Autonomic Log and Trace Analyzer for Eclipse Release Notes

© Copyright International Business Machines Corporation 2000, 2006. All rights reserved. US Government Users Restricted Rights - Use, duplication or disclosure restricted by GSA ADP Schedule Contract with IBM Corp.

Known Problems and workarounds

1.0 Log and Trace Analyzer
1.1 Views
1.1.1 Problems with the Filtering and Highlighting functions in the Log view
1.2 Importing Logs
1.2.1 Problems when importing remote log files
1.2.2 Importing log files on Linux and AIX
1.2.3 Importing WebSphere Application Server trace.log
1.2.4 Some log files cannot be imported using IBM Agent Controller V5
1.2.5 Out of memory exception when importing large log files and symptom catalogs
1.2.6 Common Base Events shown for DB2 Express Diagnostic log are not to specification
1.2.7 No validation is done by Log and Trace Analyzer on the type of log file being imported
1.2.8 Limitation importing log files generated in a different locale
1.2.9 Import of WebSphere Application Server Activity Log with Rules parser from AIX fails
1.2.10 Cannot create a time based correlation with Apache Derby version 10.1.2.1
1.2.11 Performance slow when creating and showing correlation for large logs
1.3 Database support
1.3.1 Simultaneous clients access using the same database account is not supported
1.3.2 Resource path collision
1.3.3 Large Resource Support preferences remain disabled despite valid selection
1.3.4 Improving performance of large log support
1.3.5 Restart DB2 after creating database and tables
1.3.6 Intermittent SQL exception
1.3.7 Test connection returns an error if you haven't applied preferences
1.3.8 CommonBaseEvent XPath expression is not supported when analyzing large logs
1.3.9 Only one contextDataElement is displayed for large logs
1.4 Other
1.4.1 New Log Correlation wizard not displayed properly on Linux GTK driver
1.4.2 Modifications to a plug-in's logging level after the first start of the workbench are not respected
1.4.3 Crash when creating an example on Red Hat Linux v8.0 with IBM JVM
1.4.4 Only TPTP format symptom databases can be created from the Log Navigator toolbar and popup menu
1.4.5 IBM Log Analyzer only processes messages up to 1024 bytes
1.4.6 Symptom Editor generates XPATH rules that are not valid when using complex elements
1.4.7 Symptom Editor generates XPATH rules that do not match any input Common Base Events
2.0 Profiling Tool
2.1 Agent and Process Interaction views are not supported from Sequence Diagram view
2.2 Refresh views command of the Profiling Navigator does not work for Trace Interactions
2.3 Profiling on Windows using the Sun JDK 1.4.x can end with a Microsoft(R) error report

Additional readme information associated with TPTP Log and Trace Analyzer is available in the TPTP V4.2 release notes.

1.0 Log and Trace Analyzer

1.1 Views

1.1.1 Problems with the Filtering and Highlighting functions in the Log view

In the Log View, the Filtering and Highlighting functions do not work on complex elements of the Common Base Event (for example, sourceComponentId). A fix will be provided in the next fix pack.

1.2 Importing Logs

1.2.1 Problems when importing remote log files

When importing a remote log file (File > Import... > Log File), no log entries are displayed in the Log View after refreshing the Profiling Monitor's Views (Profile > Refresh Views) or a Logging Message dialog is displayed stating that the local client has not started monitoring the remote log parser after waiting 30 seconds.

This problem is due to network latency and can be fixed by increasing the time Agent Controller waits for the local client to start monitoring the remote log parser. To work around this problem, try the following steps:

  1. Try to import the log file again.
  2. If the symptoms persist, add the following entry to the remote RemoteLogParserLoader application in Agent controller's configuration (i.e. <Agent Controller's installation directory>/plugins/com.ibm.etools.logging.parsers/config/pluginconfig.xml):
      
    <Application configuration="default" executable="RemoteLogParserLoader" 
    extends="default" location="%SYS_TEMP_DIR%" path="%JAVA_PATH%">
    ...
    <Parameter position="prepend"
    value="-Dorg.eclipse.hyades.logging.parsers.maxWaitTimeInMillis=xxxxx"/>
    ...
    </Application>
    
    where xxxxx is the maximum wait time in milliseconds (i.e. greater than 30000).
  3. Once the update is made, restart Agent Controller and try importing the log file again.

