升级Zabbix proxy到3.0.2后无法启动报错

一 问题描述

将一台zabbix proxy由2.4.5升级到3.0.2后启动了就直接崩溃了。

错误信息如下：

2367:20160508:153246.830 One child process died (PID:42385,exitcode/signal:11). Exiting ...

42367:20160508:153248.904 Zabbix Proxy stopped. Zabbix 3.0.2 (revision 59540).

将日志级别调高点，设置DebugLevel=4，查看报错

 42629:20160508:153529.004 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42628:20160508:153529.004 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42634:20160508:153529.004 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42651:20160508:153529.004 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42659:20160508:153529.004 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42661:20160508:153529.005 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42655:20160508:153529.005 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42663:20160508:153529.005 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42664:20160508:153529.005 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42666:20160508:153529.005 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42665:20160508:153529.006 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42667:20160508:153529.006 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42630:20160508:153529.006 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42668:20160508:153529.006 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42671:20160508:153529.006 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42670:20160508:153529.007 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42647:20160508:153529.007 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42674:20160508:153529.007 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42680:20160508:153529.007 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ... 42679:20160508:153529.008 Got signal [signal:15(SIGTERM),sender_pid:42623,sender_uid:498,reason:0]. Exiting ...

 9102:20160508:170455.283 End of zbx_strpool_destroy()  9102:20160508:170455.283 End of free_configuration_cache()  9102:20160508:170455.283 In free_ipmi_handler()  9102:20160508:170455.283 End of free_ipmi_handler()  9102:20160508:170455.284 In free_selfmon_collector() collector:0x7fa793e23000  9102:20160508:170455.284 End of free_selfmon_collector()  9102:20160508:170455.284 In unload_modules()  9102:20160508:170455.284 End of unload_modules()  9102:20160508:170455.284 Zabbix Proxy stopped. Zabbix 3.0.2 (revision 59540).zabbix_proxy [9152]: [file:'selfmon.c',line:375] lock failed: [22] Invalid argument

总共有4台proxy，每台proxy的操作系统版本都是CentOS6。就单独这一台zabbix proxy启动有问题，刚启动就有进程死掉。

二 问题解决

查看这台proxy服务器的内核版本是3.9.8-1.el6.elrepo.x86_64，而其他几台proxy的内核版本是2.6.32-358.el6.x86_64 刚开始也没有朝着内核版本的问题方向思考。以为还是proxy的版本或者系统参数设置有问题。

1.排查Linux下的信号量

查看Zabbix的官方BUG追踪平台Invalid argument可能是信号量达到了限制

可以尝试增大信号量的值

echo 256 40000 32 32000 > /proc/sys/kernel/sem

这个四个参数分别是

SEMMSL 每个信号量集的含有的信号量的最大值

SEMMNS 系统层面的信号量数量最大限制

SEMOPM semop(2)调用可以指定的最大操作数量

SEMMNI 系统层面的信号数量最小限制

通过echo方式更改重启后会失效，可以更改/etc/sysctl.conf

kernel.sem = 250 32000 100 10000

然后执行sysctl -p

更改参数后proxy启动仍然会崩溃

2.排查内核版本原因

Updating to latest kernel (3.10.0-327.10.1.el7.x86_64) + reboot solved it for me
The error reported originally may be something else than what affected me.

查看帖子怀疑是内核版本造成的，这台proxy以前有人升级过内核，更改linux启动项设置默认启动内核为2.6.32.重启后问题得到解决，proxy不再崩溃。但是之前proxy 2.4.5在3.9.8的内核上可以正常运行。

遇到这个BUG的人挺多的，有好几个帖子都是反馈这个问题。

三 补充知识

这个BUG涉及到Linux下的信号量相关的知识。信号量是为了解决在多进程编程下资源共享的问题。著名的用信号量来解决的就是哲学家就餐问题。

Linux下使用ipcs命令来查看共享内存，信号量和消息队列

ipcs 默认是-a参数，显示所有信息

# ipcs------ Shared Memory Segments --------key        shmid      owner      perms      bytes      nattch     status      0x6c6c6536 0          root       600        4096       0                       0x68031618 1835009    zabbix     600        8388583    0                                           0x6c030ad7 11632648   zabbixagen 600        365056     13                      0x68031672 2392073    zabbix     600        8388583    0                       0x78031672 2424842    zabbix     600        16777216   0                       0x74031672 2457611    zabbix     600        4194304    0                       0x67031672 2490380    zabbix     600        1336934400 0                       0x73031672 2523149    zabbix     600        235929600  0                                                            ------ Semaphore Arrays --------key        semid      owner      perms      nsems     0x00000000 0          root       600        1         0x00000000 32769      root       600        1         0x00000000 4325378    apache     600        1         0x7a031618 622595     zabbix     600        12        0x00000000 4358148    apache     600        1         0x7a030ad7 4292613    zabbixagen 600        13        0x7a031672 851974     zabbix     600        12        0x7a031638 1015815    zabbix     600        12        0x7a031620 1441800    zabbix     600        12             ------ Message Queues --------key        msqid      owner      perms      used-bytes   messages

ipcs -m 单独显示共享内存片段信息

ipcs -s 单独显示信号量

ipcs -q 单独显示消息队列

# ipcs -u------ Shared Memory Status --------segments allocated 71pages allocated 1795473pages resident  68130pages swapped   57636Swap performance: 0 attempts      0 successes------ Semaphore Status --------used arrays = 16allocated semaphores = 152------ Messages: Status --------allocated queues = 0used headers = 0used space = 0 bytes

参考资料：

https://support.zabbix.com/browse/ZBX-10657

http://www.cnblogs.com/forilen/p/4316358.html

https://support.zabbix.com/browse/ZBX-3974

http://blog.zabbix.com/mysterious-zabbix-problems-how-we-debug-them/1023/

https://en.wikipedia.org/wiki/Semaphore_(programming)