JVM | garbage collection

JVM | garbage collection

Java garbage collection concept

• How to decide what is garbage objects, reference counting, root search algorithm
• What is GC Roots
  1. Reference JVM stack
  2. Methods district static variable references
  3. Reference the JNI (i.e., native method) of
• The method area (permanent behalf) GC occurs right, which objects will be recovered?
  1. The method of garbage collection zone to recover primarily of two parts: waste constants and useless class
  2. Will be recycled under the constant pool no object is referenced elsewhere in the GC, it is necessary: ​​Waste Constant
  3. Useless categories: useless class must meet three conditions
     • All the instance objects are recovered, the JVM class does not exist any instance
     • Load class ClassLoader has been recovered
     • Class Class object does not have any references, there is no place to refer to the class by reflection
  4. Virtual machine may be recovered based useless satisfy the above conditions, by -Xnoclassgcclosing the recovery of class (expansion read: the JVM Parameter Description )
  5. In heavy use of reflection, dynamic proxies, custom ClassLoader scene, there will be a large number of classes are created, you need to have JVM uninstall feature class.
• When minor GC and full GC happens
  • minorGC opportunity:
    1. When the JVM not to create a new object allocation space, will trigger minor GC, such as the newly created object is larger than the remaining space Eden District
    2. Will be called once before the full GC minor GC
    3. If you set the -XX:+CMSScavengeBeforeRemarkparameters, CMS will perform a minor GC before re-mark
  • fullGC opportunity:
    1. minor GC every time the object was promoted from the new generation to the old age of the average size of> the remaining space of old age
    2. After minor GC survival of the new generation of the total size of the object> the remaining space of old age
    3. Large objects directly into the old era, then the size of the object exceeds the remaining space years old
       • By -XX:PretenureSizeThreshold=xxxcan set the standard large object, more than this value, it will be put directly to the old time, pay attention this time is not a minor GC trigger
       • PretenureSizeThreshold parameters only for Serial and ParNew two collector effective, Parallel Scavenge collector does not know this parameter, Parallel Scavenge collector generally does not need to set. If you encounter a situation you must use this parameter, consider a collector combination ParNew plus the CMS
    4. Permanent band ( PermGen JDK7 ) or meta-space ( the Metadata jdk8 ) will be insufficient to trigger full GC
       • And with permanent element spaces are Hotspot implementation of the method area.
       • Meta JDK8 default maximum space to 21M, can -XX:MetaspaceSize=xxMset the size of the space element
       • View JVM initialization parameter default values:java -XX:+PrintFlagsInitial

Garbage collector

1583730329726.drawio.html 212.12 KB

Garbage collector

Safety and security of the region point

我对于安全点的理解,把GC的过程看成街道消毒的过程,用户线程就是在街上跑步前进的人,安全点就是街道上每隔一段距离设置的一个防空洞,街上的人会有三种状态:跑步(running),在防空洞等待消毒完成(block),在街外面等人(in native),每个人状态修改时,必须将自己的状态写在一个本子上(serialization page内存).消毒的时候,必须确保所有人都已经躲进地下的防空洞中了或者在街外等人.这要怎么做呢,首先我准备消毒的时候,会设置一个信号灯(sync_state),在街上跑步的人每路过一个防空洞,都要看一下这个预备消毒的信号灯有没有亮,如果亮了,就躲进防空洞里面,停止跑步,并修改自己的状态为阻塞(block),等消毒完成后在出来继续跑.在消毒的过程中,街道办事处就不允许人修改自己的状态了(设置serialization page只读).
安全区域就是一开始我在防空洞里面睡觉(线程调用sleep),或者我到街外面等人去了(等待IO),在睡觉或者等待的时候,我会把我的状态修改好(block或者in native),而且在我睡醒或者等到人了之后,在回到街道之前,我要去修改我的状态,如果当前街道还在消毒,办事处不会让我改状态的(serialization page只读),只有等到消毒完成,我才可以重新回到街道.

• 安全点的位置

  1. 方法返回前
  2. 调用方法之后
  3. 抛出异常的位置
  4. 循环的末尾
     • 如果for循环中有明确的循环计数器变量，而且该变量有明确的起始值、终止值、步进长度的循环,它有可能被优化为循环末尾没有safepoint，这种循环被称为'counted loop'于是如果这个循环的循环次数很多、循环体里又不调用别的方法或者是调用了方法但被内联进来了，就有可能会导致进入safepoint非常耗时。
     • 解决办法: 把单层循环拆成等价的双重嵌套循环，这样其中一层循环末尾的safepoint就可能会留下来，减少进入safepoint的等待时间。

• 为啥要在这四个地方设置安全点
  JVM要等所有应用线程进入到安全点后才会分派GC任务,如果有线程一直没有进入到安全点,就会导致GC的停顿时间延长.

• 安全点应用实战

  • 生产环境推荐使用-XX:+PrintGCApplicationStoppedTime打印JVM停顿时间,如果GC日志前面有较大的停顿,要就要考虑是不是代码里有较大的循环操作.
  • 在测试环境开启安全点统计日志

    -XX:+UnlockDiagnosticVMOptions
    -XX:+LogVMOutput
    -XX:LogFile=/dev/shm/vm.log
    -XX:+PrintSafepointStatistics
    -XX:PrintSafepointStatisticsCount=1

• 看看大神是怎么解决问题的 无法进入安全点导致GC停顿时间过长问题

• 推荐 JVM安全点介绍及实战

• 推荐 安全点图解

• 推荐 CMS垃圾收集器详细介绍

ID^[1]

---

1. 12466925 ↩