共计 7367 个字符,预计需要花费 19 分钟才能阅读完成。
环境:AIX 6.1 + Oracle 10.2.0.4
现象:在 XTTS 迁移测试阶段,遇到执行几个 expdp 的导出任务,迟迟没有返回任何信息,对应日志无任何输出,查看任务状态:
SQL>
set lines 300
col OWNER_NAME for a10
col OPERATION for a15
col JOB_MODE for a20
col STATE for a15
select * from dba_datapump_jobs;
OWNER_NAME JOB_NAME OPERATION JOB_MODE STATE DEGREE ATTACHED_SESSIONS DATAPUMP_SESSIONS
———- —————————— ————— ——————– ————— ———- —————– —————–
SYS SYS_EXPORT_TRANSPORTABLE_01 EXPORT TRANSPORTABLE DEFINING 1 0 1
SYS SYS_EXPORT_TRANSPORTABLE_02 EXPORT TRANSPORTABLE DEFINING 1 1 2
SYS SYS_EXPORT_TRANSPORTABLE_03 EXPORT TRANSPORTABLE DEFINING 1 1 2
SYS SYS_EXPORT_SCHEMA_01 EXPORT SCHEMA DEFINING 1 1 2
SYS SYS_EXPORT_TRANSPORTABLE_04 EXPORT TRANSPORTABLE DEFINING 1 1 2
SYS SYS_EXPORT_SCHEMA_02 EXPORT SCHEMA DEFINING 1 1 2
6 rows selected.
可以看到所有的 expdp 导出任务的 STATE 都停留在 DEFINING 状态。
•1. 牛刀小试清异常
•2. 追本溯源查 MOS
1. 牛刀小试清异常
先强制杀掉后台执行的所有 expdp 任务:
ps -ef|grep expdp|grep -v grep|awk ‘{print $2}’|xargs kill -9
然后尝试删除这些表(其实应该在 not running 状态下删除)
select ‘drop table ‘||OWNER_NAME||’.’||JOB_NAME||’ purge;’ from dba_datapump_jobs where STATE=’NOT RUNNING’;
drop table sys.SYS_EXPORT_TRANSPORTABLE_01 purge;
..
可这样是没有作用的,查询结果不变。
甚至尝试正常 shutdown immediate 停止数据库,也无法成功,告警日志看到有活动调用:
Thu Nov 1 15:14:24 2018
Active call for process 4522064 user ‘oracle’ program ‘oracle@localhost (DM00)’
Active call for process 4456536 user ‘oracle’ program ‘oracle@localhost (DM01)’
Active call for process 10027180 user ‘oracle’ program ‘oracle@localhost (DM02)’
Active call for process 7340140 user ‘oracle’ program ‘oracle@localhost (DM03)’
Active call for process 6291888 user ‘oracle’ program ‘oracle@localhost (DM04)’
Active call for process 8126596 user ‘oracle’ program ‘oracle@localhost (DM05)’
SHUTDOWN: waiting for active calls to complete.
发现这些进程的 id 都对应了 ora_dm 的进程:
$ ps -ef|grep ora_dm
oracle 4456536 1 0 17:00:09 – 0:00 ora_dm01_xxxxdb
oracle 4522064 1 0 16:50:57 – 0:00 ora_dm00_xxxxdb
oracle 7340140 1 0 14:06:07 – 0:00 ora_dm03_xxxxdb
oracle 8126596 1 0 14:35:03 – 0:00 ora_dm05_xxxxdb
oracle 10027180 1 0 13:55:08 – 0:00 ora_dm02_xxxxdb
oracle 6291888 1 0 14:31:17 – 0:00 ora_dm04_xxxxdb
oracle 7340432 8388786 0 15:22:59 pts/4 0:00 grep ora_dm
实际上,这就是 expdp 任务的相关进程,强制杀掉这些进程:
ps -ef|grep ora_dm|grep -v grep|awk ‘{print $2}’|xargs kill -9
之后数据库关闭成功:
Thu Nov 1 15:24:37 2018
All dispatchers and shared servers shutdown
Thu Nov 1 15:24:37 2018
ALTER DATABASE CLOSE NORMAL
启动数据库后,再次查询发现已经成功清理:
SQL>
set lines 300
col OWNER_NAME for a10
col OPERATION for a15
col JOB_MODE for a20
col STATE for a15
select * from dba_datapump_jobs;
no rows selected
小结:数据泵任务与 ora_dm 进程相关;如果数据泵任务发生异常,但任务并没有退出的情况,需要同时杀掉这类进程(杀掉后状态就会变为 NOT RUNNING)。关库不是必须的,只是演示此时正常关闭被阻塞的场景。这也能说明为什么要保证在 NOT RUNNING 状态下才可以清理。
2. 追本溯源查 MOS
上面的步骤只是清理了异常的数据泵任务,但没有解决问题,再次后台执行备份任务依然会重现故障:
nohup sh expdp_xtts.sh &
$ ps -ef|grep expdp
oracle 6684914 8061208 0 15:30:07 pts/2 0:00 grep expdp
oracle 7143482 8061208 0 15:30:03 pts/2 0:00 sh expdp_xtts.sh
oracle 6685096 7143482 0 15:30:03 pts/2 0:00 expdp ‘/ as sysdba’ parfile=expdp_xtts.par
$ ps -ef|grep ora_dm
oracle 7602308 8061208 0 15:30:10 pts/2 0:00 grep ora_dm
oracle 3997964 1 1 15:30:05 – 0:00 ora_dm00_xxxxdb
$
此时查询 dba_datapump_jobs,state 依然一直是 defining 状态:
OWNER_NAME JOB_NAME OPERATION JOB_MODE STATE DEGREE ATTACHED_SESSIONS DATAPUMP_SESSIONS
———- —————————— ————— —————————— —————————— ———- —————– —————–
SYS SYS_EXPORT_TRANSPORTABLE_01 EXPORT TRANSPORTABLE DEFINING 1 1 2
