Sometimes the most obvious solutions are overlooked. As you begin your investigation, try the following steps first.

Troubleshooting and analysis require information. The more information you have the easier it is to classify a problem and search for its solution. Detailed information will also be necessary should you need to contact others for support, either through a community mailing list or through more formal Sun support channels.

As you continue your investigation, consider the following questions.

Logging is one of your most important troubleshooting tools. It is the process by which Eclipse GlassFish Server captures data about events that occur during server operation, such as configuration errors, security failures, or server malfunction. This data is recorded in log files, and is usually your first source of information when Enterprise Server problems occur. The primary purpose of log files is to provide troubleshooting information. Analyzing the log files can help determine the health of the server and identify problem areas.

By default, log information for each GlassFish Server Open Source Edition server instance is captured in a server.log file. That is, each instance, including the domain administration server (DAS), has an individual log file. By default, the log file for the DAS is located in domain-dir`/logs`, and the log file for each instance is located in instance-dir`/logs`.

In addition, for domains that use clustering, GlassFish Server Open Source Edition captures log information for each cluster instance in a cluster.log file. By default, the cluster.log file is also located in instance-dir`/logs`.

Oracle recommends using the Administration Console to view logging information. However, you can open a log file in a text editor and search for the module or message in which you are interested. GlassFish Server Open Source Edition also lets you collect log files into a ZIP file, which provides a convenient means to collect and view the log files for an instance or a domain even when it is not running.

You configure the Logging Service by setting attributes in the logging.properties file. Each server, configuration, instance, and cluster in the Eclipse GlassFish Server domain has an individual logging.properties file. The root directory in which these logging.properties files are located is the same directory as for the domain.xml file, typically domain-dir/config. The default target when configuring logging attributes is the DAS. However, you can optionally target a specific server, instance, or cluster. You can also target a configuration that is shared by one or more instances or clusters. The Logging Service can also be configured using the Administration Console.

Log levels such as SEVERE, WARNING, INFO, CONFIG, and others can be set to provide different types and amounts of information. The default setting is INFO. Each Eclipse GlassFish Server module has its own logger, and each logger has its own namespace. Log levels can be set globally for all loggers, or individually for module-specific loggers.

For information about using the Administration Console log viewer and logging functions, see the Administration Console online help. For information about using the command line for logging functions, see "Administering the Logging Service" in GlassFish Server Open Source Edition Administration Guide.

Monitoring is another helpful tool. It is the process of reviewing the statistics of a system to improve performance or solve problems. By monitoring the state of various components and services deployed in Eclipse GlassFish Server you can identify performance bottlenecks, predict failures, perform root cause analysis, and ensure that everything is functioning as expected. For more information about monitoring, including JConsole information, see "Administering the Monitoring Service" in GlassFish Server Open Source Edition Administration Guide.

Operating system utilities, such as pkginfo and showrev on Solaris and rpm on Linux, are helpful in gathering system information.

The ps -ef command provides helpful information about processes that are running, including their process identification numbers (PIDs).

A stack trace is a user-friendly snapshot of the threads and monitors in a Virtual Machine for the Java platform (Java Virtual Machine or JVM machine). A thread dump shows what every thread in a JVM is doing at a given time and is useful in debugging. When the application server freezes, hangs, or becomes sluggish for no apparent reason, you should generate and analyze a thread dump.

This section explains how to obtain a thread dump for GlassFish Server Open Source Edition. More information about analyzing the information contained in a thread dump can be found in "http://java.sun.com/developer/technicalArticles/Programming/Stacktrace[An Introduction to Java Stack Traces]" (http://java.sun.com/developer/technicalArticles/Programming/Stacktrace).

VisualVM is a Java troubleshooting tool that uses various technologies such as jvmstat, JMX, and Attach API to access monitored applications. VisualVM is a tool for visualizing data sources and by default visualizes the following types: applications, hosts, snapshots, core dumps, heap dumps, and thread dumps. These data sources are visualized in VisualVM so that they can be monitored for the purposes of analysis, management, and troubleshooting. VisualVM is commonly used to detect memory leaks.

VisualVM has a GlassFish plugin that enhances monitoring of hosted applications by adding specialized overview, a tab for monitoring the HTTP Service, and the ability to visually select and monitor any of the deployed web applications. You can experiment with VisualVM troubleshooting capabilities, but note that various features depend on the Java versions used in the client and server. Depending on your configuration, you might only get parts of the VisualVM features. For more information about VisualVM, see http://visualvm.java.net.

JVM command-line tools can be used to provide valuable information about hung Java processes or Java core dumps. These tools include the following: