Winse Blog

走走停停, 熙熙攘攘, 忙忙碌碌, 不知何畏.

Build redis-2.8

jemalloc

默认make使用的libc,在内存方面会产生比较多的碎片。可以使用jemalloc要进行内存的分配管理。

如果报make cc Command not found,需要先安装gcc。

1
2
3
4
5
6
7
8
9
tar zxvf redis-2.8.13.bin.tar.gz 
cd redis-2.8.13
cd deps/jemalloc/
# 用于产生h头文件
./configure 

cd redis-2.8.13
make MALLOC=jemalloc
src/redis-server 

查看jemalloc的include的内容如下:

1
2
3
[hadoop@localhost jemalloc]$ cd include/jemalloc/
internal/              jemalloc_defs.h.in     jemalloc_macros.h      jemalloc_mangle.h      jemalloc_mangle.sh     jemalloc_protos.h.in   jemalloc_rename.h      jemalloc.sh
jemalloc_defs.h        jemalloc.h             jemalloc_macros.h.in   jemalloc_mangle_jet.h  jemalloc_protos.h      jemalloc_protos_jet.h  jemalloc_rename.sh  

查看内存使用:

1
2
3
4
5
6
7
8
9
10
11
12
[hadoop@localhost redis-2.8.13]$ src/redis-cli info
...
# Memory
used_memory:503576
used_memory_human:491.77K
used_memory_rss:2158592
used_memory_peak:503576
used_memory_peak_human:491.77K
used_memory_lua:33792
mem_fragmentation_ratio:4.29
mem_allocator:jemalloc-3.6.0
...

redis在使用过程中,会产生碎片。重启以及libc和jemalloc的对比如下:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
# 运行中实例
# Memory
used_memory:4623527744
used_memory_human:4.31G
used_memory_rss:48304705536
used_memory_peak:38217543280
used_memory_peak_human:35.59G
used_memory_lua:33792
mem_fragmentation_ratio:10.45
mem_allocator:libc

51616 hadoop    20   0 45.1g  44g 1136 S  0.0 35.7   3410:42 /home/hadoop/redis-2.8.13/src/redis-server *:6371

# 序列化为rdb的文件大小
[hadoop@hadoop-master1 18111]$ ll
总用量 1183116
-rw-rw-r--. 1 hadoop hadoop 1210319541 1月  14 11:28 dump.rdb

# 重启后的实例
[hadoop@hadoop-master1 18111]$  ~/redis-2.8.13/src/redis-server --port 18111
[77484] 14 Jan 14:33:17.910 * DB loaded from disk: 218.337 seconds

# Memory
used_memory:4763158704
used_memory_human:4.44G
used_memory_rss:6217580544
used_memory_peak:4763158704
used_memory_peak_human:4.44G
used_memory_lua:33792
mem_fragmentation_ratio:1.31
mem_allocator:libc

77484 hadoop    20   0 6052m 5.8g 1200 S  0.0  4.6   3:38.39 /home/hadoop/redis-2.8.13/src/redis-server *:18111

# 使用jemalloc替换libc的实例
[hadoop@hadoop-master1 18111]$ ~/redis-jemalloc/redis-2.8.13/src/redis-server --port 18888
[14793] 14 Jan 14:50:11.250 * DB loaded from disk: 209.839 seconds

# Memory
used_memory:4527760088
used_memory_human:4.22G
used_memory_rss:4625887232
used_memory_peak:4527760088
used_memory_peak_human:4.22G
used_memory_lua:33792
mem_fragmentation_ratio:1.02
mem_allocator:jemalloc-3.6.0

14793 hadoop    20   0 4538m 4.3g 1360 S  0.0  3.4   3:28.10 /home/hadoop/redis-jemalloc/redis-2.8.13/src/redis-server *:18888                                                                                                                       

tcmalloc

  • root安装

如果有root用户的话操作比较简单。现在gperftoolslibunwind-0.99-beta

1
2
3
4
5
6
7
8
9
cd libunwind-0.99-beta
./configure 
make && make install
cd /home/hadoop/gperftools-2.4
./configure 
make && make install

cd redis-2.8.13
make MALLOC=tcmalloc

如果出现./libtool: line 1125: g++: command not found的错误,缺少编译环境;

1
[root@localhost gperftools-2.4]# yum -y install gcc+ gcc-c++

编译后,运行报错src/redis-server: error while loading shared libraries: libtcmalloc.so.4: cannot open shared object file: No such file or directory,需要配置环境变量:

1
2
[hadoop@localhost redis-2.8.13]$ export LD_LIBRARY_PATH=/usr/local/lib
[hadoop@localhost redis-2.8.13]$ src/redis-server 

或者按照网上的做法:

1
2
echo "/usr/local/lib" > /etc/ld.so.conf.d/usr_local_lib.conf  
/sbin/ldconfig  

检查tcmalloc是否生效lsof -n | grep tcmalloc,出现以下信息说明生效

1
redis-ser 1716    hadoop  mem       REG  253,0  2201976  936349 /usr/local/lib/libtcmalloc.so.4.2.6

修改配置文件找到daemonize,将后面的no改为yes,让其可以以服务方式运行。

  • 普通用户安装

考虑到可以各台机器上面复制,指定编译目录这种方式会比较方便。

1
2
3
4
5
6
7
8
9
10
cd libunwind-0.99-beta
CFLAGS=-fPIC ./configure --prefix=/home/hadoop/redis
make && make install

cd gperftools-2.4
./configure -h
export LDFLAGS="-L/home/hadoop/redis/lib"
export CPPFLAGS="-I/home/hadoop/redis/include"
./configure --prefix=/home/hadoop/redis
make && make install

