How to debug the impala source?

use gdb

Hi,when I use gdb to debug the service/impalad,I encountered the problem as
follow:
[impala@hadoop-01 service]$ gdb -q impalad
Reading symbols from
/usr/local/src/impala/be/build/debug/service/impalad...done.
(gdb) break main
Breakpoint 1 at 0x9adba1: file
/usr/local/src/impala/be/src/service/impalad-main.cc, line 71.
(gdb) set args -use_statestore=false -nn=hadoop-01.localdomain -nn_port=8030
(gdb) clear main
Deleted breakpoint 1
(gdb) quit
[impala@hadoop-01 service]$ gdb -q impalad
Reading symbols from
/usr/local/src/impala/be/build/debug/service/impalad...done.
(gdb) set args -use_statestore=false -nn=hadoop-01.localdomain -nn_port=8030
(gdb) break main
Breakpoint 1 at 0x9adba1: file
/usr/local/src/impala/be/src/service/impalad-main.cc, line 71.
(gdb) r
Starting program: /usr/local/src/impala/be/build/debug/service/impalad
-use_statestore=false -nn=hadoop-01.localdomain -nn_port=8030
[Thread debugging using libthread_db enabled]

Breakpoint 1, main (argc=4, argv=0x7fffffffbc48) at
/usr/local/src/impala/be/src/service/impalad-main.cc:71
71 InitDaemon(argc, argv);
Missing separate debuginfos, use: debuginfo-install
boost-date-time-1.41.0-11.el6_1.2.x86_64
boost-regex-1.41.0-11.el6_1.2.x86_64 boost-thread-1.41.0-11.el6_1.2.x86_64
bzip2-libs-1.0.5-7.el6_0.x86_64 glibc-2.12-1.80.el6_3.6.x86_64
libevent-1.4.13-4.el6.x86_64 libgcc-4.4.6-4.el6.x86_64
libicu-4.2.1-9.1.el6_2.x86_64 libstdc++-4.4.6-4.el6.x86_64
zlib-1.2.3-27.el6.x86_64
(gdb) cont
Continuing.
[New Thread 0x7ffff2dd8700 (LWP 8180)]
[New Thread 0x7ffff2cd7700 (LWP 8181)]
[New Thread 0x7ffff2a3a700 (LWP 8182)]
[New Thread 0x7ffff2939700 (LWP 8183)]
[New Thread 0x7ffff2838700 (LWP 8184)]
[New Thread 0x7fffec8a6700 (LWP 8185)]
[New Thread 0x7fffec7a5700 (LWP 8186)]
[New Thread 0x7fffec6a4700 (LWP 8187)]
[New Thread 0x7fffec5a3700 (LWP 8188)]
[New Thread 0x7fffec4a2700 (LWP 8189)]

Program received signal SIGSEGV, Segmentation fault.
0x00007ffff326f4e8 in ?? ()

Could you tell me the reasons for the Segmentation fault and how to debug
the impala in details?
Thanks！
Qiang


2013/1/21 Jung-Yup Lee <ljykjh@gmail.com>

Try this way
1. Run impalad
2. gdb>attach <pidofimpalad>
2013. 1. 26. 오후 2:16에 "Qiang Zhou" <zqlongqiang@gmail.com>님이 작성:

Debugging the BE by attaching to a running Impalad is easy. As Jung-Yup
mentioned, just use: gdb --pid <impalad pid> .

However, the embedded JVM makes debugging the Impala BE from startup a bit
tedious. The JVM generates a bunch of segmentation faults that are meant to
be handled by the JVM, but get trapped by gdb.
You can either manually continue on all of these traps (not fun) or tell
gdb to ignore them (and let Java handle them) using:
*
*
gdb> handle SIGSEGV nostop noprint pass

Debugging the FE (Java) can be done using Eclipse or jdb. If Impala is
started with the following environment variable defined it will allow you
to attach to the JVM:
Hope this helps.

Thanks,
Lenni
Software Engineer - Cloudera

Thank you! I have tried this way,but the problem occured as follows when i
execute this sql 'select count(*) from tab1' in the shell:
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffec7a5700 (LWP 7862)]
0x00007ffff6ce74e3 in ciMethod::method_data() () from
/usr/java/default/jre/lib/amd64/server/libjvm.so
I really don't know what happens in the /usr/java/default/...
Qiang


2013/1/28 Jung-Yup Lee <ljykjh@gmail.com>

Sorry, gdb may also trap on SIGSEGV's meant for the JVM at runtime as well.
Please try the suggestion I made on ignoring these - after attaching to
impalad, but before running the query issue the following gdb command:

gdb> handle SIGSEGV nostop noprint pass

Let me know if that doesn't solve your problem.

Thanks,
Lenni
Software Engineer - Cloudera

I just want to reply your given suggestion, the problem that the SIGSEGV
incurs the segmentation fault has solved just now.
Thank your enthusiastic guidance!
And could you give me some the debugging infomations about impala?
Thanks again!
Qiang


2013/1/29 Lenni Kuff <lskuff@cloudera.com>