[20191119]探究ipcs命令输出.txt

[20191119]探究ipcs命令输出.txt $ man ipcs IPCS(8)                    Linux Programmer's Manual                   IPCS(8) NAME        ipcs - provide information on ipc facilities SYNOPSIS        ipcs [ -asmq ] [ -tclup ]        ipcs [ -smq ] -i id        ipcs -h DESCRIPTION        ipcs provides information on the ipc facilities for which the calling process has read access.        The -i option allows a specific resource id to be specified.  Only information on this id will be printed.        Resources may be specified as follows:        -m     shared memory segments        -q     message queues        -s     semaphore arrays        -a     all (this is the default)        The output format may be specified as follows:        -t     time        -p     pid        -c     creator        -l     limits        -u     summary --//ipcs主要用来显示shared memory segments(-m)以及semaphore arrays(-s) --//好像message queues(-q)没有相关信息.由于oracle使用要使用shared memory segments,而且配置需求很大,一些发布版本要修改 --//参数文件/etc/sysctl.conf,满足应用需求.很少人具体了解一些内核参数具体含义,仅仅copy and paste来安装配置数据库, --//而且现在的安装如果不满足条件,oracle会提示如何修改相关参数.自己根据学习做一些测试.千万不要在生产系统做这些测试!! --//从网上转抄的: set kernel parameters by adding the following lines in "/etc/sysctl.conf" fs.aio-max-nr = 1048576 fs.file-max = 6815744 kernel.shmall = 2097152 kernel.shmmax = 536870912 kernel.shmmni = 4096 kernel.sem = 250 32000 100 128 net.ipv4.ip_local_port_range = 9000 65500 net.core.rmem_default = 262144 net.core.rmem_max = 4194304 net.core.wmem_default = 262144 net.core.wmem_max = 1048576 --//kernel.shmall * kernel.shmmni = 2097152*4096 = 8589934592 --//8589934592/1024/1024/1024 = 8G shmmax-> This parameter represents the size in bytes of a single shared memory segment shmmni-> This parameter represents total no.of shared memory segments system wide(4096 is most probably enough) shmall-> We specify this parameter in pages. The SHMALL defines the largest amount of shared memory pages that can be          used at one time on the system. kernel.sem -> This parameter is related to semaphores e.g., kernel.sem = SEMMSL SEMMNS SEMOPM SEMMNI where semmsl:  The number of semaphores per set semmns:  The total number of semaphores available semopm:  The number of operations which can be made per semaphore call semmni:  The maximum number of shared memory segments available in the system --//主要学习了解如下参数,因为这些与ipcs命令输出有关. kernel.shmall = 2097152 kernel.shmmax = 536870912 kernel.shmmni = 4096 kernel.sem = 250 32000 100 128 --//为了测试方便,我建立了实例文件并没有数据文件. $ cat /tmp/test.ora test.__db_cache_size=1258291200 test.__oracle_base='/u01/app/oracle'#ORACLE_BASE set from environment test.__shared_pool_size=1048576000 *.db_name='test' *.processes=2596 *.sga_target=2500M *.undo_management='auto' --//注:test.__shared_pool_size=1G,test.__db_cache_size=1.2G. --//内核相关参数配置如下: # grep "^kernel.s[he]" /etc/sysctl.conf ##kernel.shmmax = 68719476736            -> 与下面重复了 kernel.shmall = 4294967296 kernel.shmmax = 19327352832 kernel.shmmni = 4096 kernel.sem = 2600 332800 2600 128 --//68719476736/1024/1024/1024 = 64G --//4294967296/1024/1024/1024 = 4G --//19327352832/1024/1024/1024 = 18G --//注:你可以发现我配置kernel.shmmax重复,最后1个配置有效. 1.测试1: --//首先关闭一些测试库,避免影响.