Java項目一般使用jstack, jmap等Java工具都可以分析進程bug，但是有時候jvm自身異常退出，此時往往沒有hprof文件，而是生成了core文件，特別是在使用JNI技術的項目中，這時候就要用充分挖掘core文件中的有用信息了。

core dump介紹

當程序運行的過程中異常終止或崩潰，操作系統(tǒng)會將程序當時的內(nèi)存狀態(tài)記錄下來，保存在一個文件中，這種行為就叫做Core Dump（中文有的翻譯成“核心轉(zhuǎn)儲”)。我們可以認為 core dump 是“內(nèi)存快照”，但實際上，除了內(nèi)存信息之外，還有些關鍵的程序運行狀態(tài)也會同時 dump 下來，例如寄存器信息（包括程序指針、棧指針等）、內(nèi)存管理信息、其他處理器和操作系統(tǒng)狀態(tài)和信息。core dump 對于編程人員診斷和調(diào)試程序是非常有幫助的，因為對于有些程序錯誤是很難重現(xiàn)的，例如指針異常，而 core dump 文件可以再現(xiàn)程序出錯時的情景。

生成core dump

由于 core dump 文件會占據(jù)一定的磁盤空間，默認情況下，Linux 不允許生成 core dump 文件。例如，下面的命令顯示，Linux 允許的最大 core dump 文件大小為 0：