编译好后,把东西redis目录内容移到redis-2.8.13/src下。然后修改src/Makefile:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
[hadoop@master1 redis-2.8.13]$ vi src/Makefile
# Include paths to dependencies
FINAL_CFLAGS+= -I../deps/hiredis -I../deps/linenoise -I../deps/lua/src
    
ifeq ($(MALLOC),tcmalloc)
        #FINAL_CFLAGS+= -DUSE_TCMALLOC
        #FINAL_LIBS+= -ltcmalloc
        FINAL_CFLAGS+= -DUSE_TCMALLOC -I./include
        FINAL_LIBS+= -L./lib  -ltcmalloc -ldl

endif

ifeq ($(MALLOC),tcmalloc_minimal)
        FINAL_CFLAGS+= -DUSE_TCMALLOC
        FINAL_LIBS+= -ltcmalloc_minimal
endif

然后编译:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
[hadoop@master1 redis-2.8.13]$ export LD_LIBRARY_PATH=/home/hadoop/redis-2.8.13/src/lib
[hadoop@master1 redis-2.8.13]$ make MALLOC=tcmalloc
cd src && make all
make[1]: Entering directory `/home/hadoop/redis-2.8.13/src'
    LINK redis-server
    INSTALL redis-sentinel
    CC redis-cli.o
In file included from zmalloc.h:40,
                 from redis-cli.c:50:
./include/google/tcmalloc.h:35:2: warning: #warning is a GCC extension
./include/google/tcmalloc.h:35:2: warning: #warning "google/tcmalloc.h is deprecated. Use gperftools/tcmalloc.h instead"
    LINK redis-cli
    CC redis-benchmark.o
In file included from zmalloc.h:40,
                 from redis-benchmark.c:47:
./include/google/tcmalloc.h:35:2: warning: #warning is a GCC extension
./include/google/tcmalloc.h:35:2: warning: #warning "google/tcmalloc.h is deprecated. Use gperftools/tcmalloc.h instead"
    LINK redis-benchmark
    CC redis-check-dump.o
    LINK redis-check-dump
    CC redis-check-aof.o
    LINK redis-check-aof

Hint: To run 'make test' is a good idea ;)

make[1]: Leaving directory `/home/hadoop/redis-2.8.13/src'
[hadoop@master1 redis-2.8.13]$ 

redis3集群安装cluster

编译安装和2.8一样,configuration/make/makeinstall即可。

1
2
3
4
5
6
[hadoop@umcc97-44 cluster-test]$ cat cluster.conf 
port .
cluster-enabled yes
cluster-config-file nodes.conf
cluster-node-timeout 5000
appendonly yes

比较苦逼的是需要安装ruby,服务器不能上网!其实ruby在能访问的机器上面安装就可以了!初始化集群的脚本其实就是客户端连接服务端,初始化集群而已。 还有就是在调用命令的时刻要加上-c,这样才是使用集群模式,不然仅仅连单机,读写其他集群服务会报错!

Cygwin

开发环境系统都是在windows,想调试一步步的看源码就得编译下redis。由于cygwin环境,模拟的linux,有部分的变量没有定义,需要进行修改。修改如下:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
$ git log -1
commit 0c211a1953afeda3d0d45126653e2d4c38bd88cb
Author: antirez <antirez@gmail.com>
Date:   Fri Dec 5 10:51:09 2014 +010

$ git branch
* 2.8

$ git diff
diff --git a/deps/hiredis/net.c b/deps/hiredis/net.c
index bdb84ce..6e95f22 100644
--- a/deps/hiredis/net.c
+++ b/deps/hiredis/net.c
@@ -51,6 +51,13 @@
 #include "net.h"
 #include "sds.h"

+/* Cygwin Fix */
+#ifdef __CYGWIN__
+#define TCP_KEEPCNT 8
+#define TCP_KEEPINTVL 150
+#define TCP_KEEPIDLE 14400
+#endif
+
 /* Defined in hiredis.c */
 void __redisSetError(redisContext *c, int type, const char *str);

diff --git a/src/Makefile b/src/Makefile
index 8b3e959..a72b2f2 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -63,6 +63,9 @@ else
 ifeq ($(uname_S),Darwin)
        # Darwin (nothing to do)
 else
+ifeq ($(uname_S),CYGWIN_NT-6.3-WOW64)
+       # cygwin (nothing to do)
+else
 ifeq ($(uname_S),AIX)
         # AIX
         FINAL_LDFLAGS+= -Wl,-bexpall
@@ -75,6 +78,7 @@ else
 endif
 endif
 endif
+endif
 # Include paths to dependencies
 FINAL_CFLAGS+= -I../deps/hiredis -I../deps/linenoise -I../deps/lua/src

然后编译:

1
2
3
4
5
cd deps/
make lua hiredis linenoise

cd ..
make

编译成功后,把程序导入eclipse CDT环境进行运行调试。导入后需要重新构建一下,不然调试的时刻会按照/cygwin的路径来查找源码。

  • Import,然后选择C/C++目录下的[Existing Code as Makefile project]
  • 在[Existing Code Location]填入redis程序对应的目录,在[Toolchain for Indexer Settings]选择Cygwin GCC
  • 导入完成后,右键选择[Build Configuration]->[Build All]
  • Run然后选择执行redis-server即可。

好像也可以远程调试

1
[root@Frankzfz]$gdbserver 10.27.10.48:9000 ./test_seg_fault

参考

–END

Comments