环境：
操作系统为：SunOS 5.10
数据库版本：Oracle RAC 11.2.0.3.0
该主机上有2个库，一套RAC的节点1在上面，还有另外一个库，之所以使用的是DB下的监听而没用GRID下的监听，是为了避免停掉CRS时影响另外一个库的使用。


故障现象：
数据库监听不定期出现异常 ,从应用tnsping数据库，时间花费很长甚至连不上


故障分析处理过程：


从应用主机tnsping如下：


racdb1_scenemon$tnsping racdb_new
 
TNS Ping Utility for Solaris: Version 11.2.0.3.0 - Production on 19-2月 -2014 15:46:23
 
Copyright (c) 1997, 2011, Oracle. All rights reserved.
 
已使用的参数文件:


已使用 TNSNAMES 适配器来解析别名
 尝试连接 (DESCRIPTION = (ADDRESS = (PROTOCOL = TCP)(HOST = 11.111.11.1)(PORT = 1521)) (ADDRESS = (PROTOCOL = TCP)(HOST = 11.111.11.2)(PORT = 1521)) (LOAD_BALANCE = yes) (CONNECT_DATA = (SERVER = DEDICATED) (SERVICE_NAME = racdb) (FAILOVER_MODE = (TYPE = SELECT) (METHOD = BASIC) (RETRIES = 180) (DELAY = 5))))
 
OK (60000 毫秒)


从数据库服务器上查看监听状态：


$ lsnrctl status
 
LSNRCTL for Solaris: Version 11.2.0.3.0 - Production on 19-FEB-2014 15:45:58
 
Copyright (c) 1991, 2011, Oracle. All rights reserved.
 
Connecting to (ADDRESS=(PROTOCOL=tcp)(HOST=)(PORT=1521))
 




TNS-12535: TNS:operation timed out
 TNS-12560: TNS:protocol adapter error
 TNS-00505: Operation timed out   


将监听进程予以kill并重启监听：




$ ps -ef |grep tns
 oracle 18504 18468 0 15:45:15 pts/10 0:00 grep tns
 oracle 24037 1 0 Nov 28 ? 6:08 /oracle/app/asm/11.2.0/grid/bin/tnslsnr LISTENER -inherit
 oracle 9488 1 3 Jan 24 ? 4447:11 /oracle/app/db/product/11.2.0/db/bin/tnslsnr LISTENER -inherit
 $ 
 $ kill -9 9488
 $ lsnrctl start
 
LSNRCTL for Solaris: Version 11.2.0.3.0 - Production on 19-FEB-2014 15:48:22
 
Copyright (c) 1991, 2011, Oracle. All rights reserved.
 
Starting /oracle/app/db/product/11.2.0/db/bin/tnslsnr: please wait...
 
TNSLSNR for Solaris: Version 11.2.0.3.0 - Production
 Log messages written to /oracle/app/db/product/11.2.0/db/log/diag/tnslsnr/racdb/listener/alert/log.xml
 Listening on: (DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=racdb)(PORT=1521)))
 
Connecting to (ADDRESS=(PROTOCOL=tcp)(HOST=)(PORT=1521))
 STATUS of the LISTENER
 ------------------------
 Alias LISTENER
 Version TNSLSNR for Solaris: Version 11.2.0.3.0 - Production
 Start Date 19-FEB-2014 15:48:22
 Uptime 0 days 0 hr. 0 min. 0 sec
 Trace Level off
 Security ON: Local OS Authentication
 SNMP OFF
 Listener Log File /oracle/app/db/product/11.2.0/db/log/diag/tnslsnr/racdb/listener/alert/log.xml
 Listening Endpoints Summary...
 (DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=racdb)(PORT=1521)))
 The listener supports no services
 The command completed successfully   


检查listener.log如下：




17-FEB-2014 16:02:26 * (CONNECT_DATA=(SERVER=DEDICATED)(SERVICE_NAME=racdb)(FAILOVER_MODE=(TYPE=SELECT)(METHOD=BASIC)(RETRIES=180)(DELAY=5))(CID=(PROGRAM=sqlplus@racdb1)(HOST=racdb1)(USER=tmn))) * (ADDRESS=(PROTOCOL=tcp)(HOST=11.111.11.89)(PORT=41603)) * establish * racdb * 0
 17-FEB-2014 16:02:26 * (CONNECT_DATA=(SERVER=DEDICATED)(SERVICE_NAME=racdb)(FAILOVER_MODE=(TYPE=SELECT)(METHOD=BASIC)(RETRIES=180)(DELAY=5))(CID=(PROGRAM=sqlplus@racdb1)(HOST=racdb1)(USER=tmn))) * (ADDRESS=(PROTOCOL=tcp)(HOST=11.111.11.89)(PORT=41659)) * establish * racdb * 12518
 TNS-12518: TNS:listener could not hand off client connection
 TNS-12560: TNS:protocol adapter error
 TNS-00530: Protocol adapter error
 Solaris Error: 24: Too many open files <<<<<<<<<
 17-FEB-2014 16:02:26 * (CONNECT_DATA=(SERVER=DEDICATED)(SERVICE_NAME=racdb)(FAILOVER_MODE=(TYPE=SELECT)(METHOD=BASIC)(RETRIES=180)(DELAY=5))(CID=(PROGRAM=sqlplus@racdb1)(HOST=racdb1)(USER=tmn))) * (ADDRESS=(PROTOCOL=tcp)(HOST=11.111.11.89)(PORT=41652)) * establish * racdb * 0
 17-FEB-2014 16:02:26 * (CONNECT_DATA=(SERVER=DEDICATED)(SERVICE_NAME=racdb)(FAILOVER_MODE=(TYPE=SELECT)(METHOD=BASIC)(RETRIES=180)(DELAY=5))(CID=(PROGRAM=sqlplus@racdb1)(HOST=racdb1)(USER=tmn))) * (ADDRESS=(PROTOCOL=tcp)(HOST=11.111.11.89)(PORT=41669)) * establish * racdb * 0
 17-FEB-2014 16:02:26 * (CONNECT_DATA=(SERVER=DEDICATED)(SERVICE_NAME=racdb)(FAILOVER_MODE=(TYPE=SELECT)(METHOD=BASIC)(RETRIES=180)(DELAY=5))(CID=(PROGRAM=sqlplus@racdb1)(HOST=racdb1)(USER=tmn))) * (ADDRESS=(PROTOCOL=tcp)(HOST=11.111.11.89)(PORT=41682)) * establish * racdb * 12518
 TNS-12518: TNS:listener could not hand off client connection
 TNS-12560: TNS:protocol adapter error
 TNS-00530: Protocol adapter error
 Solaris Error: 24: Too many open files <<<<<<<<<<