$ ulimit -a | grep core core file size (blocks, -c) 0

可以通過下面設置，允許 Linux 生成 core dump 文件：

$ ulimit -c unlimited

注意到，這個設置只對當前登錄回話有效。如果想要這個設置持久有效，可以把它寫入到/etc/security/limits.conf文件中：

$ sudo vi /etc/security/limits.conf * soft core unlimited * soft hard unlimited

gdb

gdb $JAV{A}_{H}OME$/bin/java core-26492
進入gdb上下文后再執(zhí)行bt命令，可以看到進程退出時的線程棧。

$ gdb /data/soft/jdk/bin/java core.11668 GNU gdb (GDB) Red Hat Enterprise Linux 7.6.1-80.el7 Copyright (C) 2013 Free Software Foundation, Inc. License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html> This is free software: you are free to change and redistribute it. There is NO WARRANTY, to the extent permitted by law. Type "show copying" and "show warranty" for details. This GDB was configured as "x86_64-redhat-linux-gnu". For bug reporting instructions, please see: <http://www.gnu.org/software/gdb/bugs/>... Reading symbols from /data/soft/jdk1.8.0_141/bin/java...Missing separate debuginfo for /data/soft/jdk1.8.0_141/bin/java Try: yum --enablerepo='*debug*' install /usr/lib/debug/.build-id/c9/0f19ee0af98c47ccaa7181853cfd14867bc931.debug (no debugging symbols found)...done. [New LWP 14367] ... [New LWP 14355] [New LWP 11685] [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib64/libthread_db.so.1". Missing separate debuginfo for /data/soft/jdk1.8.0_141/jre/lib/amd64/server/libjvm.so Try: yum --enablerepo='*debug*' install /usr/lib/debug/.build-id/43/a2bc419314aa566cd5ba54779ae18b47022cad.debug Missing separate debuginfo for /data/soft/jdk1.8.0_141/jre/lib/amd64/libverify.so Try: yum --enablerepo='*debug*' install /usr/lib/debug/.build-id/b3/de2701fa3aef182d8e7b2db8d294a91af1eb7c.debug Missing separate debuginfo for /data/soft/jdk1.8.0_141/jre/lib/amd64/libmanagement.so Try: yum --enablerepo='*debug*' install /usr/lib/debug/.build-id/ee/7e746ef83f2c2fbe8ea0cb42706b63cd74de3f.debug Missing separate debuginfo for /data/soft/jdk1.8.0_141/jre/lib/amd64/libinstrument.so Try: yum --enablerepo='*debug*' install /usr/lib/debug/.build-id/13/552fe85b2af551ab806735bc5e5f2a5e00fbb1.debug Core was generated by `/data/soft/jdk/bin/java -Djava.util.logging.config.file=/data/home/aics/taskbot'. Program terminated with signal 6, Aborted. #0 0x00007f50d652b1d7 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56 56 return INLINE_SYSCALL (tgkill, 3, pid, selftid, sig); (gdb) bt #0 0x00007f50d652b1d7 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56 #1 0x00007f50d652c8c8 in __GI_abort () at abort.c:90 #2 0x00007f50397417f4 in tensorflow::internal::LogMessageFatal::~LogMessageFatal() () from /data/home/aics/taskbot-tomcat/temp/tensorflow_native_libraries-1545395060637-0/libtensorflow_framework.so #3 0x00007f5039721cf0 in tensorflow::monitoring::CollectionRegistry::Register(tensorflow::monitoring::AbstractMetricDef const*, std::function<void (tensorflow::monitoring::MetricCollectorGetter)> const&) ()from /data/home/aics/taskbot-tomcat/temp/tensorflow_native_libraries-1545395060637-0/libtensorflow_framework.so #4 0x00007f501520951f in ?? () from /data/home/aics/taskbot-tomcat/temp/tensorflow_native_libraries-1545633543384-0/libtensorflow_jni.so #5 0x00007f50151334db in ?? () from /data/home/aics/taskbot-tomcat/temp/tensorflow_native_libraries-1545633543384-0/libtensorflow_jni.so #6 0x00007f50d6efd1e3 in call_init (env=0x7f50d00c4460, argv=0x7ffe7bc01488, argc=21, l=<optimized out>) at dl-init.c:82 #7 _dl_init (main_map=main_map@entry=0x7f50249fa900, argc=21, argv=0x7ffe7bc01488, env=0x7f50d00c4460) at dl-init.c:131 #8 0x00007f50d6f018f6 in dl_open_worker (a=a@entry=0x7f503b2f7b98) at dl-open.c:560 #9 0x00007f50d6efcff4 in _dl_catch_error (objname=objname@entry=0x7f503b2f7b88, errstring=errstring@entry=0x7f503b2f7b90, mallocedp=mallocedp@entry=0x7f503b2f7b80, operate=operate@entry=0x7f50d6f01540 <dl_open_worker>, args=args@entry=0x7f503b2f7b98) at dl-error.c:177 #10 0x00007f50d6f00feb in _dl_open (file=0x7f5024008550 "/data/home/aics/taskbot-tomcat/temp/tensorflow_native_libraries-1545633543384-0/libtensorflow_jni.so", mode=-2147483647, caller_dlopen=<optimized out>, nsid=-2, argc=21, argv=0x7ffe7bc01488, env=0x7f50d00c4460) at dl-open.c:650 #11 0x00007f50d68b8fbb in dlopen_doit (a=a@entry=0x7f503b2f7da0) at dlopen.c:66 #12 0x00007f50d6efcff4 in _dl_catch_error (objname=0x7f5024ae73e0, errstring=0x7f5024ae73e8, mallocedp=0x7f5024ae73d8, operate=0x7f50d68b8f60 <dlopen_doit>, args=0x7f503b2f7da0) at dl-error.c:177 #13 0x00007f50d68b95bd in _dlerror_run (operate=operate@entry=0x7f50d68b8f60 <dlopen_doit>, args=args@entry=0x7f503b2f7da0) at dlerror.c:163 #14 0x00007f50d68b9051 in __dlopen (file=<optimized out>, mode=<optimized out>) at dlopen.c:87 #15 0x00007f50d5e2a24e in os::dll_load(char const*, char*, int) () from /data/soft/jdk1.8.0_141/jre/lib/amd64/server/libjvm.so #16 0x00007f50d5c2198c in JVM_LoadLibrary () from /data/soft/jdk1.8.0_141/jre/lib/amd64/server/libjvm.so #17 0x00007f50d4ac7df8 in Java_java_lang_ClassLoader_00024NativeLibrary_load () from /data/soft/jdk1.8.0_141/jre/lib/amd64/libjava.so

jstack

之前從來沒想過，居然還可以使用java工具來分析core文件

使用jstack可以查看到進程core_dump時的線程執(zhí)行棧
/data/soft/jdk/bin/jstack /data/soft/jdk/bin/java core.26492

