Re: DWBR and dirty buffers

Naive question, but did you give Oracle some time to do the writes after the switch/checkpoint ? It will probably take sometime to write 10,000 blocks (especially in a RAC system where there may be a need for cross-instance chat to clear the status of any PIs that are older versions of the XCURs).

It is interesting that you don't have any PIs in your report. Is your RAC system active/passive, or is it very clearly split by function ? Do you normally see some PIs if you run the query outside the scope of this test.

-- 
Regards

Jonathan Lewis

http://www.jlcomp.demon.co.uk/faq/ind_faq.html
The Co-operative Oracle Users' FAQ

http://www.jlcomp.demon.co.uk/seminar.html
Public Appearances - schedule updated Jan 21st 2005






"Noons" <wizofoz2k_at_yahoo.com.au> wrote in message 
news:1109297099.819349.279810_at_l41g2000cwc.googlegroups.com...


>

> Jonathan Lewis wrote:

>> Check the status of the dirty blocks.

>> You may find that they are all CR.

>>

>> When a dirty current block is cloned to create

>> a CR block, there seem to be cases where the

>> dirty bit is carried to the CR block - and CR blocks

>> are never written, so the bit never changes.

>>

>

> Interesting.

> Check this from my RAC db, 9ir2:

>

> Purr...>select dirty,status,count(1)

>  2  from v$bh

>  3  group by dirty,status;

> Press Return to continue...

>

> D STATU   COUNT(1)

> - ----- ----------

> N cr            73

> N free           1

> N xcur      145846

> Y cr            63

> Y xcur       10173

>

> Elapsed: 00:00:01.00

>

> those dirty blocks don't get cleaned in

> a switch logfile.

>