修改环境变量ORACLE_SID=test. $ export ORACLE_SID=test $ ipcs ------ Shared Memory Segments -------- key        shmid      owner      perms      bytes      nattch     status ------ Semaphore Arrays -------- key        semid      owner      perms      nsems ------ Message Queues -------- key        msqid      owner      perms      used-bytes   messages SYS@test> startup nomount pfile='/tmp/@.ora'; ORACLE instance started. Total System Global Area 2622255104 bytes Fixed Size                  2256112 bytes Variable Size            1124074256 bytes Database Buffers         1476395008 bytes Redo Buffers               19529728 bytes $ ipcs ------ Shared Memory Segments -------- key        shmid      owner      perms      bytes      nattch     status 0x00000000 401735681  oracle    640        33554432   18 0x00000000 401768450  oracle    640        167772160  18 0x00000000 401801219  oracle    640        16777216   18 0x00000000 401833989  oracle    640        2415919104 18 0x8a931fb8 401866758  oracle    640        2097152    18 ------ Semaphore Arrays -------- key        semid      owner      perms      nsems 0xc13ea218 326893568  oracle    640        2600 ------ Message Queues -------- key        msqid      owner      perms      used-bytes   messages --//你可以发现建立5个Shared Memory Segments.nattch表示进程连接数量. --//实际上你看连接数量有多少,直接看ipcs 的nattch就很快.而不用执行下面命令. # ps -ef | grep "tes[t]"   |wc      18     145    1163 --//查看Semaphore Arrays. $ ipcs -s ------ Semaphore Arrays -------- key        semid      owner      perms      nsems 0xc13ea218 326893568  oracle    640        2600 $ ipcs -s -i 326893568 Semaphore Array semid=326893568 uid=502  gid=502         cuid=502        cgid=502 mode=0640, access_perms=0640 nsems = 2600 otime = Tue Nov 19 09:26:22 2019 ctime = Tue Nov 19 09:26:22 2019 semnum     value      ncount     zcount     pid 0          0          0          0          52844 1          2559       0          0          52844 2          19574      0          0          52844 3          1          0          0          52844 4          0          0          0          0 5          0          0          0          0 6          0          0          0          0 7          0          1          0          52849 8          0          0          0          0 9          0          1          0          52855 10         0          1          0          0 11         0          1          0          0 12         0          1          0          0 13         0          1          0          52863 14         0          1          0          0 15         0          1          0          0 16         0          1          0          0 17         0          1          0          0 18         0          1          0          52873 19         0          1          0          0 20         0          1          0          0 21         0          1          0          52879 22         0          1          0          0 23         0          0          0          0 24         0          0          0          52885 25         0          0          0          52887 ...--//截断. 2590       0          0          0          0 2591       0          0          0          0 2592       0          0          0          0 2593       0          0          0          0 2594       0          0          0          0 2595       0          0          0          0 2596       0          0          0          0 2597       0          0          0          0 2598       0          0          0          0 2599       0          0          0          52844 --//查看sem可以发现nsems=2600,实际上我数据库参数文件中定义*.processes=2596,多出4个与参数kernel.sem = 2600 332800 2600 128 的SEMMSL. --//kernel.sem = SEMMSL SEMMNS SEMOPM SEMMNI --//semmsl:  The number of semaphores per set ,你可以理解每一个进程有需要一个信号与共享内存段通讯.