1.2.2 Importing log files on Linux and AIX

Some log file imports do not work at all on the Linux(R) and AIX(R) platforms.

This occurs when the user does not have proper permissions to access the log files. To resolve this issue, the log files should be copied into the user's home directory and imported from there.

1.2.3 Importing WebSphere Application Server trace.log

An Internal Correlation is created when a WebSphere(R) Application Server trace.log file is imported.

The user should not delete these correlations as they contain information related to the internal structure of the corresponding agent.

1.2.4 Some log files cannot be imported using IBM Agent Controller V5

Remote log files other than activity.log cannot be imported using a V5 IBM(R) Agent Controller. When importing log files remotely, IBM Agent Controller on the remote machine must be at the same level or higher as IBM Log and Trace Analyzer for Eclipse.

To work around this limitation, install IBM Agent Controller shipped with IBM Log and Trace Analyzer for Eclipse on the remote machine.

1.2.5 Out of memory exception when importing large log files and symptom catalogs

Log files and symptom catalogs up to 25MB can be imported into the workspace and opened into Log and Trace Analyzer.

The amount of time it takes to open the log file depends on the number of log records in the file, and you may get an "out of memory" exception. To work around this problem, use the large log support feature.

If you get an "out of memory" exception when importing large symptom catalogs, increase the heap size by adding the following vmargs:

-vmargs -Xmx1000m

If using the large log support feature does not fix the "out of memory" exception problem when importing large logs, increasing the heap size should help.

1.2.6 Common Base Events shown for DB2 Express Diagnostic log are not to specification

If you import an IBM DB2(R) Express Diagnositic log, the resulting Common Base Events shown in Log and Trace Analyzer break some of the Common Base Event specification. This is due to a bug in the db2diag.exe utility which is used to convert DB2 Diagnostic logs to Common Base Events. This should not otherwise affect the import operation. There is no known work-around until db2diag.exe is fixed.

1.2.7 No validation is done by Log and Trace Analyzer on the type of log file being imported

No validation is done by Log and Trace Analyzer on the type of log file being imported. When importing a log file ensure that the correct log type and version is chosen for the log file that is being imported. If a problem occurs during the import, no records or incorrect records are shown in the Log View and an error message will be displayed. For more detailed information about the error, view the details from the error dialog.

1.2.8 Limitation importing log files generated in a different locale

When importing log files that contain timestamps in a locale different than the locale that Log and Trace Analyzer is executing in, the rules-based log file parser sometimes does not parse the timestamp and other log record data correctly. For example, when importing an English IBM HTTP Server access log file into Log and Trace Analyzer running on a Japanese machine, the creationTime values in the resulting Common Base Events will have a value of zero and the Common Base Event msg field will be blank.

To work around this problem, when importing log files containing timestamps in a different locale than the current machine locale, specify the locale of the log file on the Details tab in the Log Import wizard, if a locale field exists.

1.2.9 Import of WebSphere Application Server Activity Log with Rules parser from AIX fails

Importing a binary WAS Activity log with the Rules parser from an AIX system fails with the following error displayed in the error dialog:

IWAT0030E An error occurred during the execution of the remote log 
parser "com.ibm.etools.logging.adapter.config.StaticParserExtension": 
IWAT0412E Errors occurred parsing the log file /home/tfoun/logs/activity.log.
java.lang.Exception: IWAT0239E Converter command failed: java.lang.Exception: 
IWAT0238E Converter process ended with exit value 1

The converter command to convert the binary log file to text so that it can be parsed fails on AIX. To work around this problem, either use the Static parser to import the binary activity log file or convert the binary activity log file to text using the WAS showlog utility and import the resulting text file into Log and Trace Analyzer. The WAS showlog utility can be found in the bin directory in the WAS installation directory. For example, use the following command to convert the binary log file to a text file called activity.txt:

/opt/WebSphere/AppServer/bin/showlog activity.log activity.txt

Then import activity.txt into Log and Trace Analyzer.

