1. 概述
在本快速教程中，我们将介绍最常用的 Java 虚拟机配置选项。
2. 堆大小配置 – Xms 与 Xmx 选项
One of the most common performance-related practices is to initialize the heap memory as per the application requirements.
That's why we should specify minimal and maximal heap size. Below parameters can be used for achieving it:
-Xms<heap size>[unit]
-Xmx<heap size>[unit]
Here, unit denotes the unit in which the memory (indicated by heap size) is to be initialized. Units can be marked as ‘g' for GB, ‘m' for MB and ‘k' for KB.
For example, if we want to assign minimum 2 GB and maximum 5 GB to JVM, we need to write:
-Xms2G -Xmx5G
Starting with Java 8, the size of Metaspace is not defined. Once it reaches the global limit, JVM automatically increases it, However, to overcome any unnecessary instability, we can set Metaspace size with:
-XX:MaxMetaspaceSize=<metaspace size>[unit]
Here, metaspace size denotes the amount of memory we want to assign to Metaspace.
As per Oracle guidelines, after total available memory, the second most influential factor is the proportion of the heap reserved for the Young Generation. By default, the minimum size of the YG is 1310 MB, and maximum size is unlimited.
We can assign them explicitly:
-XX:NewSize=<young size>[unit]
-XX:MaxNewSize=<young size>[unit]
3. 垃圾回收器配置
选择正确的垃圾回收算法，对于提高应用程序稳定性至关重要。
JVM 有 4 种类型的 GC 实现 :
- Serial 垃圾收集器
- Parallel 垃圾收集器
- CMS 垃圾收集器
- G1 垃圾收集器
可以使用如下参数进行配置：
-XX:+UseSerialGC
-XX:+UseParallelGC
-XX:+USeParNewGC
-XX:+UseG1GC
更多关于垃圾回收 实现细节，可以访问这篇文章。
4. GC 日志配置
To strictly monitor the application health, we should always check the JVM's Garbage Collection performance. The easiest way to do this is to log the GC activity in human readable format.
Using the following parameters, we can log the GC activity:
-XX:+UseGCLogFileRotation
-XX:NumberOfGCLogFiles=< number of log files >
-XX:GCLogFileSize=< file size >[ unit ]
-Xloggc:/path/to/gc.log
UseGCLogFileRotation specifies the log file rolling policy, much like log4j, s4lj, etc. NumberOfGCLogFiles denotes the max number of log files that can be written for a single application life cycle. GCLogFileSize specifies the max size of the file. Finally, loggc denotes its location.
Point to note here is that, there are two more JVM parameters available (-XX:+PrintGCTimeStamps and -XX:+PrintGCDateStamps) which can be used to print date-wise timestamp in the GC log.
For example, if we want to assign a maximum of 100 GC log files, each having a maximum size of 50 MB and want to store them in ‘*/home/user/log/'* location, we can use below syntax:
-XX:+UseGCLogFileRotation
-XX:NumberOfGCLogFiles=10
-XX:GCLogFileSize=50M
-Xloggc:/home/user/log/gc.log
However, the problem is that one additional daemon thread is always used for monitoring system time in the background. This behavior may create some performance bottleneck; that's why it's always better not to play with this parameter in production.
5. 处理内存溢出
It's very common for a large application to face out of memory error which, in turn, results in the application crash. It's a very critical scenario and very hard to replicate to troubleshoot the issue.
That's why JVM comes with some parameters which dump heap memory into a physical file which can be used later for finding out leaks:
-XX:+HeapDumpOnOutOfMemoryError
-XX:HeapDumpPath=./java_pid<pid>.hprof
-XX:OnOutOfMemoryError="< cmd args >;< cmd args >"
-XX:+UseGCOverheadLimit
A couple of points to note here:
- HeapDumpOnOutOfMemoryError instructs the JVM to dump heap into physical file in case of OutOfMemoryError
- HeapDumpPath denotes the path where the file is to be written; any filename can be given; however, if JVM finds a
<pid>tag in the name, the process id of the current process causing the out of memory error will be appended to the file name with .hprof format - OnOutOfMemoryError is used to issue emergency commands to be executed in case of out of memory error; proper command should be used in the space of cmd args. For example, if we want to restart the server as soon as out of memory occur, we can set the parameter:
-XX:OnOutOfMemoryError="shutdown -r"
- UseGCOverheadLimit is a policy that limits the proportion of the VM's time that is spent in GC before an OutOfMemory error is thrown
6. 32/64 位
In the OS environment where both 32 and 64-bit packages are installed, the JVM automatically chooses 32-bit environmental packages.
If we want to set the environment to 64 bit manually, we can do so using below parameter:
-d<OS bit>
OS bit can be either 32 or 64. More information about this can be found here.
7. 其他配置
- -server: enables “Server Hotspot VM”; this parameter is used by default in 64 bit JVM
- -XX:+UseStringDeduplication: Java 8u20 has introduced this JVM parameter for reducing the unnecessary use of memory by creating too many instances of the same String; this optimizes the heap memory by reducing duplicate String values to a single global char[] array
- -XX:+UseLWPSynchronization: sets LWP (Light Weight Process) – based synchronization policy instead of thread-based synchronization
- -XX:LargePageSizeInBytes: sets the large page size used for the Java heap; it takes the argument in GB/MB/KB; with larger page sizes we can make better use of virtual memory hardware resources; however, this may cause larger space sizes for the PermGen, which in turn can force to reduce the size of Java heap space
- -XX:MaxHeapFreeRatio: sets the maximum percentage of heap free after GC to avoid shrinking.
- -XX:MinHeapFreeRatio: sets the minimum percentage of heap free after GC to avoid expansion; to monitor the heap usage you can use VisualVM shipped with JDK.
- -XX:SurvivorRatio: Ratio of eden/survivor space size – for example, -XX:SurvivorRatio=6 sets the ratio between each survivor space and eden space to be 1:6,
- -XX:+UseLargePages: use large page memory if it is supported by the system; please note that OpenJDK 7 tends to crash if using this JVM parameter
- -XX:+UseStringCache: enables caching of commonly allocated strings available in the String pool
- -XX:+UseCompressedStrings: use a byte[] type for String objects which can be represented in pure ASCII format
- -XX:+OptimizeStringConcat: it optimizes String concatenation operations where possible
8. 总结
In this quick article, we learned about some important JVM parameters – which can be used to tune and improve general application performance.
Some of these can also be used for debugging purposes.
If you want to explore the reference parameters in more detail, you can get started here.