这个参数与oracle的processes有关. $ ipcs -s -i 326893568 | awk '/semnum/,/^2599/' | awk '$5>0 {print $5}' | sort | uniq -c       5 52844       1 52849       1 52855       1 52863       1 52873       1 52879       1 52885       1 52887       1 pid --//并没有18个进程与信号对应.而且并没有pid=52844进程.也许仅仅安装到nomount的缘故. $ ps -fp 52844 UID        PID  PPID  C STIME TTY          TIME CMD $ ps -eLf | grep "tes[t]" | sort -nk2 UID        PID  PPID   LWP  C NLWP STIME TTY          TIME CMD oracle   52847     1 52847  0    1 09:26 ?        00:00:00 ora_pmon_test oracle   52849     1 52849  0    1 09:26 ?        00:00:00 ora_psp0_test oracle   52851     1 52851  1    1 09:26 ?        00:00:19 ora_vktm_test oracle   52855     1 52855  0    1 09:26 ?        00:00:00 ora_gen0_test oracle   52857     1 52857  0    1 09:26 ?        00:00:00 ora_diag_test oracle   52859     1 52859  0    1 09:26 ?        00:00:00 ora_dbrm_test oracle   52861     1 52861  0    1 09:26 ?        00:00:05 ora_dia0_test oracle   52863     1 52863  0    1 09:26 ?        00:00:00 ora_mman_test oracle   52865     1 52865  0    1 09:26 ?        00:00:00 ora_dbw0_test oracle   52867     1 52867  0    1 09:26 ?        00:00:00 ora_dbw1_test oracle   52869     1 52869  0    1 09:26 ?        00:00:00 ora_dbw2_test oracle   52871     1 52871  0    1 09:26 ?        00:00:00 ora_lgwr_test oracle   52873     1 52873  0    1 09:26 ?        00:00:00 ora_ckpt_test oracle   52875     1 52875  0    1 09:26 ?        00:00:00 ora_smon_test oracle   52877     1 52877  0    1 09:26 ?        00:00:00 ora_reco_test oracle   52879     1 52879  0    1 09:26 ?        00:00:00 ora_mmon_test oracle   52881     1 52881  0    1 09:26 ?        00:00:01 ora_mmnl_test oracle   52882 52843 52882  0    1 09:26 ?        00:00:00 oracletest (DESCRIPTION=(LOCAL=YES)(ADDRESS=(PROTOCOL=beq))) $ ipcs -m -p ------ Shared Memory Creator/Last-op -------- shmid      owner      cpid       lpid 401735681  oracle     52844      52893 401768450  oracle     52844      52893 401801219  oracle     52844      52893 401833989  oracle     52844      52893 401866758  oracle     52844      52893 --//加入-p参数可以发现cpid(Creator)是52844.而lpid(Last-op)是52893(这个进程也不存在) --//如果看ps -eLF最后一行PPID=52843,也就是建立Shared Memory Segments的进程已经消失. --//启动一个新连接看看. SYS@test> @ spid        SID    SERIAL# PROCESS SERVER             SPID       PID  P_SERIAL# C50 ---------- ---------- ------- ------------------ ------ ------- ---------- --------------------------------------------------       3281          3 53194   DEDICATED          53195       20          2 alter system kill session '3281,3' immediate; SYS@test> select sysdate from dual ; SYSDATE ------------------- 2019-11-19 09:59:51 $ ipcs -m -p ------ Shared Memory Creator/Last-op -------- shmid      owner      cpid       lpid 401735681  oracle     52844      53195 401768450  oracle     52844      53195 401801219  oracle     52844      53195 401833989  oracle     52844      53195 401866758  oracle     52844      53195 --//SPID=53195,这次可以对上,估计在nomount阶段,一些进程完成某些操作已经退出. $ ipcs ------ Shared Memory Segments -------- key        shmid      owner      perms      bytes      nattch     status 0x00000000 401735681  oracle    640        33554432   19 0x00000000 401768450  oracle    640        167772160  19 0x00000000 401801219  oracle    640        16777216   19 0x00000000 401833989  oracle    640        2415919104 19 0x8a931fb8 