最近两天测试环境有一个服务总是会挂（两到三天一次），JVM虚拟机总是会崩溃。所以有必要了解JVM崩溃的原因是什么。
当JVM发生致命错误导致崩溃时，会生成一个hs_err_pid_xxx.log这样的文件，该文件包含了导致 JVM crash 的重要信息，我们可以通过分析该文件定位到导致 JVM Crash 的原因，从而修复保证系统稳定。
默认情况下，该文件是生成在工作目录下的，当然也可以通过 JVM 参数指定生成路径：
-XX:ErrorFile=/var/log/hs_err_pid<pid>.log
这个文件主要包含如下内容：
- 日志头文件
- 导致 crash 的线程信息
- 所有线程信息
- 安全点和锁信息
- 堆信息
- 本地代码缓存
- 编译事件
- gc 相关记录
- jvm 内存映射
- jvm 启动参数
- 服务器信息
下面就根据这个文件内容逐步解析。
日志头文件
内容如下
#
# A fatal error has been detected by the Java Runtime Environment:
#
# SIGSEGV (0xb) at pc=0x0000003797807a91, pid=29071, tid=139901421901568
#
# JRE version: Java(TM) SE Runtime Environment (8.0_45-b14) (build 1.8.0_45-b14)
# Java VM: Java HotSpot(TM) 64-Bit Server VM (25.45-b02 mixed mode linux-amd64 compressed oops)
# Problematic frame:
# C [libresolv.so.2+0x7a91] __libc_res_nquery+0x1c1
#
# Failed to write core dump. Core dumps have been disabled. To enable core dumping, try "ulimit -c unlimited" before starting Java again
#
# If you would like to submit a bug report, please visit:
# http://bugreport.java.com/bugreport/crash.jsp
# The crash happened outside the Java Virtual Machine in native code.
# See problematic frame for where to report the bug.
#
这段内容主要简述了导致 JVM Crash 的原因。常见的原因有 JVM 自身的 bug，应用程序错误，JVM 参数，服务器资源不足，JNI 调用错误等。当然还有一些版本和配置信息，
SIGSEGV (0xb) at pc=0x0000003797807a91, pid=29071, tid=139901421901568
非预期的错误被 JRE 检测到了，其中
- SIGSEGV ：信号量
- 0xb ：信号码
- pc=0x0000003797807a91 ：程序计数器的值
- pid=29071 ：进程号
- tid=139901421901568 ：线程号
SIGSEGV(0xb) 表示 JVM Crash 时正在执行 JNI 代码，常见的描述还有EXCEPTION_ACCESS_VIOLATION，该描述表示 JVM Crash 时正在执行 JVM 自身的代码，这往往是因为 JVM 的 Bug 导致的 Crash；另一种常见的描述是EXCEPTION_STACK_OVERFLOW，该描述表示这是个栈溢出导致的错误，这往往是应用程序中存在深层递归导致的。
# JRE version: Java(TM) SE Runtime Environment (8.0_45-b14) (build 1.8.0_45-b14)
# Java VM: Java HotSpot(TM) 64-Bit Server VM (25.45-b02 mixed mode linux-amd64 compressed oops)
JRE 和 JVM 的版本信息
# Problematic frame:
# C [libresolv.so.2+0x7a91] __libc_res_nquery+0x1c1
这个信息比较重要，问题帧信息：
C 表示帧类型为本地帧，还有其他类型：
- j ： 解释的Java帧
- V : 虚拟机帧
- v ：虚拟机生成的存根栈帧
- J：其他帧类型，包括编译后的Java帧
[libresolv.so.2+0x7a91] __libc_res_nquery+0x1c1和程序计数器（pc）表达的含义一样，但是用的是本地so库+偏移量的方式。
导致 crash 的线程信息
内容如下：
--------------- T H R E A D ---------------
Current thread (0x0000000001e94800): JavaThread "pool-1-thread-2" [_thread_in_native, id=30111, stack(0x00007f3d567e5000,0x00007f3d568e6000)]
siginfo: si_signo: 11 (SIGSEGV), si_code: 1 (SEGV_MAPERR), si_addr: 0x0000000000000003
Registers:
RAX=0x0000000000000000, RBX=0x0000000000000000, RCX=0x0000000000000000, RDX=0x0000000000000050
RSP=0x00007f3d568e2280, RBP=0x00007f3d568e2570, RSI=0x0000000000000000, RDI=0x00000000ffffffff
R8 =0x0000000000000000, R9 =0x0000000000000000, R10=0x000000000007a337, R11=0x0000000000000213
R12=0x00007f3d568e2ef0, R13=0x00007f3d568e22b0, R14=0x0000000000000000, R15=0x00007f3d568e5db8
RIP=0x0000003797807a91, EFLAGS=0x0000000000010246, CSGSFS=0x0000000000000033, ERR=0x0000000000000004
TRAPNO=0x000000000000000e
Top of Stack: (sp=0x00007f3d568e2280)
0x00007f3d568e2280: b8e4bfb900000800 00007f3d568e3760
0x00007f3d568e2290: 00007f3d568e3758 00007f3d568e377c
0x00007f3d568e22a0: 00007f3d568e3778 6f6e6b6e56a88a58
0x00007f3d568e22b0: 00000100000149a0 7a68710800000000
0x00007f3d568e22c0: 6970067363642d78 6d6f63036e61676e
....省略
Instructions: (pc=0x0000003797807a91)
0x0000003797807a71: 48 89 45 b8 48 8b 4d b8 0f b6 51 03 89 d3 83 e3
0x0000003797807a81: 0f 75 0d 0f b7 49 06 66 c1 c9 08 66 85 c9 75 4f
0x0000003797807a91: 0f b6 48 03 bf 0f 00 00 00 40 20 cf 75 0d 0f b7
0x0000003797807aa1: 70 06 66 c1 ce 08 66 85 f6 75 34 83 e1 0f 83 e2
Register to memory mapping:
RAX=0x0000000000000000 is an unknown value
RBX=0x0000000000000000 is an unknown value
RCX=0x0000000000000000 is an unknown value
RDX=0x0000000000000050 is an unknown value
RSP=0x00007f3d568e2280 is pointing into the stack for thread: 0x0000000001e94800
RBP=0x00007f3d568e2570 is pointing into the stack for thread: 0x0000000001e94800
... 省略
Stack: [0x00007f3d567e5000,0x00007f3d568e6000], sp=0x00007f3d568e2280, free space=1012k
Native frames: (J=compiled Java code, j=interpreted, Vv=VM code, C=native code)
C [libresolv.so.2+0x7a91] __libc_res_nquery+0x1c1
C [libresolv.so.2+0x7fd1]
Java frames: (J=compiled Java code, j=interpreted, Vv=VM code)
J 15056 java.net.Inet6AddressImpl.lookupAllHostAddr(Ljava/lang/String;)[Ljava/net/InetAddress; (0 bytes) @ 0x00007f3d7492af8c [0x00007f3d7492af40+0x4c]
J 14966 C1 java.net.InetAddress.getAddressesFromNameService(Ljava/lang/String;Ljava/net/InetAddress;)[Ljava/net/InetAddress; (245 bytes) @ 0x00007f3d75466754 [0x00007f3d754662c0+0x494]
J 14291 C2 java.net.InetAddress.getAllByName(Ljava/lang/String;Ljava/net/InetAddress;)[Ljava/net/InetAddress; (387 bytes) @ 0x00007f3d7534b718 [0x00007f3d7534ae20+0x8f8]
J 14178 C1 java.net.InetSocketAddress.<init>(Ljava/lang/String;I)V (47 bytes) @ 0x00007f3d752ce0f4 [0x00007f3d752cdec0+0x234]
j sun.security.ssl.SSLSocketImpl.<init>(Lsun/security/ssl/SSLContextImpl;Ljava/lang/String;ILjava/net/InetAddress;I)V+144
