memcached proxy error

[jianhuihi]

twemproxy 0.4 memcached proxy
run code
$memcache = new Memcached;
$memcache->addServer('192.168.8.56', 22122);
$a = $memcache->getMulti(array('sun6456', 'sun444'));

twemproxy log

[2014-11-13 13:59:57.734] nc_proxy.c:206 p 7 listening on '192.168.8.56:6380' in redis pool 1 'redis' with 3 servers
[2014-11-13 13:59:57.734] nc_proxy.c:206 p 8 listening on '192.168.8.56:22122' in memcache pool 2 'memcached' with 4 servers
[2014-11-13 14:00:02.886] nc_proxy.c:366 accepted c 9 on p 8 from '192.168.12.71:58439'
src/nutcracker(nc_stacktrace_fd+0x20)[0x412700]
src/nutcracker(signal_handler+0xa6)[0x410306]
/lib64/libpthread.so.0[0x3cf380f500]
src/nutcracker(req_recv_done+0x12a)[0x40b14a]
src/nutcracker(msg_recv+0xf0)[0x409f30]
src/nutcracker(core_core+0x111)[0x405a81]
src/nutcracker(event_wait+0xb8)[0x41d418]
src/nutcracker(core_loop+0x19)[0x405889]
src/nutcracker(main+0x5b8)[0x413438]
/lib64/libc.so.6(__libc_start_main+0xfd)[0x3cf341ecdd]
src/nutcracker[0x405609]
[.......................] signal 11 (SIGSEGV) received, core dumping

[idning]

thanks for your report, it looks a bug introduced by mget code refactor.

please run this cmd and get backtrace:

cat nutcracker.log | grep '^src/nutcracker' | awk -F '[' '{print $2}' | addr2line -e bin/nutcracker

or get gdb stack from core file.

[jianhuihi]

[2014-11-13 16:24:47.751] nc.c:187 nutcracker-0.4.0 built for Linux 2.6.32-220.4.1.el6.centos.plus.x86_64 x86_64 started on pid 8219
[2014-11-13 16:24:47.751] nc.c:192 run, rabbit run / dig that hole, forget the sun / and when at last the work is done / don't sit down / it's time to dig another one
[2014-11-13 16:24:47.755] nc_core.c:43 max fds 65535 max client conns 65489 max server conns 14
[2014-11-13 16:24:47.755] nc_stats.c:851 m 4 listening on '0.0.0.0:22222'
[2014-11-13 16:24:47.755] nc_epoll.c:63 e 5 with nevent 1024
[2014-11-13 16:24:47.755] nc_proxy.c:206 p 7 listening on '192.168.8.56:8880' in redis pool 0 'ssdb' with 4 servers
[2014-11-13 16:24:47.755] nc_proxy.c:206 p 8 listening on '192.168.8.56:6380' in redis pool 1 'redis' with 3 servers
[2014-11-13 16:24:47.755] nc_proxy.c:206 p 9 listening on '192.168.8.56:22122' in memcache pool 2 'memcached' with 4 servers
[2014-11-13 16:25:04.925] nc_proxy.c:366 accepted c 10 on p 9 from '192.168.12.71:64747'
src/nutcracker(nc_stacktrace_fd+0x20)[0x412700]
src/nutcracker(signal_handler+0xa6)[0x410306]
/lib64/libpthread.so.0[0x3cf380f500]
src/nutcracker(req_recv_done+0x12a)[0x40b14a]
src/nutcracker(msg_recv+0xf0)[0x409f30]
src/nutcracker(core_core+0x111)[0x405a81]
src/nutcracker(event_wait+0xb8)[0x41d418]
src/nutcracker(core_loop+0x19)[0x405889]
src/nutcracker(main+0x5b8)[0x413438]
/lib64/libc.so.6(__libc_start_main+0xfd)[0x3cf341ecdd]
src/nutcracker[0x405609]
[.......................] signal 11 (SIGSEGV) received, core dumping

[jianhuihi]

[root@pppoe56 twemproxy]# cat nutcracker.log | grep '^src/nutcracker' | awk -F '[' '{print $2}' | addr2line -e src/nutcracker
/root/twemproxy/src/nc_util.c:310
/root/twemproxy/src/nc_signal.c:115
/root/twemproxy/src/nc_request.c:638
/root/twemproxy/src/nc_message.c:591
/root/twemproxy/src/nc_core.c:195
/root/twemproxy/src/event/nc_epoll.c:249
/root/twemproxy/src/nc_core.c:353
/root/twemproxy/src/nc.c:531

[idning]

this is a known problem with old gcc, compile with CFLAGS="-ggdb3 -O1" will solve it.

ref:

• SIGSEGV nutcracker-0.2.3 #115
• Running problems with twemproxy. #146

I will try to understand how this happen.

[idning]

I found it's the -f-strict-aliasing flag which break this, try to build with this:

    ./configure CFLAGS="-O3 -fno-strict-aliasing && make clean && make 

[idning]

add -fno-strict-aliasing as default option in this commit:

1f17bc1

[manjuraj]

This is a great find. Good job @idning