401866758  oracle    640        2097152    19 ------ Semaphore Arrays -------- key        semid      owner      perms      nsems 0xc13ea218 326893568  oracle    640        2600 ------ Message Queues -------- key        msqid      owner      perms      used-bytes   messages --//nattch=19,增加1个链接. --//另外还有一个参数-t,可以显示时间. $ ipcs  -t ------ Shared Memory Attach/Detach/Change Times -------- shmid      owner      attached             detached             changed 401735681  oracle      Nov 19 09:55:49      Nov 19 09:27:47      Nov 19 09:26:11 401768450  oracle      Nov 19 09:55:49      Nov 19 09:27:47      Nov 19 09:26:11 401801219  oracle      Nov 19 09:55:49      Nov 19 09:27:47      Nov 19 09:26:11 401833989  oracle      Nov 19 09:55:49      Nov 19 09:27:47      Nov 19 09:26:11 401866758  oracle      Nov 19 09:55:49      Nov 19 09:55:49      Nov 19 09:26:11 ------ Semaphore Operation/Change Times -------- semid    owner      last-op                    last-changed 326893568 oracle      Tue Nov 19 09:26:22 2019   Tue Nov 19 09:26:22 2019 ------ Message Queues Send/Recv/Change Times -------- msqid    owner      send                 recv                 change --//你可以从这些时间判断是否存在活动. 2.测试2: --//我前面定义kernel.shmmax = 19327352832 等于18G,如果定义很小会什么问题呢? --//假如定义kernel.shmmax=100M,也就是100*1024*1024 = 104857600,实际上就建立许多Shared Memory Segments,实际使用不会有什么问题, --//这也是为什么许多文档设置该参数为机器物理内存的1半的缘故. --//当然最好避免建立许多Shared Memory Segments.相互通讯应该有开销,还有就是Attach/Detach需要一点点时间(也就是影响登录与退出) # grep "^kernel.s[he]" /etc/sysctl.conf kernel.shmall = 4294967296 kernel.shmmax = 104857600 kernel.shmmni = 4096 kernel.sem = 2600 332800 2600 128 # sysctl -p --//重新启动实例到nomount看看. $ ps -ef | grep sqlplus | grep sysdb[a] oracle   52843 52842  0 09:26 pts/10   00:00:00 /u01/app/oracle/product/11.2.0.4/dbhome_1/bin/sqlplus   as sysdba --//使用strace跟踪pid=52843. $ strace -f -p 52843 -e ipc -o /tmp/52843.txt SYS@test> startup nomount pfile='/tmp/@.ora'; ORACLE instance started. Total System Global Area 2622255104 bytes Fixed Size                  2256112 bytes Variable Size            1124074256 bytes Database Buffers         1476395008 bytes Redo Buffers               19529728 bytes $ strace -f -p 52843 -e ipc -o /tmp/52843.txt Process 52843 attached - interrupt to quit Process 53315 attached Process 53316 attached Process 53316 detached Process 53317 attached Process 53318 attached Process 53317 detached Process 53319 attached Process 53320 attached Process 53319 detached Process 53321 attached Process 53322 attached (waiting for parent) Process 53322 resumed (parent 53321 ready) Process 53321 detached Process 53323 attached Process 53322 suspended Process 53324 attached (waiting for parent) Process 53324 resumed (parent 53323 ready) Process 53322 resumed Process 53323 detached Process 53324 detached Process 53325 attached Process 53326 attached (waiting for parent) Process 53326 resumed (parent 53325 ready) Process 53325 detached Process 53327 attached (waiting for parent) Process 53327 resumed (parent 53320 ready) Process 53328 attached (waiting for parent) Process 53328 resumed (parent 53327 ready) Process 53327 detached Process 53329 attached Process 53330 attached Process 53329 detached Process 53331 attached Process 53332 attached Process 53331 detached Process 53333 attached Process 53334 attached Process 53333 detached Process 53335 attached Process 53336 attached Process 53335 detached Process 53337 attached Process 53338 attached Process 