网上很多关于这个错误的文章，有些还引用了Oracle文档中的说明。
Setting Memory Target at Instance Startup on Linux
Starting with Oracle Database 11g Release 1 (11.1), Oracle provides the option of automatically managing SGA and PGA with a combined MEMORY_TARGETparameter without having to set SGA_TARGET and PGA_AGGREGATE_TARGET explicitly. This is supported on Linux, Windows, Solaris, HPUX, and AIX (reference Bug 7258378).
If you see the ORA-00845 error reported on Linux machines at Oracle instance startup when using the MEMORY_TARGET parameter, then check the size of /dev/shm. If /dev/shm is not configured, then mount it sized to be at least the value of MEMORY_TARGET. If /dev/shm is configured but the amount of available space reported (through df -k /dev/shm) is less than MEMORY_TARGET, then free the space or mount a larger /dev/shm to satisfy the MEMORY_TARGET size. Note that if you set theMEMORY_MAX_TARGET parameter greater than MEMORY_TARGET, then ensure that /dev/shm is sized to be at least the value of MEMORY_MAX_TARGET.
Memory Target for Oracle Database InstancesRunning Database Configuration Assistant (DBCA) defaults to this Automatic Memory Management option. In the case of upgrade or manual database creation,MEMORY_TARGET can be specified in the initialization parameter file.
大部分人理解为，要求/dev/shm至少必须不小于MEMORY_MAX_TARGET的大小。
但是根据测试的实际情况，必须是/dev/shm的可用空间不小于MEMORY_MAX_TARGET的大小。
例如
SQL> show parameter target
NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
archive_lag_target integer 0
db_flashback_retention_target integer 1440
fast_start_io_target integer 0
fast_start_mttr_target integer 0
memory_max_target big integer 6336M
memory_target big integer 6336M
parallel_servers_target integer 16
pga_aggregate_target big integer 0
sga_target big integer 0
根据参数，memory_max_target 参数为6.18G
[[email protected] ~]# mount -o remount,size=10G /dev/shm
[[email protected] ~]# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/vg_centosdbo-lv_root 50G 42G 5.3G 89% /
tmpfs 10G 9.9G 144M 99% /dev/shm
/dev/sda2 485M 39M 421M 9% /boot
/dev/sda1 200M 260K 200M 1% /boot/efi
/dev/mapper/vg_centosdbo-lv_home 1.3T 504G 744G 41% /home
[[email protected] ~]# su – oracle
即便这里我设置size大小为10G。但是可用空间只有144M.
这时我启动数据库：
SQL*Plus: Release 11.2.0.1.0 Production on Wed Mar 30 14:29:21 2016
Copyright (c) 1982, 2009, Oracle. All rights reserved.
SQL> conn /as sysdba
Connected to an idle instance.
SQL> startup
ORA-00845: MEMORY_TARGET not supported on this system
仍然会报错。
如下所示，如果保证可用空间大于memory_max_target，则数据库正常启动。
[[email protected] ~]# mount -o remount,size=16G /dev/shm
[[email protected] ~]# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/vg_centosdbo-lv_root 50G 42G 5.3G 89% /
tmpfs 16G 6.2G 9.9G 39% /dev/shm
/dev/sda2 485M 39M 421M 9% /boot
/dev/sda1 200M 260K 200M 1% /boot/efi
/dev/mapper/vg_centosdbo-lv_home 1.3T 504G 744G 41% /home
[[email protected] ~]# sqlplus /nolog
-bash: sqlplus: command not found
[[email protected] ~]# su - oracle
[[email protected] ~]$ sqlplus /nolog
SQL*Plus: Release 11.2.0.1.0 Production on Wed Mar 30 14:30:41 2016
Copyright (c) 1982, 2009, Oracle. All rights reserved.
SQL> conn /as sysdba
Connected to an idle instance.
SQL> startup
ORACLE instance started.
Total System Global Area 6614106112 bytes
Fixed Size 2213896 bytes
Variable Size 3556771832 bytes
Database Buffers 3019898880 bytes
Redo Buffers 35221504 bytes
Database mounted.
Database opened.
SQL>
由此可能的情况是，/dev/shm默认为系统内存的一半，虽然理论上可能已经大于MEMORY_TARGET，但数据库启动是可能仍然会报错。
比如我所用的测试库，MEMORY_TARGET是6.18G
而默认/dev/shm为内存的一半，即7.7G左右。但数据库启动时仍然报错。
[[email protected] ~]# free -g
total used free shared buffers cached
Mem: 15 11 4 0 0 10
-/+ buffers/cache: 0 14
Swap: 7 0 7
[[email protected] ~]#