其他的导出任务都一样,不再赘述。
为了方便测试,写一个简单的单表 expdp 导出,现象也一样。
expdp \’/ as sysdba\’ directory=XTTS tables=query.test dumpfile=query_test.dmp logfile=query_test.log
根据故障现象,用如下关键字在 MOS 中搜索: expdp state DEFINING,匹配到文档:
•DataPump Export/Import Hangs With “DEFINING” Status When Using A Directory On NFS Filesystem (文档 ID 2262196.1)
正好这次测试是在 NFS 文件系统上,MOS 建议移动到本地文件系统导出。
这次再将 expdp 进程全部杀掉:
ps -ef|grep ora_dm|grep -v grep|awk ‘{print $2}’|xargs kill -9
ps -ef|grep expdp|grep -v grep|awk ‘{print $2}’|xargs kill -9
此时查询 dba_datapump_jobs:
OWNER_NAME JOB_NAME OPERATION JOB_MODE STATE DEGREE ATTACHED_SESSIONS DATAPUMP_SESSIONS
———- —————————— ————— —————————— ————— ———- —————– —————–
SYS SYS_EXPORT_TABLE_04 EXPORT TABLE NOT RUNNING 0 0 0
SYS SYS_EXPORT_SCHEMA_01 EXPORT SCHEMA NOT RUNNING 0 0 0
SYS SYS_EXPORT_TABLE_02 EXPORT TABLE NOT RUNNING 0 0 0
SYS SYS_EXPORT_TABLE_05 EXPORT TABLE NOT RUNNING 0 0 0
SYS SYS_EXPORT_TABLE_03 EXPORT TABLE NOT RUNNING 0 0 0
SYS SYS_EXPORT_TABLE_01 EXPORT TABLE NOT RUNNING 0 0 0
SYS SYS_EXPORT_TRANSPORTABLE_01 EXPORT TRANSPORTABLE NOT RUNNING 0 0 0
7 rows selected.
清理 NOT RUNNING 的 master 表:
select ‘drop table ‘||OWNER_NAME||’.’||JOB_NAME||’ purge;’ from dba_datapump_jobs where STATE=’NOT RUNNING’;
– 执行结果用来执行,再次查看结果为空:
SQL> select * from dba_datapump_jobs;
no rows selected
按 MOS 建议,将导出任务移动到本地文件系统:
AIX 源端导出 XTTS 源数据至源端 /hxbak/xtts_exp 目录中,而后 copy 至 nfs 共享存储 /xtts/dmp 中:
mkdir /hxbak/xtts_exp
chown oracle:dba /hxbak/xtts_exp
ls -ld /hxbak/xtts_exp
select * from dba_directories;
create or replace directory XTTS as ‘/hxbak/xtts_exp’;
此时测试 expdp 任务可正常运行:
$ expdp \’/ as sysdba\’ directory=XTTS tables=query.test dumpfile=query_test.dmp logfile=query_test.log
Export: Release 10.2.0.4.0 – 64bit Production on Thursday, 01 November, 2018 16:03:21
Copyright (c) 2003, 2007, Oracle. All rights reserved.
Connected to: Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 – 64bit Production
With the Partitioning, OLAP, Data Mining and Real Application Testing options
Starting “SYS”.”SYS_EXPORT_TABLE_01″: ‘/******** AS SYSDBA’ directory=XTTS tables=query.test dumpfile=query_test.dmp logfile=query_test.log
Estimate in progress using BLOCKS method…
Processing object type TABLE_EXPORT/TABLE/TABLE_DATA
Total estimation using BLOCKS method: 8 MB
Processing object type TABLE_EXPORT/TABLE/TABLE
Processing object type TABLE_EXPORT/TABLE/STATISTICS/TABLE_STATISTICS
. . exported “QUERY”.”TEST” 6.743 MB 72593 rows
Master table “SYS”.”SYS_EXPORT_TABLE_01″ successfully loaded/unloaded
******************************************************************************
Dump file set for SYS.SYS_EXPORT_TABLE_01 is:
/hxbak/xtts_exp/query_test.dmp
Job “SYS”.”SYS_EXPORT_TABLE_01″ successfully completed at 16:03:57
SQL> select * from dba_datapump_jobs;
OWNER_NAME JOB_NAME OPERATION JOB_MODE STATE DEGREE ATTACHED_SESSIONS DATAPUMP_SESSIONS
———- —————————— ————— —————————— ————— ———- —————– —————–
SYS SYS_EXPORT_TABLE_01 EXPORT TABLE EXECUTING 1 1 3
再次导出其他元数据:
#expdp_xtts.sh (about 5min)
nohup sh expdp_xtts.sh &
#expdp_xtts_other.sh(about 5min)
nohup sh expdp_xtts_other.sh &
#expdp_tmp_table
nohup sh expdp_tmp_table01.sh &
nohup sh expdp_tmp_table02.sh &
nohup sh expdp_tmp_table03.sh &
nohup sh expdp_tmp_table04.sh &
最后将这些导出文件再移动到 /xtts/dmp/ 下,供后续 xtts 测试目标端导入使用:
$ pwd
/hxbak/xtts_exp
$ cp -rp * /xtts/dmp/
目标端导入时只需要有读这些文件的权限,即可,实际测试恢复 OK。
小结:在自己的 linux 环境测试过是可以直接 expdp 到 nfs 文件系统的,AIX 看来有区别,MOS 的建议也只是一个 workaround,但也能满足需求,毕竟元数据导出文件没多大。
: