有人试过在aix 5.2 ML06+oracle9i+jfs2文件系统下打开jfs2的concurrent i/o功能吗？

wuchengeng

数据库使用的是dedicate server。YanBing兄如果能上metalink,帮我下228059.1的文档。谢谢！我的msn:wuchengeng1@126.com

wuchengeng

help!!

wuchengeng

关于涉及到oracle方面问题及日志，见我itpub上的贴子：http://www.itpub.net/showthread. ... 10&pagenumber=1

yanbing

Subject:  ALERT: Jobs Scheduled with DBMS_JOB May Not Run at Scheduled Intervals
  Doc ID:  Note:228059.1 Type:  ALERT
  Last Revision Date:  20-OCT-2005 Status:  PUBLISHED


ALERT: Jobs Scheduled with DBMS_JOB May Not Run at Scheduled Intervals
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Versions Affected
~~~~~~~~~~~~~~~~~
This issue is introduced in Oracle9i in version 9.0.1.0.  It affects version
9.0.1.0 through 9.0.1.4 inclusive and versions 9.2.0.1 and 9.2.0.2.

Platforms Affected
~~~~~~~~~~~~~~~~~~
  GENERIC.

Description
~~~~~~~~~~~  
  An attempt is made herein to increase the visibility of Bug 2649244 which
  many customers could face as they move on to OEE 9.0.1.x or 9.2.x.
  It is intended for customers who use the DBMS_JOB API which is an integral
  part of replication and also is used to schedule other administrative
  activities.  The symptoms associated with this bug are that jobs may not
  run at scheduled intervals thus holding up other processes.

Likelihood of Occurrence
~~~~~~~~~~~~~~~~~~~~~~~~
  If you are using DBMS_JOB API to schedule jobs then this problem could
  affect you.  To check if you are hitting this bug:

     SQL> SELECT last_date, last_sec, next_date, next_sec FROM dba_jobs;

  See if the next_date, next_sec are updated after a job is run.

  If not you would have to run the job manually using DBMS_JOB.RUN(jobid).

Possible Symptoms
~~~~~~~~~~~~~~~~~
  a) JOBS may not run at scheduled intervals even if the job is not BROKEN.
  b) The job coordinator CJQ process produces trace files with stack trace
     as   kkjqawi kkjawiq kkjex1s kkjcjexe kkjssrh ksbcti
  c) Next_date in DBA_JOBS is not updated.

Workaround
~~~~~~~~~~
  Set job_queue_processes=0 and set it back to the old value after a couple
  of minutes.  When we set job_queue_processes=0 this would kill all the snp
  related processes and then after a couple of minutes set the value of
  job_queue_processes to the original value.

  For example, show parameter job_queue_processes and if the output is say 5

      SQL> ALTER SYSTEM SET JOB_QUEUE_PROCESSES = 0;

  would kill all Jxxx processes.

  After 2-3 minutes;

      SQL> ALTER SYSTEM SET JOB_QUEUE_PROCESSES = 5;

  Alternatively use the cron utility to run the jobs.

Patches
~~~~~~~
  Fix to this bug is addressed in patchset 9.2.0.3.0 and Oracle Server
  Release 10i.

  Patches for this bug can be found on MetaLink by following these steps:
    1. Login to MetaLink - http://metalink.oracle.com
    2. Choose Patches from the Menu.
    3. Input 2649244 into the Patch Number field.
    4. Click the Go button.

References
~~~~~~~~~~
  Bug 2649244 JOBS may not run at scheduled intervals   
______________________________________________________________________________
                                                      Oracle Support Services

wuchengeng

事情似乎有了进展！！！
    用户发现原因只要执行完exp之后，数据库就开始变慢，最终导致客户端无法登录。这个exp作业以前也是一直在执行的。
    但问题的原因更让人迷惑：
    1.以前在jfs格式下，这个exp作业一直正常
    2.修改了jfs2格式之后的一个多星期内也没有出现问题
为什么现在突然变成这样？难道这只是一个表面现象么？

corysave

你可能要去看smit -->system enviroment-->Change / Show Characteristics of Operating System 中
  high water mark and low water mark 選項是否為0 ,if 為0 表示在大量copy 時會用到所有I/O Block ,此時主機會變慢 ,標準參數值
   high water mark  = 33
    low water mark    = 24