How to use a Java profiler like JVisualVM or JProfiler?

The key to using Java Profiler is to understand its functionality and follow the steps. 1. Start the tool and connect to the target application. JVisualVM can directly run and automatically identify local processes. JProfiler needs to be installed and supports remote connections. 2. Analyze CPU and memory. JProfiler provides "Call Tree" and "Hot Spots". JVisualVM samples CPUs through "Sampler", and both can view memory trends and object allocations. 3. Locate thread problems. JVisualVM view state changes through "Threads" tag. JProfiler can detect deadlocks and display resource waiting conditions. 4. Combined with the external environment troubleshooting problems, view GC logs and application logs, and use JVM parameters to assist in analyzing performance bottlenecks.

The key to using Java Profilers such as JVisualVM or JProfiler is to understand their functional positioning and operational processes. Both types of tools can help you analyze application performance issues, but the focus is slightly different: JVisualVM is a free tool built into JDK, suitable for basic monitoring; while JProfiler has more comprehensive functions, suitable for in-depth diagnosis of performance bottlenecks.

1. Start and connect to the target application

No matter which tool it is, the first step is to start and connect to the Java application you want to analyze.

• JVisualVM : The way to open is very simple. Enter the bin directory of the JDK and run jvisualvm . It will automatically list the local running Java processes and click to select to connect.
• JProfiler : It needs to be installed separately. After the installation is completed, select "Remote Integration" through the GUI interface or directly attach to the local process.

Note: If it is an application on a remote server, some configuration is required to connect, such as turning on JMX or adding agent parameters. This process is common in production environments, but it is necessary to ensure that the firewall allows communication to the relevant ports.

2. Analyze CPU and memory usage

Once the connection is successful, you can start to view the application's running status.

• CPU Analysis :

  
  • In JProfiler, you can select "CPU Views", then select "Call Tree" or "Hot Spots" to quickly find the method that takes up the most CPU time.
  • JVisualVM needs to first click the "Sampler" tab, then enable CPU sampling, and then run for a period of time before stopping viewing the results.

• Memory analysis :

  • It mainly depends on the use trend of heap memory, whether there is frequent GC, and whether object creation is reasonable.
  • JProfiler provides the "Allocation Hot Spots" function, which can directly see which classes are constantly generating new objects.
  • JVisualVM can switch to the "Monitor" tab to view the heap memory usage, or do memory sampling to see the current distribution of surviving objects.

Suggestion: If you find that a certain method is called in a particularly large number, or some classes are instantiated very frequently, it may be the entry point for performance optimization.

3. Locate thread issues and deadlocks

Common threading problems in Java applications, such as deadlock, blocking, insufficient thread pool, etc.

• JVisualVM : Click the "Threads" tab to see the thread state change diagram. If the thread is in BLOCKED state for a long time, you can click "Thread Dump" to view the details.
• JProfiler : There is a dedicated "Threads/Thread Status" view, which can also detect potential deadlocks and show the situation where threads are waiting for resources.

Tips: When you suspect a deadlock, you can directly send a thread dump (thread dump) and then search the keyword "deadlock" in the tool.

4. Cooperate with the external environment to check

Many times, performance problems are not problems with the code itself, but are caused by factors such as slow database, network latency, and unreasonable GC settings.

• View GC situation: You can see the frequency and time-consuming of Full GC in both tools. If Full GC is frequently used, it means that there may be a problem with the memory setting.
• Combined with logs: Although profiler can provide a lot of information, sometimes it is necessary to cooperate with the application log to confirm which section of logic is wrong.
• Use JVM parameter assistance: For example, add -XX: PrintGCDetails to output GC logs to help determine whether GC is the bottleneck.

Basically that's it. By mastering these commonly used functions, most common performance problems can be located. Not complicated, but it is easy to ignore details.

The above is the detailed content of How to use a Java profiler like JVisualVM or JProfiler?. For more information, please follow other related articles on the PHP Chinese website!