53337 detached Process 53339 attached Process 53340 attached Process 53339 detached Process 53341 attached Process 53342 attached Process 53341 detached Process 53343 attached Process 53344 attached Process 53343 detached Process 53345 attached Process 53346 attached (waiting for parent) Process 53346 resumed (parent 53345 ready) Process 53345 detached Process 53347 attached Process 53348 attached Process 53347 detached Process 53349 attached Process 53350 attached Process 53349 detached Process 53351 attached Process 53352 attached (waiting for parent) Process 53352 resumed (parent 53351 ready) Process 53351 detached Process 53353 attached Process 53315 detached Process 53354 attached Process 53355 attached Process 53354 detached Process 53356 attached Process 53357 attached (waiting for parent) Process 53357 resumed (parent 53356 ready) Process 53356 detached Process 53355 detached Process 53357 detached $ ipcs -a ------ Shared Memory Segments -------- key        shmid      owner      perms      bytes      nattch     status 0x00000000 401932289  oracle    640        33554432   18 0x00000000 401965058  oracle    640        100663296  18 0x00000000 401997827  oracle    640        50331648   18 0x00000000 402030597  oracle    640        33554432   18 0x00000000 402063366  oracle    640        100663296  18 0x00000000 402096135  oracle    640        100663296  18 0x00000000 402128904  oracle    640        100663296  18 0x00000000 402161673  oracle    640        100663296  18 0x00000000 402194442  oracle    640        100663296  18 0x00000000 402227211  oracle    640        100663296  18 0x00000000 402259980  oracle    640        100663296  18 0x00000000 402292749  oracle    640        100663296  18 0x00000000 402325518  oracle    640        100663296  18 0x00000000 402358287  oracle    640        100663296  18 0x00000000 402391056  oracle    640        100663296  18 0x00000000 402423825  oracle    640        100663296  18 0x00000000 402456594  oracle    640        100663296  18 0x00000000 402489363  oracle    640        100663296  18 0x00000000 402522132  oracle    640        100663296  18 0x00000000 402554901  oracle    640        100663296  18 0x00000000 402587670  oracle    640        100663296  18 0x00000000 402620439  oracle    640        100663296  18 0x00000000 402653208  oracle    640        100663296  18 0x00000000 402685977  oracle    640        100663296  18 0x00000000 402718746  oracle    640        100663296  18 0x00000000 402751515  oracle    640        100663296  18 0x00000000 402784284  oracle    640        100663296  18 0x00000000 402817053  oracle    640        100663296  18 0x8a931fb8 402849822  oracle    640        2097152    18 --//注意仅仅1个key有值0x8a931fb8. ------ Semaphore Arrays -------- key        semid      owner      perms      nsems 0xc13ea218 327057408  oracle    640        2600 ------ Message Queues -------- key        msqid      owner      perms      used-bytes   messages $ ipcs -m -p ------ Shared Memory Creator/Last-op -------- shmid      owner      cpid       lpid 401932289  oracle     53315      53357 401965058  oracle     53315      53357 401997827  oracle     53315      53357 402030597  oracle     53315      53357 402063366  oracle     53315      53357 402096135  oracle     53315      53357 402128904  oracle     53315      53357 402161673  oracle     53315      53357 402194442  oracle     53315      53357 402227211  oracle     53315      53357 402259980  