j sun.security.ssl.SSLSocketFactoryImpl.createSocket(Ljava/lang/String;ILjava/net/InetAddress;I)Ljava/net/Socket;+13
j com.ufclub.daq.qhzx.utils.SSLProtocolSocketFactory.createSocket(Ljava/lang/String;ILjava/net/InetAddress;I)Ljava/net/Socket;+15
.... 省略代码
这部分内容包含出发 JVM 致命错误的线程详细信息和线程栈。
线程信息
Current thread (0x0000000001e94800): JavaThread "pool-1-thread-2" [_thread_in_native, id=30111, stack(0x00007f3d567e5000,0x00007f3d568e6000)]
- 0x0000000001e94800：出错的线程指针
- JavaThread：线程类型，可能的类型包括
- JavaThread：Java线程
- VMThread : JVM 的内部线程
- CompilerThread：用来调用JITing，实时编译装卸class 。 通常，jvm会启动多个线程来处理这部分工作，线程名称后面的数字也会累加，例如：CompilerThread1
- GCTaskThread：执行gc的线程
- WatcherThread：JVM 周期性任务调度的线程，是一个单例对象
- ConcurrentMarkSweepThread：jvm在进行CMS GC的时候，会创建一个该线程去进行GC，该线程被创建的同时会创建一个SurrogateLockerThread（简称SLT）线程并且启动它，SLT启动之后，处于等待阶段。CMST开始GC时，会发一个消息给SLT让它去获取Java层Reference对象的全局锁：Lock
- pool-1-thread-2：线程名称
- _thread_in_native：当前线程状态。该描述还包含有：
- _thread_in_native：线程当前状态，状态枚举包括：
- _thread_uninitialized：线程还没有创建，它只在内存原因崩溃的时候才出现
- _thread_new：线程已经被创建，但是还没有启动
- _thread_in_native：线程正在执行本地代码，一般这种情况很可能是本地代码有问题
- _thread_in_vm：线程正在执行虚拟机代码
- _thread_in_Java：线程正在执行解释或者编译后的Java代码
- _thread_blocked：线程处于阻塞状态
- …_trans：以_trans结尾，线程正处于要切换到其它状态的中间状态
- id=30111：线程ID
- stack(0x00007f3d567e5000,0x00007f3d568e6000)：栈区间
siginfo: si_signo: 11 (SIGSEGV), si_code: 1 (SEGV_MAPERR), si_addr: 0x0000000000000003
表示导致虚拟机终止的非预期的信号信息。
Top of Stack: (sp=0x00007f3d568e2280)
0x00007f3d568e2280: b8e4bfb900000800 00007f3d568e3760
0x00007f3d568e2290: 00007f3d568e3758 00007f3d568e377c
0x00007f3d568e22a0: 00007f3d568e3778 6f6e6b6e56a88a58
0x00007f3d568e22b0: 00000100000149a0 7a68710800000000
0x00007f3d568e22c0: 6970067363642d78 6d6f63036e61676e
....省略
Instructions: (pc=0x0000003797807a91)
0x0000003797807a71: 48 89 45 b8 48 8b 4d b8 0f b6 51 03 89 d3 83 e3
0x0000003797807a81: 0f 75 0d 0f b7 49 06 66 c1 c9 08 66 85 c9 75 4f
0x0000003797807a91: 0f b6 48 03 bf 0f 00 00 00 40 20 cf 75 0d 0f b7
0x0000003797807aa1: 70 06 66 c1 ce 08 66 85 f6 75 34 83 e1 0f 83 e2
栈顶程序计数器旁的操作码，它们可以被反汇编成系统崩溃前执行的指令。
Stack: [0x00007f3d567e5000,0x00007f3d568e6000], sp=0x00007f3d568e2280, free space=1012k
Native frames: (J=compiled Java code, j=interpreted, Vv=VM code, C=native code)
C [libresolv.so.2+0x7a91] __libc_res_nquery+0x1c1
C [libresolv.so.2+0x7fd1]
Java frames: (J=compiled Java code, j=interpreted, Vv=VM code)
J 15056 java.net.Inet6AddressImpl.lookupAllHostAddr(Ljava/lang/String;)[Ljava/net/InetAddress; (0 bytes) @ 0x00007f3d7492af8c [0x00007f3d7492af40+0x4c]
