一、概述

在Linux上安装oracle，需要对内核参数进行调整，其中有shmmax和shmall这两个参数，那这两个参数是什么意思，又该如何设置呢？

二、官方文档

在oracle的官方文档（ https://docs.oracle.com/en/database/oracle/oracle-database/19/ladbi/minimum-parameter-settings-for-installation.html#GUID-CDEB89D1-4D48-41D9-9AC2-6AD9B0E944E3 ）中对这两个参数，设置了最小的标准值。
微信截图_20220406104623.png
shmall - Greater than or equal to the value of shmmax, in pages.
shmmax - Half the size of physical memory in bytes. See My Oracle Support Note 567506.1 for additional information about configuring shmmax.

再根据redhat的官方文档（ https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html/performance_tuning_guide/sect-red_hat_enterprise_linux-performance_tuning_guide-configuration_tools-configuring_system_memory_capacity ），去查这两个参数所表达的含义。
微信截图_20220406113052.png
shmall - Defines the total amount of shared memory pages that can be used on the system at one time. A page is 4096 bytes on the AMD64 and Intel 64 architecture, for example.
shmmax - Defines the maximum size (in bytes) of a single shared memory segment allowed by the kernel.

以上两段英文翻译过来：shmmax单个最大共享内存段，shmall同一时刻能使用的所有共享内存页。shmmax最小一半的物理内存，shmall >= shmmax/4096。

oracle的sga（Shared Global Area）使用的就是共享内存，共享内存的优势redhat官方文档（ https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/5/html/tuning_and_optimizing_red_hat_enterprise_linux_for_oracle_9i_and_10g_databases/chap-oracle_9i_and_10g_tuning_guide-setting_shared_memory#sect-Oracle_9i_and_10g_Tuning_Guide-Setting_Shared_Memory-Setting_SHMMAX_Parameter_ ）中也有提及。直白点说就是多进程使用共享内存交流数据最快。例如：服务器进程从磁盘读取数据到sga的buffer cache，dbwn进程从buffer cache将数据写回到磁盘，操作的是同一片内存区域。如果没有共享内存，那么就需要将服务器进程操作的这片内存复制一份到dbwn所操作的内存中去，来完成读取和写入操作。
微信截图_20220406113004.png
Shared memory allows processes to access common structures and data by placing them in shared memory segments. It is the fastest form of inter-process communication available since no kernel involvement occurs when data is passed between the processes. In fact, data does not need to be copied between the processes.
Oracle uses shared memory segments for the Shared Global Area (SGA) which is an area of memory that is shared by Oracle processes. The size of the SGA has a significant impact to Oracle’s performance since it holds database buffer cache and much more.

从上面的官方文档我们了解了这两个参数的含义，但是oracle只给了shmmax和shmall的最小值。接下来我们就通过实验来看看这两个参数对oracle的影响。

三、实验

我的实验机器物理内存是1877M，设置SGA_TAEGET为1000M。接下来测试几个场景。

a. shmmax 200M, shmall 200M

将/etc/sysctl.conf参数设置为
kernel.shmmax = 209715200
kernel.shmall = 51200

oracle启动直接报错

SQL> startup nomount pfile='/home/oracle/test.ora'
ORA-27102: out of memory
Linux-x86\_64 Error: 28: No space left on device
Additional information: 209715200
Additional information: 1

b. shmmax 1200M, shmall 200M

将/etc/sysctl.conf参数设置为
kernel.shmmax = 1258291200
kernel.shmall = 51200

oracle启动报跟上面一样的错

SQL> startup nomount pfile='/home/oracle/test.ora' 
ORA-27102: out of memory
Linux-x86\_64 Error: 28: No space left on device
Additional information: 1035993088
Additional information: 1

从a和b的实验结果来看，oracle是否能够正常启动跟shmmax参数无关，只与shmall有关。shmall不能设置的比SGA_TAEGET小。

c. shmmax 200M, shmall 1200M

将/etc/sysctl.conf参数设置为
kernel.shmmax = 209715200
kernel.shmall = 307200

数据库能够正常启动

SQL> startup nomount pfile='/home/oracle/test.ora'
ORACLE instance started.Total System Global Area 1043886080 bytes
Fixed Size                  2259840 bytes
Variable Size             327156864 bytes
Database Buffers          708837376 bytes
Redo Buffers                5632000 bytes

查看共享内存的信息

[root@oracletest ~]# ipcs -m------ Shared Memory Segments --------
key        shmid      owner      perms      bytes      nattch     status      
0x00000000 229376     oracle     640        12582912   18                      
0x00000000 262145     oracle     640        209715200  18                      
0x00000000 294914     oracle     640        209715200  18                      
0x00000000 327683     oracle     640        209715200  18                      
0x00000000 360452     oracle     640        209715200  18                      
0x00000000 393221     oracle     640        197132288  18                      
0x276f5044 425990     oracle     640        2097152    18

把上面的共享内存段bytes全部加起来（12582912+209715200…+2097152）/1024/1024=1002MB。可以看到oracle分配内存段的时候，单个共享内存段的确没有超过shmmax（209715200）。总的共享内存刚好等于SGA_TAEGET。

d. shmmax 1200M, shmall 1200M

将/etc/sysctl.conf参数设置为
kernel.shmmax = 1258291200
kernel.shmall = 307200

数据库同样能够正常启动

SQL> startup nomount pfile='/home/oracle/test.ora'
ORACLE instance started.Total System Global Area 1043886080 bytes
Fixed Size                  2259840 bytes
Variable Size             327156864 bytes
Database Buffers          708837376 bytes
Redo Buffers                5632000 bytes

查看共享内存的信息

[root@oracletest ~]# ipcs -m------ Shared Memory Segments --------
key        shmid      owner      perms      bytes      nattch     status      
0x00000000 557056     oracle     640        12582912   18                      
0x00000000 589825     oracle     640        1035993088 18                      
0x276f5044 622594     oracle     640        2097152    18

把上面的共享内存段bytes全部加起来（12582912+1035993088+2097152）/1024/1024=1002MB。总的共享内存仍然刚好等于SGA_TAEGET。内存段的数量却只有三个，最大的内存段达到1035993088/1024/1024=988M

f. shmmax 2400M, shmall 2400M

将/etc/sysctl.conf参数设置为
kernel.shmmax = 2516582400
kernel.shmall = 614400

SQL> startup nomount pfile='/home/oracle/test.ora'
ORACLE instance started.Total System Global Area 1043886080 bytes
Fixed Size                  2259840 bytes
Variable Size             327156864 bytes
Database Buffers          708837376 bytes
Redo Buffers                5632000 bytes

[root@oracletest ~]# ipcs -m------ Shared Memory Segments --------
key        shmid      owner      perms      bytes      nattch     status      
0x00000000 688128     oracle     640        12582912   18                      
0x00000000 720897     oracle     640        1035993088 18                      
0x276f5044 753666     oracle     640        2097152    18

可以看到f跟e没啥区别，shmmax这个值你就算设置超过了物理内存也不受影响。因为oracle实际上分配的共享内存不会超过SGA_TAEGET。