Thread 15040: (state = IN_NATIVE)- java.lang.ClassLoader$NativeLibrary.load(java.lang.String, boolean) @bci=0 (Interpreted frame)- java.lang.ClassLoader.loadLibrary0(java.lang.Class, java.io.File) @bci=328, line=1941 (Interpreted frame)- java.lang.ClassLoader.loadLibrary(java.lang.Class, java.lang.String, boolean) @bci=48, line=1824 (Interpreted frame)- java.lang.Runtime.load0(java.lang.Class, java.lang.String) @bci=57, line=809 (Interpreted frame)- java.lang.System.load(java.lang.String) @bci=7, line=1086 (Interpreted frame)- org.tensorflow.NativeLibrary.load() @bci=212, line=101 (Interpreted frame)- org.tensorflow.TensorFlow.init() @bci=0, line=66 (Interpreted frame)- org.tensorflow.TensorFlow.<clinit>() @bci=0, line=70 (Interpreted frame)- org.tensorflow.Graph.<clinit>() @bci=0, line=258 (Interpreted frame)

jmap

$ /data/soft/jdk/bin/jmap /data/soft/jdk/bin/java core.14652
結(jié)果如下：

Attaching to core core.14652 from executable /data/soft/jdk/bin/java, please wait... Debugger attached successfully. Server compiler detected. JVM version is 25.141-b15 0x0000000000400000 7K /data/soft/jdk/bin/java 0x00007fbc261fb000 41K /lib64/libonion.so 0x00007fbc25ec8000 139K /lib64/libpthread.so.0 0x00007fbc25cb2000 100K /data/soft/jdk1.8.0_141/bin/../lib/amd64/jli/libjli.so 0x00007fbc25aae000 18K /lib64/libdl.so.2 0x00007fbc256ec000 2063K /lib64/libc.so.6 0x00007fbc260e4000 151K /lib64/ld-linux-x86-64.so.2 0x00007fbc246f5000 16615K /data/soft/jdk1.8.0_141/jre/lib/amd64/server/libjvm.so 0x00007fbc243f3000 1114K /lib64/libm.so.6 0x00007fbc240ea000 42K /lib64/librt.so.1 0x00007fbc1fdf1000 64K /data/soft/jdk1.8.0_141/jre/lib/amd64/libverify.so 0x00007fbc1fbc5000 220K /data/soft/jdk1.8.0_141/jre/lib/amd64/libjava.so 0x00007fbc1f9aa000 121K /data/soft/jdk1.8.0_141/jre/lib/amd64/libzip.so 0x00007fbbed321000 49K /data/soft/jdk1.8.0_141/jre/lib/amd64/libmanagement.so 0x00007fbbed10a000 113K /data/soft/jdk1.8.0_141/jre/lib/amd64/libnet.so 0x00007fbbeca48000 90K /data/soft/jdk1.8.0_141/jre/lib/amd64/libnio.so 0x00007fbbec500000 251K /data/soft/jdk1.8.0_141/jre/lib/amd64/libsunec.so 0x00007fbbec2ea000 86K /lib64/libgcc_s.so.1 0x00007fbb78e4c000 50874K /data/home/aics/taskbot-tomcat/temp/tensorflow_native_libraries-1545634810127-0/libtensorflow_jni.so 0x00007fbb944f4000 15740K /data/home/aics/taskbot-tomcat/temp/tensorflow_native_libraries-1545634810127-0/libtensorflow_framework.so 0x00007fbb941d7000 972K /lib64/libstdc++.so.6 0x00007fbb69848000 50874K /data/home/aics/taskbot-tomcat/temp/tensorflow_native_libraries-1545641641135-0/libtensorflow_jni.so 看起來內(nèi)存占用情況還好，沒有特別的內(nèi)存，但是不知道為什么libtensorflow_jni.so占用了兩個50874K的堆。

轉(zhuǎn)成hprof后分析

/data/soft/jdk/bin/jmap -dump:format=b,file=dump.hprof /data/soft/jdk/bin/java core.14652
轉(zhuǎn)換之后就可以很方面的使用MAT或VisualVM進行JVM堆內(nèi)存分析

常見問題

Missing separate debuginfo for: 執(zhí)行命令debuginfo-install glibc-2.17-157.tl2.2.x86_64 libgcc-4.8.5-4.el7.x86_64 libstdc+±4.8.5-4.el7.x86_64

這里一般按照錯誤中的提示來執(zhí)行命令，不過jvm相關的我也沒有安裝成功，也許是要使用OpenJDK才有調(diào)試信息包吧

參考

linux下core dump【總結(jié)】
在 Linux 生成 core dump 文件
如何抓獲JVM crash的幕后黑手？（一）
https://docs.oracle.com/javase/8/docs/technotes/guides/troubleshoot/tooldescr016.html#BABBAIJC
https://docs.oracle.com/javase/7/docs/technotes/tools/share/jmap.html

總結(jié)

以上是生活随笔為你收集整理的JDK core dump分析的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯，歡迎將生活随笔推薦給好友。