J 14966 C1 java.net.InetAddress.getAddressesFromNameService(Ljava/lang/String;Ljava/net/InetAddress;)[Ljava/net/InetAddress; (245 bytes) @ 0x00007f3d75466754 [0x00007f3d754662c0+0x494]
J 14291 C2 java.net.InetAddress.getAllByName(Ljava/lang/String;Ljava/net/InetAddress;)[Ljava/net/InetAddress; (387 bytes) @ 0x00007f3d7534b718 [0x00007f3d7534ae20+0x8f8]
J 14178 C1 java.net.InetSocketAddress.<init>(Ljava/lang/String;I)V (47 bytes) @ 0x00007f3d752ce0f4 [0x00007f3d752cdec0+0x234]
j sun.security.ssl.SSLSocketImpl.<init>(Lsun/security/ssl/SSLContextImpl;Ljava/lang/String;ILjava/net/InetAddress;I)V+144
j sun.security.ssl.SSLSocketFactoryImpl.createSocket(Ljava/lang/String;ILjava/net/InetAddress;I)Ljava/net/Socket;+13
j com.ufclub.daq.qhzx.utils.SSLProtocolSocketFactory.createSocket(Ljava/lang/String;ILjava/net/InetAddress;I)Ljava/net/Socket;+15
.... 省略代码
线程栈信息。包含了地址、栈顶、栈计数器和线程尚未使用的栈信息。到这里就基本上已经确定了问题所在原因了。
线程信息
Java Threads: ( => current thread )
0x00007f3d5d0a0800 JavaThread "logback-8" daemon [_thread_blocked, id=489, stack(0x00007f3d56de7000,0x00007f3d56ee8000)]
0x00007f3d5d09f800 JavaThread "logback-7" daemon [_thread_blocked, id=30974, stack(0x00007f3d53fc3000,0x00007f3d540c4000)]
0x00007f3d5c47f800 JavaThread "logback-6" daemon [_thread_blocked, id=25662, stack(0x00007f3d545c9000,0x00007f3d546ca000)]
0x00007f3d5c2a4000 JavaThread "logback-5" daemon [_thread_blocked, id=20922, stack(0x00007f3d552d2000,0x00007f3d553d3000)]
0x0000000003291800 JavaThread "logback-4" daemon [_thread_blocked, id=16768, stack(0x00007f3d542c6000,0x00007f3d543c7000)]
0x0000000002320000 JavaThread "logback-3" daemon [_thread_blocked, id=14730, stack(0x00007f3d54bcd000,0x00007f3d54cce000)]
0x0000000002d05000 JavaThread "logback-2" daemon [_thread_blocked, id=6569, stack(0x00007f3d549cb000,0x00007f3d54acc000)]
所有线程信息，一目了然。_thread_blocked表示阻塞。
安全点和锁信息
VM state:not at safepoint (normal execution)
虚拟机状态。not at safepoint 表示正常运行。其余状态：
- at safepoint：所有线程都因为虚拟机等待状态而阻塞，等待一个虚拟机操作完成；
- synchronizing：一个特殊的虚拟机操作，要求虚拟机内的其它线程保持等待状态。
VM Mutex/Monitor currently owned by a thread: None
虚拟机的 Mutex 和 Monito r目前没有被线程持有。Mutex 是虚拟机内部的锁，而 Monitor 则关联到了 Java 对象。
堆信息
Heap:
PSYoungGen total 178688K, used 25522K [0x00000000eab00000, 0x00000000f8d80000, 0x0000000100000000)
eden space 177664K, 13% used [0x00000000eab00000,0x00000000ec343d30,0x00000000f5880000)
from space 1024K, 65% used [0x00000000f8c80000,0x00000000f8d28d20,0x00000000f8d80000)
to space 1024K, 0% used [0x00000000f8b80000,0x00000000f8b80000,0x00000000f8c80000)
ParOldGen total 360448K, used 47193K [0x00000000c0000000, 0x00000000d6000000, 0x00000000eab00000)
