|
# 42941
Chi
来自: Mexico
|
|
|
 To: Eygle
When you answered another post about the iowait%, you said the 25% was not high.However, from one of reference "http://www.aixtips.com/AIXtip/iowait.htm";, the author think > 15% is high according to the filemon monitor (on the AIX OS) and therefore need to do something.What is your iowait% was based on?
If the db_block_lru_latches is a default on 10g R2? I could not find it under my system. So I can play with the db_writer_process.Because according to 10g R2 admin guide, both parameters are related.Am I wrong or missed something?My AIX has 2 CPU but partitioned into 4.
p.s Tried more of your Canon 400 yet.
From: Chi
2007.06.15 02:56
|
|
|
|
|
To: Chi
我没仔细看你说的tips.
我的意思是，iowait在25%，说明系统的IO能力还远没达到满负荷，在这种情况下系统已经出现严重的性能问题，一定说明其他资源或设置上存在问题，还并未充分利用IO。
From: eygle
2007.06.18 09:03
|
|
|
|
|
# 42932
stockxs333
|
|
|
To: eygle
你好：上次的问题（42930），按照你的建议（查看是否IO瓶颈）用iostat -x 5查看在程序插入挂掉时 %iowait为 25，wkB/s 为4000--7000，我从其他资料看到iowait大于15以上就说明IO有瓶颈， 可否认为只有加磁盘，这个问题才可以最终解决？
From: stockxs333
2007.06.12 17:37
|
|
|
|
|
To: stockxs333
1.你的事务是否可以分次提交?不要一次插入那么多数据。
2.iowait 25%不算高，可以继续增加使用多个dbwr，看i/o能不能上去
3.将redo/arch放到和数据文件不同的物理硬盘上，降低i/o竞争。
From: eygle
2007.06.13 08:40
|
|
|
|
|
# 42931
Alex
|
|
|
To: eygle
我又来了请教你了.
现在想系统学下 oracle 想考个OCP 但,不知道考 9i 还是10g 呢???
很犹豫,9实际中用的好像多..但,11都快出了 10是趋势..
不知道
考那个??????????????????????????????????
From: Alex
2007.06.12 17:16
|
|
|
|
|
To: Alex
学好了，OCP证书可以不考的；如果要考，个人认为10g比9i好一点
From: eygle
2007.06.13 08:41
|
|
|
|
|
# 42929
Chi
来自: Mexico
|
|
|
To: Eygle
Can you explain specifically what do u mean the “锁定的问题”? in the discussion of the “alter system archive log current”.From your display, it seems to me that when you switch the archive current log, the previous log sequence is still running under the same thread, so there is a competition for the same spid.While you issue a
SQL> archive log stop;
then
SQL> archive log start;
you generated “ARC0: Unable to archive log 5 thread 1 sequence 324” because OS has give priority to next sequence log; but when you just issue “archive log stop”, you force to stop the previous sequence log.Then with “alter system archive log current”, you requests OS to allocate a new process, which is the same one.Am I right?
p.s. Nice picture.
From: Chi
2007.06.11 22:34
|
|
|
|
|
# 43366
SNOW
|
|
|
To: eygle
盖老师：
我的数据库报出错误ora-04030
我的内存设置为
sga=1.59G
data buffer=1350M
shared_pool_size=200M
pga=800M
我们的软件供应商数据库技术支持分析ora-04030 是因为我内存参数设置不合理导致的
他的意思是说SGA+PGA和 应该小于 1.7G
然后在biti老大的一篇文章《关于SGA设置的一点总结》中提到32bit 通常 SGA有 1.7G 的限制
我的理解是SGA小于1.7G,而不是 SGA+PGA<1.7G
From: SNOW
2007.06.11 21:43
|
|
|
|
|
eygle.com
我要留言
回复
编辑
,发现redolog还没有轮换完（还有好多inactive的log），程序就无法对目标表执行插入了（每次总是插入1500万条左右就挂了），看ADDM “数据库写入进程 (DBWR) 无法满足对空闲缓冲区的需要”是影响最大的因素，是否我们的这种配置（2 Xeon CPU，2G 内存，146G SCSI硬盘）压根达不到这种需求（10000条纪录/秒）呢？