oracle     53315      53357 402292749  oracle     53315      53357 402325518  oracle     53315      53357 402358287  oracle     53315      53357 402391056  oracle     53315      53357 402423825  oracle     53315      53357 402456594  oracle     53315      53357 402489363  oracle     53315      53357 402522132  oracle     53315      53357 402554901  oracle     53315      53357 402587670  oracle     53315      53357 402620439  oracle     53315      53357 402653208  oracle     53315      53357 402685977  oracle     53315      53357 402718746  oracle     53315      53357 402751515  oracle     53315      53357 402784284  oracle     53315      53357 402817053  oracle     53315      53357 402849822  oracle     53315      53357 --//共建立29个Shared Memory Segments.如果你看前面的strace,就可以发现cpid,lpid对应进程号. $ ipcs -m -p | grep oracle|wc      29     116    1276 SYS@test> @ spid  SID    SERIAL# PROCESS SERVER    SPID       PID  P_SERIAL# C50 ---- ---------- ------- --------- ------ ------- ---------- -------------------------------------------------- 3117          1 52843   DEDICATED 53353       19          1 alter system kill session '3117,1' immediate; SYS@test> quit $ grep 53353 /tmp/52843.txt | egrep  "shmat|shmdt" 53353 shmat(402849822, 0, 0)            = ? 53353 shmdt(0x7f9410123000)             = 0 53353 shmat(402849822, 0xfd000000, 0)   = ? 53353 shmat(401932289, 0x60000000, 0)   = ? 53353 shmat(401965058, 0x62000000, 0)   = ? 53353 shmat(401997827, 0x68000000, 0)   = ? 53353 shmat(402030597, 0x6b000000, 0)   = ? 53353 shmat(402063366, 0x6d000000, 0)   = ? 53353 shmat(402096135, 0x73000000, 0)   = ? 53353 shmat(402128904, 0x79000000, 0)   = ? 53353 shmat(402161673, 0x7f000000, 0)   = ? 53353 shmat(402194442, 0x85000000, 0)   = ? 53353 shmat(402227211, 0x8b000000, 0)   = ? 53353 shmat(402259980, 0x91000000, 0)   = ? 53353 shmat(402292749, 0x97000000, 0)   = ? 53353 shmat(402325518, 0x9d000000, 0)   = ? 53353 shmat(402358287, 0xa3000000, 0)   = ? 53353 shmat(402391056, 0xa9000000, 0)   = ? 53353 shmat(402423825, 0xaf000000, 0)   = ? 53353 shmat(402456594, 0xb5000000, 0)   = ? 53353 shmat(402489363, 0xbb000000, 0)   = ? 53353 shmat(402522132, 0xc1000000, 0)   = ? 53353 shmat(402554901, 0xc7000000, 0)   = ? 53353 shmat(402587670, 0xcd000000, 0)   = ? 53353 shmat(402620439, 0xd3000000, 0)   = ? 53353 shmat(402653208, 0xd9000000, 0)   = ? 53353 shmat(402685977, 0xdf000000, 0)   = ? 53353 shmat(402718746, 0xe5000000, 0)   = ? 53353 shmat(402751515, 0xeb000000, 0)   = ? 53353 shmat(402784284, 0xf1000000, 0)   = ? 53353 shmat(402817053, 0xf7000000, 0)   = ? 53353 shmdt(0x60000000)                 = 0 53353 shmdt(0x62000000)                 = 0 53353 shmdt(0x68000000)                 = 0 53353 shmdt(0x6b000000)                 = 0 53353 shmdt(0x6d000000)                 = 0 53353 shmdt(0x73000000)                 = 0 53353 shmdt(0x79000000)                 = 0 53353 shmdt(0x7f000000)                 = 0 53353 shmdt(0x85000000)                 = 0 53353 shmdt(0x8b000000)                 = 0 53353 shmdt(0x91000000)                 = 0 53353 shmdt(0x97000000)                 = 0 53353 shmdt(0x9d000000)                 = 0 53353 shmdt(0xa3000000)                 = 0 53353 shmdt(0xa9000000)                 = 0 53353 shmdt(0xaf000000)                 = 0 53353 shmdt(0xb5000000)                 = 0 53353 shmdt(0xbb000000)                 = 0 53353 shmdt(0xc1000000)                 = 0 53353 shmdt(0xc7000000)                 = 0 53353 shmdt(0xcd000000)                 = 0 53353 shmdt(0xd3000000)                 = 0 53353 shmdt(0xd9000000)                 = 0 53353 shmdt(0xdf000000)                 = 0 53353 shmdt(0xe5000000)                 = 0 53353 shmdt(0xeb000000)                 = 0 53353 shmdt(0xf1000000)                 = 0 53353 shmdt(0xf7000000)                 = 0 53353 shmdt(0xfd000000)                 = 0 --//你可以1个连接要调用shmat attach共享内存段29次,而退出连接一个使用shmdt函数29次.