1.2.10 Cannot create a time based correlation with Apache Derby version 10.1.2.1

Although not officially supported by TPTP, a TPTP user attempted to use Derby 10.1.2.1 in conjunction with large log support and experienced this problem. When trying to create a time based correlation with a large log, an ArrayIndexOutOfBoundsException occurred which caused the correlation to run continuously without stopping. This is a known Derby bug that is fixed in Derby version 10.2.0.0. For more information, see the description of the Derby bug at http://www.archivum.info/derby-dev@db.apache.org/2006-03/msg01624.html.

Note that this exception occurs on the server side and is not seen on the user's workbench. It is displayed in the command window where the command is issued to start the Derby server.

1.2.11 Performance slow when creating and showing correlation for large logs

When creating or showing a correlation for a large log, performance may be slow with Apache Derby and DB2.

To improve performance when creating a correlation, import logs with an import filter so that unnecessary or unrelated events are not included in the database. Using a "Show correlated events only" filter can improve performance when opening the correlation in the Log Interactions view.

1.3 Database Support

1.3.1 Simultaneous clients access using the same database account is not supported

Concurrent client access to a database might lock some tables. To unlock the tables, restart the Workbench and the database server.

1.3.2 Resource path collision

If you use the same database and database account for multiple workspaces, resource path collisions may result and resource will not be persisted into the database.

To work around this problem, use different project and monitor names in each workspace to avoid resource path collisions.

1.3.3 Large Resource Support preferences remain disabled despite valid selection

In Linux/GTK, on the Large Resource Support preferences page, the database settings fields will remain disabled even when a valid database type is selected.

To work around this problem, select DB2 as the database type, and apply the settings. Close and reopen the preferences dialog; the fields will be enabled.

1.3.4 Improving performance of large log support

To improve the performance of large log support, run the following command after you have imported large logs:

db2 -tvf plugins/com.ibm.etools.ac.resources.database_x_x_x
/scripts/runStatsForAllHyadesTablesDB2-8.1.sql
where x_x_x is the version number on the plug-in directory.

1.3.5 Restart DB2 after creating database and tables

After running the database and tables creation script, CreateDatabaseAndTablesDB2-8.1.sql, you must restart DB2. From a DB2 command window run db2stop and then db2start in order to apply the changes to the configuration parameters made by the script.

1.3.6 Intermittent SQL exception

Occasionally, the following SQL exception may occur:

com.ibm.db2.jcc.b.SQLException: NULLID.SYSSH203 0X5359534C564C3031

When this problem occurs, restart the workbench and the database, if using DB2 or CloudscapeTM in network mode. Logs imported when this exception occurs will not be valid and will have to be re-imported.

1.3.7 Test connection returns an error if you haven't applied preferences

In the Large Log Support page of the Profiling and Logging section of the Preferences dialog, if you click Test connection before you have applied changes to settings, you may get a connection error. To avoid this, click Apply after making any setting changes and then click Test connection.

1.3.8 CommonBaseEvent XPath expression is not supported when analyzing large logs

The following XPath expression is not supported by IBM Log Analyzer when analyzing large logs.

	<expression:xpathExpression>/CommonBaseEvent</expression:xpathExpression>

1.3.9 Only one contextDataElement is displayed for large logs

If DB2 is used for large log support, only one contextDataElement for an event is displayed in the Log view, even if there are multiple contextDataElements for that event.

1.4 Other

1.4.1 New Log Correlation wizard not displayed properly on Linux GTK driver

In the New Log Correlation wizard, only the Available Logs list is shown, and the list is empty. This is due to the fact that only the left portion of the wizard page is displayed.

To work around this problem, try resizing the wizard page to have the Selected Logs list and actions buttons shown. An alternate workaround is to select the log files for correlation in the Log Navigator view before clicking the New Log Correlation button.

1.4.2 Modifications to a plug-in's logging level after the first start of the workbench are not respected

If you modify a plug-in's logging level in its plugin.xml file (via the com.ibm.etools.common.logging.commonLoggingOptions or com.ibm.etools.logging.util.loggingOptions extension points) after the first start of a workbench, the change is not respected in subsequent launches of the workbench. The logging level specified in the plug-in's plugin.xml file is cached by the Eclipse run-time on the first start of the workbench. Subsequent starts of the workbench use the cached value of the plug-in's logging level despite modification in the plugin.xml file.

