IBM JVMs v1.3.1, v1.4.x and v1.5.0.

When using JREs other than the Integration default one, make sure to make the tools.jar available, usually by putting it into the ext folder. This library is not included with the JRE installations, but is part of the JDK. The Integration Server default JVM includes this library (even though it is not guaranteed that it will always do so) for it's needed for some functionality, such as JSPs.

The Service Profiler can run in any of the Operating Systems and CPU architectures supported by the webMethods Integration Server. However, it is not ready available in all of them.

The list of currently ready-available Operating and CPU architectures are as follows:

• Windows 2000, XP & 2003 Server (intel 32-bit)
• Windows 2003 Server x64 (AMD 64-bit)
• Linux Redhat (intel 32-bit) (AMD 64-bit)
• Linux Suse (intel 32)
• SUN Solaris 8 (SPARC 32-bit)
• SUN Solaris 9 (SPARC 32-bit & 64-bit)
• HP-UX B10 & B11 (PA RISC 32-bit & 64-bit) (Itanium 32-bit & 64-bit)
• Mac OS X (intel 32 & 64-bit)
• AIX (ppc 32 & 64-bit)
  Introduced with patch #2 as experimental.
  Any feedback is welcome.

The list of supported Operating Systems may change through the release of new patches.

• When the command-line profiler options are actived Norton Internet Security may prevent the Integration Server from starting;
  Manually adding the java.exe file to the permitted files of the firewall configuration usually overcomes this problem.
  If it does not work, disable Norton Internet Security while the server is starting.
  This situation may also happen with other security/firewall software, even though it has not been tested.
• Globalization;
  The tool is only available in English language.
• Missing menu link;
  The Licensing menu link may happen to not appear at all.
  The reason for this is not yet known, and usually refreshing the page makes it reappear.
• Grayed/disabled links.
  This is not an issue.
  When functionality menu links are disabled, this means that either the installation of the tool is incomplete or some error/setting is preventing the tool core libraries from loading.
  Check logs and if it is a problem, please call for support.
• Files downloaded with Save As... option of the browser are truncated/corrupted.
  This has been withnessed with Mozilla v1.7.12.
  Neither IExplorer or Firefox v1.5.x have reproduced this behaviour.
  This problem has been noted when copying the PDF documents in the info page or the (link) files generated on the Export To File functionality.
  Alternative workaround is to open the file by clicking its link and:
  • PDF;
    Use the Save Copy toolbar button of the plug-in.
  • CSV or XML.
    Use the This Frame / View Frame Source context menu and save it to a file.
• Pagination of the tree of service nodes limits find feature;
  If the number of nodes in a Snapshot triggers the tree pagination in the dashboard, the find a node in the tree is limited to the current page.
  On the special case of the navigation link from the View Per Service to the Browse Snapshot, the navigation & find is always (and only) made onto the first page. However, jumping to the next/previous page and repeat the find is always possible, but hits on the entire snapshot are not seen all at once.
• In the snapshot identification the host port may appear as -1 or missing.
  This will not happen if using the Service Profiler manually; i.e., start the profiling when the Integration Server has already completed its startup sequence, and the Service Profiler is also already loaded.
  The -1 value will happen if the port is inquired to the Integration Server during a package startup sequence.
  During the Integration Server startup sequence, its listening port number is also undetermined. Depending how it is inquired, it may result in -1 or nothing at all.