object space 360448K, 13% used [0x00000000c0000000,0x00000000c2e16518,0x00000000d6000000)
Metaspace used 79300K, capacity 80628K, committed 80936K, reserved 1120256K
class space used 9401K, capacity 9645K, committed 9768K, reserved 1048576K
Card table byte_map: [0x00007f3d729f1000,0x00007f3d72bf2000] byte_map_base: 0x00007f3d723f1000
新生代、老年代、元空间一目了然。
Card table表示一种卡表，是 jvm 维护的一种数据结构，用于记录更改对象时的引用，以便 gc 时遍历更少的 table 和 root。
本地代码缓存
CodeCache: size=245760Kb used=41374Kb max_used=41402Kb free=204385Kb
bounds [0x00007f3d72fb2000, 0x00007f3d75872000, 0x00007f3d81fb2000]
total_blobs=12767 nmethods=12130 adapters=549
compilation: enabled
一块用于编译和保存本地代码的内存。
编译事件
Compilation events (10 events):
Event: 501041.566 Thread 0x0000000001b6e000 16334 3 sun.security.rsa.RSAKeyFactory::<init> (5 bytes)
Event: 501041.566 Thread 0x0000000001b6e000 nmethod 16334 0x00007f3d74985790 code [0x00007f3d74985900, 0x00007f3d74985b10]
Event: 501041.569 Thread 0x0000000001b6e000 16335 3 sun.security.pkcs.PKCS8Key::<init> (5 bytes)
Event: 501041.570 Thread 0x0000000001b6e000 nmethod 16335 0x00007f3d74736290 code [0x00007f3d74736400, 0x00007f3d747365b0]
Event: 501041.575 Thread 0x0000000001b6e000 16336 3 sun.security.ssl.BaseSSLSocketImpl::<init> (15 bytes)
Event: 501041.576 Thread 0x0000000001b6e000 nmethod 16336 0x00007f3d73f9b450 code [0x00007f3d73f9b5c0, 0x00007f3d73f9b7e8]
Event: 501041.578 Thread 0x0000000001b6e000 16337 3 javax.net.ssl.SSLSocket::<init> (5 bytes)
Event: 501041.580 Thread 0x0000000001b6e000 nmethod 16337 0x00007f3d739c7210 code [0x00007f3d739c7380, 0x00007f3d739c7508]
Event: 501041.582 Thread 0x0000000001b6e000 16338 3 javax.net.ssl.SNIServerName::<init> (66 bytes)
Event: 501041.583 Thread 0x0000000001b6e000 nmethod 16338 0x00007f3d74c15cd0 code [0x00007f3d74c15ea0, 0x00007f3d74c164f8]
记录10次编译事件。这里的信息也印证了上面的结论。
gc 日志
GC Heap History (10 events):
Event: 476166.948 GC heap before
{Heap before GC invocations=149 (full 3):
PSYoungGen total 194560K, used 193984K [0x00000000eab00000, 0x00000000f8e80000, 0x0000000100000000)
eden space 193536K, 100% used [0x00000000eab00000,0x00000000f6800000,0x00000000f6800000)
from space 1024K, 43% used [0x00000000f8c80000,0x00000000f8cf0000,0x00000000f8d80000)
to space 1024K, 0% used [0x00000000f8d80000,0x00000000f8d80000,0x00000000f8e80000)
ParOldGen total 360448K, used 47161K [0x00000000c0000000, 0x00000000d6000000, 0x00000000eab00000)
object space 360448K, 13% used [0x00000000c0000000,0x00000000c2e0e518,0x00000000d6000000)