To work around this problem, restart the workbench with the -clean command line argument. This option will clean the caches used to store bundle dependency resolution and eclipse extension registry data and force Eclipse to reinitialize these caches. Add the option after the eclipse.exe in the <workbench installation directory>\eclipse\ac.bat (Windows(R)) or <workbench installation directory>/eclipse/ac.sh (Linux). Alternatively, you can modify the plug-in's logging level on the Logging preferences page (Window > Preferences > Logging) and restart the workbench.

1.4.3 Crash when creating an example on Red Hat Linux v8.0 with IBM JVM

When creating an example in Log and Trace Analyzer using the Example Creation wizard (New > Example), the Workbench crashes. This happens when the Example Creation wizard tries to open the readme.html file as part of the creation steps. This problem is found on Red Hat Linux v8.0 with the IBM JavaTM Virtual Machine (JVM), and is an SWT bug (https://bugs.eclipse.org/bugs/show_bug.cgi?id=76515).

To work around this problem, try adding -Xj9 as a JVM argument in ac.sh to launch using the IBM J9 JVM. (for example, ./eclipse -vmargs -Xj9 -Xmx500m). For more information about running SWT on Red Hat Linux v8.0, check the SWT FAQ (http://dev.eclipse.org/viewcvs/index.cgi/%7Echeckout%7E/platform-swt-home/faq.html).

Note: As stated on Eclipse reference platforms table (http://www.eclipse.org/eclipse/development/eclipse_project_plan_3_0.html#TargetOperatingEnvironments), only the following Linux systems are supported: version 2.2.1 of the GTK+ widget toolkit and associated librares (GLib, Pango); SWT HTML viewer requires Mozilla 1.4GTK2. For Motif on other Linux systems: Open Motif 2.1 (included); SWT HTML viewer requires Mozilla 1.4GTK2.

1.4.4 Only TPTP format symptom databases can be created from the Log Navigator toolbar and popup menu

Because of a TPTP limitation users cannot create both types of symptom databases from the log navigator toolbar button and context menu. A workaround is to select File > New > Other > Profiling and Logging from the workbench menu and choose the type of symptom database you want to create.

1.4.5 IBM Log Analyzer only processes messages up to 1024 bytes

In XPATH rules and IBM correlation rules, the string in the msg Common Base Event property should not be more than 1024 bytes. If the string is more than 1024 bytes, they will not be processed by IBM Log Analyzer.

1.4.6 Symptom Editor generates XPATH rules that are not valid when using complex elements

When using complex elements of the Common Base Event (for example, sourceComponentID), Symptom Editor generates XPATH rules that are not valid. A fix will be provided in the next fix pack.

1.4.7 Symptom Editor generates XPATH rules that do not match any input Common Base Events

If any of the ExtendedDataElements are selected when creating a new XPATH expression, the resulting symptom catalog contains extendedDataElement instead of extendedDataElements. As a result, Symptom Editor generates XPATH rules that do not match any input Common Base Events.

2.0 Profiling Tool

2.1 Agent and Process Interaction views are not supported from Sequence Diagram view

The Agent Interactions view and the Process Interactions view are not supported from the Sequence Diagram view for log files.

There is currently no workaround for this problem.

2.2 Refresh views command of the Profiling Navigator does not work for Trace Interactions

Refresh views command of the Profiling Navigator does not work for Trace Interactions. However, the trace interactions automatically refresh on their own at regular intervals.

A workaround is to select another node in the Profiling tree and select again the previous node.

2.3 Profiling on Windows using the Sun JDK 1.4.x can end with a Microsoft error report

Bugzilla defect: 103058

The profiling or application terminates with the message "java.exe has encountered a problem and needs to close. We are sorry for the inconvenience." The message contains information and choices for submitting an error report to Microsoft. See bugzilla 103058 for a screen capture of the message.

This problem has been observed in a number of modes: Time Analysis (with or without "Show Instances"), Memory Analysis, and in combinations of these modes. You can work around the problem by re-running the profiling or appliaction with JIT turned off, i.e. by supplying "-Djava.compiler=NONE" as a JVM argument. The problem has been encountered with the SUN JDK 1.4.2_08-b03 build for Windows.