总之尽量避免设置kernel.shmmax过小. 3.测试3: shmmni-> This parameter represents total no.of shared memory segments system wide(4096 is most probably enough) --//此参数表示系统范围内共享内存段的总数(4096很可能足够),我以前使用它与shmall相乘完全是理解错误.因为它正好定义是4096. --//先测试修改kernel.shmmni = 28看看. # grep "^kernel.s[he]" /etc/sysctl.conf kernel.shmall = 4294967296 kernel.shmmax = 104857600 kernel.shmmni = 28 kernel.sem = 2600 332800 2600 128 # sysctl -p SYS@test> startup nomount pfile='/tmp/@.ora'; ORA-27102: out of memory Linux-x86_64 Error: 28: No space left on device Additional information: 100663296 Additional information: 1 --//增加3个看看.注我测试时忽略1个问题,已经存在2个shared memory segments.使用root用户就可以发现,oracle用户看不到如下信息. # ipcs ------ Shared Memory Segments -------- key        shmid      owner      perms      bytes      nattch     status 0x740202db 2523136    root      600        4          0 0x00000000 348422148  gdm       600        393216     2          dest ------ Semaphore Arrays -------- key        semid      owner      perms      nsems ------ Message Queues -------- key        msqid      owner      perms      used-bytes   messages # grep "^kernel.s[he]" /etc/sysctl.conf kernel.shmall = 4294967296 kernel.shmmax = 104857600 kernel.shmmni = 31 kernel.sem = 2600 332800 2600 128 # sysctl -p SYS@test> startup nomount pfile='/tmp/@.ora'; ORACLE instance started. Total System Global Area 2622255104 bytes Fixed Size                  2256112 bytes Variable Size            1124074256 bytes Database Buffers         1476395008 bytes Redo Buffers               19529728 bytes --//OK.这样就可以很准确验证自己的判断. 4.测试4: --//kernel.shmall 参数. shmall-> We specify this parameter in pages. The SHMALL defines the largest amount of shared memory pages that can be          used at one time on the system. --//这个定义最大shared memory pages数量.每页4096.也就是shmall*4096表示当前服务器最大共享内存段大小. # grep "^kernel.s[he]" /etc/sysctl.conf kernel.shmall = 4294967296 kernel.shmmax = 19327352832 kernel.shmmni = 4096 kernel.sem = 2600 332800 2600 128 --//我的定义超大,实际上以前对参数不理解.kernel.shmall = 4294967296 --//4294967296*4096/1024/1024/1024 = 16384G. --//实际上这个参数表示总的sga使用量,也许还要大一些. --//我当前数据库定义 *.sga_target=2500M --//假设定义2400*1024*1024/4096 = 614400,相当于2400M. # grep "^kernel.s[he]" /etc/sysctl.conf kernel.shmall = 614400 kernel.shmmax = 19327352832 kernel.shmmni = 4096 kernel.sem = 2600 332800 2600 128 # sysctl -p SYS@test> startup nomount pfile='/tmp/@.ora'; ORA-27102: out of memory Linux-x86_64 Error: 28: No space left on device Additional information: -1879048192 Additional information: 1 --//如果修改如下:2600*1024*1024/4096 = 665600 # grep "^kernel.s[he]" /etc/sysctl.conf kernel.shmall = 665600 kernel.shmmax = 19327352832 kernel.shmmni = 4096 kernel.sem = 2600 332800 2600 128 # sysctl -p SYS@test> startup nomount pfile='/tmp/@.ora'; ORACLE instance started. Total System Global Area 2622255104 bytes Fixed Size                  2256112 bytes Variable Size            1124074256 bytes Database Buffers         1476395008 bytes Redo Buffers               19529728 bytes --//OK启动正常. # ipcs -m| grep "^0x" | awk '{print $5/4096}' 0.000976562 8192 40960 4096 96 589824 512 # ipcs -m| grep "^0x" | awk '{print $5/4096}'| paste -sd+| bc -l 643680.000976562 --//也就是我定义643681可以启动数据库,小于643681无法启动数据库. # grep "^kernel.s[he]" /etc/sysctl.conf kernel.shmall = 643681 kernel.shmmax = 19327352832 kernel.shmmni = 4096 kernel.sem = 2600 332800 2600 128 # sysctl -p SYS@test> startup nomount pfile='/tmp/@.ora'; ORACLE instance started. Total System Global Area 2622255104 bytes Fixed Size                  2256112 bytes Variable Size            1124074256 bytes Database Buffers         1476395008 bytes Redo Buffers               19529728 bytes --//OK可以启动数据库. --//减少1个看看. # grep "^kernel.s[he]" /etc/sysctl.conf kernel.shmall = 643680 kernel.shmmax = 19327352832 kernel.shmmni = 4096 kernel.sem = 2600 332800 2600 128 # sysctl -p SYS@test> startup nomount pfile='/tmp/@.ora'; ORA-27102: out of memory Linux-x86_64 Error: 28: No space left on device Additional information: 2097152 --//OK.验证自己的判断. --//还原内核参数. # grep "^kernel.s[he]" /etc/sysctl.conf kernel.shmmax = 68719476736 kernel.shmall = 4294967296  -->不改小了. kernel.shmmni = 4096 kernel.sem = 2600 332800 2600 128 # sysctl -p 5.sysrev命令 --//另外oracle提供1个sysresv 命令. $ sysresv -h sysresv: invalid option -- h usage   : sysresv [-if] [-d <on/off>] [-l sid1 <sid2> ...]           -i : Prompt before removing ipc resources for each sid           -f : Remove ipc resources silently, oevrrides -i option           -d <on/off> : List ipc resources for each sid if on           -l sid1 <sid2> .. : apply sysresv to each sid Default : sysresv -d on -l $ORACLE_SID Note    : ipc resources will be attempted to be deleted for a           sid only if there is no currently running instance           with that sid. $ sysresv  -l test IPC Resources for ORACLE_SID "test" : Shared Memory: ID              KEY 406683649       0x00000000 406716418       0x00000000 406749187       0x00000000 406781957       0x00000000 406814726       0x8a931fb8 Semaphores: ID              KEY 327647232       0xc13ea218 Oracle Instance alive for sid "test" --//才发现sysresv命令-f参数可以Remove ipc resources.而且这个命令相对安全. $ sysresv  -f -l test IPC Resources for ORACLE_SID "test" : Shared Memory: ID              KEY 406683649       0x00000000 406716418       0x00000000 406749187       0x00000000 406781957       0x00000000 406814726       0x8a931fb8 Semaphores: ID              KEY 327647232       0xc13ea218 Oracle Instance alive for sid "test" SYSRESV-005: Warning         Instance maybe alive - aborting remove for sid "test" --//监测alive,移除ipc失败. $ sysresv  -if -l test IPC Resources for ORACLE_SID "test" : Shared Memory: ID              KEY 406683649       0x00000000 406716418       0x00000000 406749187       0x00000000 406781957       0x00000000 406814726       0x8a931fb8 Semaphores: ID              KEY 327647232       0xc13ea218 Oracle Instance alive for sid "test" SYSRESV-005: Warning         Instance maybe alive - aborting remove for sid "test" 总结: --//剩下kernel.sem = SEMMSL SEMMNS SEMOPM SEMMNI没写,下一篇了.再次摘抄前面内容: shmmax-> This parameter represents the size in bytes of a single shared memory segment shmmni-> This parameter represents total no.of shared memory segments system wide(4096 is most probably enough) shmall-> We specify this parameter in pages. The SHMALL defines the largest amount of shared memory pages that can be          used at one time on the system. --//我自己仅仅希望这些测试很好理解这些参数的具体含义以及ipcs的相关输出.