Too many open files意味着Maximum Number Of Open Files Per Process 达到了上限。因此listener hang住的原因可能是该limit设置过小，进一步查看硬件限制如下：




$ ulimit -Ha
 time(seconds) unlimited
 file(blocks) unlimited
 data(kbytes) unlimited
 stack(kbytes) unlimited
 coredump(blocks) unlimited
 nofiles(descriptors) 65536
 memory(kbytes) unlimited


查看软件限制如下：




$ ulimit -Sa
 time(seconds) unlimited
 file(blocks) unlimited
 data(kbytes) unlimited
 stack(kbytes) 8192
 coredump(blocks) unlimited
 nofiles(descriptors) 256
 memory(kbytes) unlimited


查看/etc/system如下：




*ident  "@(#)system    1.18    97/06/27 SMI" /* SVR4 1.5 */


*


* SYSTEM SPECIFICATION FILE


*


* moddir:


*


*      Set the search path for modules.  This has a format similar to the


*      csh path variable. If the module isn't found in the first directory


*      it tries the second and so on. The default is /kernel /usr/kernel


*


*      Example:


*              moddir: /kernel /usr/kernel /other/modules


 
* root device and root filesystem configuration:


*


*      The following may be used to override the defaults provided by


*      the boot program:


*


*      rootfs:        Set the filesystem type of the root.


*


*      rootdev:        Set the root device.  This should be a fully


*                      expanded physical pathname.  The default is the


*                      physical pathname of the device where the boot


*                      program resides.  The physical pathname is


*                      highly platform and configuration dependent.


*


*      Example:


*              rootfs:ufs


*              rootdev:/sbus@1,f8000000/esp@0,800000/sd@3,0:a


*


*      (Swap device configuration should be specified in /etc/vfstab.)
 
* exclude:


*


*      Modules appearing in the moddir path which are NOT to be loaded,


*      even if referenced. Note that `exclude' accepts either a module name,


*      or a filename which includes the directory.


*


*      Examples:


*              exclude: win


*              exclude: sys/shmsys


* forceload:


*


*      Cause these modules to be loaded at boot time, (just before mounting


*      the root filesystem) rather than at first reference. Note that


*      forceload expects a filename which includes the directory. Also


*      note that loading a module does not necessarily imply that it will


*      be installed.


*


*      Example:


*              forceload: drv/foo


* set:


*


*      Set an integer variable in the kernel or a module to a new value.


*      This facility should be used with caution.  See system(4).


*


*      Examples:


*


*      To set variables in 'unix':


*


*              set nautopush=32


*              set maxusers=40


*


*      To set a variable named 'debug' in the module named 'test_module'


*


*              set test_module:debug = 0x13


 




* Begin FJSVssf (do not edit)


set ftrace_atboot = 1


set kmem_flags = 0x100


set kmem_lite_maxalign = 8192


* End FJSVssf (do not edit)


forceload: drv/fjpfca


* Begin MDD root info (do not edit)


rootdev:/pseudo/md@0:0,0,blk


* End MDD root info (do not edit)


*** Begin EMCpower added lines *** DO NOT EDIT BELOW THIS LINE ***


forceload: drv/emcpsf


forceload: drv/sd


forceload: drv/ssd


forceload: drv/emcp


forceload: misc/emcpgpx


forceload: misc/emcpmpx


forceload: misc/emcpvlumd


forceload: misc/emcpxcrypt


forceload: misc/emcpdm


forceload: misc/emcpioc


set emcp:bPxEnableInit=1


*** End EMCpower added lines *** DO NOT EDIT ABOVE THIS LINE ***


exec_user_stack = 1                           


set noexec_user_stack_log = 1


查看/etc/project如下：




system:0::::


user.root:1::::


noproject:2::::


default:3::::


group.staff:10::::


user.oracle:100::oracle::




综上所述，从listener.log的输出看，nofiles的soft limit过小，只有256，这会导致Solaris Error: 24: Too many open files 错误。从/etc/system和/etc/project，我们没有看到设置了soft limit。


故障处理小结及后续建议


将oracle用户的soft limit提升为至少1024，然后重新oracle用户登录，检验ulimit合格后，重新启动数据库和监听。


 具体解决办法如下：


1、在/etc/system增加以下行


set rlim_fd_max=65536


set rlim_fd_cur=4096


2、重新登录ORACLE并检验oracle用户的限制


su – oracle


ulimit -Ha


ulimit –Sa


3、重新启动数据库和监听