Metaspace used 79243K, capacity 80500K, committed 80680K, reserved 1120256K
class space used 9400K, capacity 9645K, committed 9768K, reserved 1048576K
Event: 476166.985 GC heap after
Heap after GC invocations=149 (full 3):
PSYoungGen total 190464K, used 448K [0x00000000eab00000, 0x00000000f8e00000, 0x0000000100000000)
eden space 189952K, 0% used [0x00000000eab00000,0x00000000eab00000,0x00000000f6480000)
from space 512K, 87% used [0x00000000f8d80000,0x00000000f8df0000,0x00000000f8e00000)
to space 1024K, 0% used [0x00000000f8c00000,0x00000000f8c00000,0x00000000f8d00000)
ParOldGen total 360448K, used 47161K [0x00000000c0000000, 0x00000000d6000000, 0x00000000eab00000)
object space 360448K, 13% used [0x00000000c0000000,0x00000000c2e0e518,0x00000000d6000000)
Metaspace used 79243K, capacity 80500K, committed 80680K, reserved 1120256K
class space used 9400K, capacity 9645K, committed 9768K, reserved 1048576K
}
... 省略
同样是记录10次 GC。
jvm 内存映射
Dynamic libraries:
00400000-00401000 r-xp 00000000 fd:00 2108521 /usr/java/jdk1.8.0_45/bin/java
00600000-00601000 rw-p 00000000 fd:00 2108521 /usr/java/jdk1.8.0_45/bin/java
019e9000-04f5e000 rw-p 00000000 00:00 0 [heap]
c0000000-d6000000 rw-p 00000000 00:00 0
d6000000-eab00000 ---p 00000000 00:00 0
...
7f3d6df48000-7f3d6df6a000 r--s 0038e000 fd:00 2108900 /usr/java/jdk1.8.0_45/jre/lib/ext/cldrdata.jar
7f3d6df6a000-7f3d6df73000 r--s 07db3000 fd:00 2374798 /opt/risk/service/xxx-xxx-container/xxx-xxxx-container.jar
...
这些信息是虚拟机崩溃时的虚拟内存列表区域。它可以告诉你崩溃原因时哪些类库正在被使用，位置在哪里，还有堆栈和守护页信息。
- 00400000-00401000：内存区域
- r-xp：权限，r/w/x/p/s分别表示读/写/执行/私有/共享
- 00000000：文件内的偏移量
- fd:00：文件位置的majorID和minorID
- 2108521：索引节点号
- /usr/java/jdk1.8.0_45/bin/java：文件位置
从/opt/risk/service/xxx-xxx-container/xxx-xxxx-container.jar我们可以确认是那个jar出问题了。
jvm 启动参数
VM Arguments:
jvm_args: -Xmx1024M -Xms512M -XX:PermSize=128M -XX:MaxPermSize=256M
java_command: /opt/risk/service/xxx-xxx-xxx-container/xxx-xxx-xxx-container.jar
java_class_path (initial): /opt/risk/service/xxx-xxx-xxx-container/xxx-xxx-xxx-container.jar
Launcher Type: SUN_STANDARD
Environment Variables:
JAVA_HOME=/usr/java/jdk1.8.0_45
CLASSPATH=.:/usr/java/jdk1.8.0_45/lib/dt.jar:/usr/java/jdk1.8.0_45/lib/tools.jar
PATH=/usr/java/jdk1.8.0_45/bin:/bin:/usr/local/erlang/bin:/usr/local/maven3/bin:/usr/local/git/bin:/usr/java/jdk1.8.0_45/bin:/bin:/usr/local/erlang/bin:/usr/local/maven3/bin:/usr/local/git/bin:/usr/lib64/qt-3.3/bin:/usr/local/sbin:/usr/bin:/bin:/usr/sbin:/sbin:/root/bin
SHELL=/bin/bash
jvm 虚拟机参数和环境变量。
