遭遇ORA-07445[ACCESS_VIOLATION][unable_to_trans_pc

像平常一样，准备启动自己机器上的测试数据库看点东西，net start oracleservicening，正常，sqlplus以sysdba登入，奇怪，这次怎么10秒了还没反映，嗯，机器 不好，先做点别的事情吧。 几分钟后再来看看，还是挂起在那里，不对劲，第一反应，看alert，果然有

像平常一样，准备启动自己机器上的测试数据库看点东西，net start oracleservicening，正常，sqlplus以sysdba登入，奇怪，这次怎么10秒了还没反映，嗯，机器不好，先做点别的事情吧。

几分钟后再来看看，还是挂起在那里，不对劲，第一反应，看alert，果然有机关：

PMON started with pid=2, OS id=404
PSP0 started with pid=3, OS id=888
MMAN started with pid=4, OS id=3628
DBW0 started with pid=5, OS id=148
LGWR started with pid=6, OS id=3944
CKPT started with pid=7, OS id=252
SMON started with pid=8, OS id=2336
RECO started with pid=9, OS id=2452
CJQ0 started with pid=10, OS id=3772
MMON started with pid=11, OS id=3728
MMNL started with pid=12, OS id=3204
Fri Mar 16 09:09:47 2007
Errors in file e:oracleora10adminningudumpning_ora_3848.trc:
ORA-07445: exception encountered: core dump [ACCESS_VIOLATION] [unable_to_trans_pc] [PC:0x77E7FA69] [ADDR:0x7D42] [UNABLE_TO_WRITE] []

晕，一直好好的，怎么突然出07445错误了，而且google，metalink转了一遍，也没发现什么有价值的地方。回想之前，唯一的改动好像是将

buffer cache调小了点，内存不够用，没办法。alert再往前看看

__db_cache_size = 41943040
db_cache_size = 41943040

嗯，只有40M，记得之前好像是48M，一直用的spfile，之前也没有保存pfile，只有先startup nomout，然后create pfile from spfile来生成一个了。

停止服务，修改注册表，将启动服务时自动startup数据库的选项改为false，然后重新启动服务

C:>sqlplus /nolog

SQL*Plus: Release 10.2.0.3.0 - Production on Fri Mar 16 09:15:04 2007

Copyright (c) 1982, 2006, Oracle. All Rights Reserved.

@>conn / as sysdba
Connected to an idle instance.

SYS@ning>startup nomount;
ORA-24324: service handle not initialized
ORA-24323: value not allowed
ORA-03113: end-of-file on communication channel

我晕，这招不灵光，根本instance都起不来，也对，如果是内存参数有问题，起不来也正常。看来不能偷懒，还是利用alert里面记录的非默认参数自己弄一个pfile了，alert真是信息丰富啊:)

System parameters with non-default values:
processes = 150
__shared_pool_size = 83886080
shared_pool_size = 83886080
large_pool_size = 0
java_pool_size = 0
control_files = E:ORACLEORADATANINGCONTROL01.CTL, E:ORACLEORADATANINGCONTROL02.CTL,

E:ORACLEORADATANINGCONTROL03.CTL
db_block_size = 8192
__db_cache_size = 41943040
db_cache_size = 41943040
compatible = 10.2.0.1.0
db_file_multiblock_read_count= 16
dml_locks = 2
undo_management = AUTO
undo_tablespace = UNDOTBS1
remote_login_passwordfile= EXCLUSIVE
db_domain =
job_queue_processes = 10
audit_file_dest = E:ORACLEORA10ADMINNINGADUMP
background_dump_dest = E:ORACLEORA10ADMINNINGBDUMP
user_dump_dest = E:ORACLEORA10ADMINNINGUDUMP
core_dump_dest = E:ORACLEORA10ADMINNINGCDUMP
db_name = ning
open_cursors = 300
pga_aggregate_target = 95420416

SYS@ning>startup nomount pfile='d:initning.ora';
ORA-24324: service handle not initialized
ORA-24323: value not allowed
ORA-03113: end-of-file on communication channel

偶分特，还不行。这次alert里面还报了一堆没有参数的额07445了
Fri Mar 16 09:53:54 2007
Errors in file e:oracleora10adminningudumpning_ora_3812.trc:
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [] [] [] [] [] []
ORA-07445: exception encountered: core dump [ACCESS_VIOLATION] [unable_to_trans_pc] [PC:0x7C911F52] [ADDR:0x6C63616F]

[UNABLE_TO_READ] []
ORA-07445: exception encountered: core dump [ACCESS_VIOLATION] [unable_to_trans_pc] [PC:0x7C911F52] [ADDR:0x6C63616F]

[UNABLE_TO_READ] []

嗯，不好意思，刚才好像只改了db_cache_size，忘记改__db_cache_size了，再来

SYS@ning>startup nomount pfile='d:initning.ora';
ORACLE instance started.

Total System Global Area 142606336 bytes
Fixed Size 1289340 bytes
Variable Size 83886980 bytes
Database Buffers 50331648 bytes
Redo Buffers 7098368 bytes

SYS@ning>alter database mount;

Database altered.

SYS@ning>alter database open;

Database altered.

终于起来了，哎，不过将buffer cache改小了8M，至于弄个07445的错误来吓我么。

重新生成spfile

SYS@ning>create spfile from pfile;

File created.

不过这也告诉我们，生产库上的参数，如果没有经过充分测试，千万不要随便乱改，尤其是一些高危参数，比如内存大小啊，cursor_sharing啊什么的。并且做任何改变之前，先做好文档记录，万一有问题，还能追查到出问题之前对于系统的改动情况，这个对于排错是有相当大的帮助的。