Skip to Main Content
Forums

Oracle Database Discussions

Announcement

For appeals, questions and feedback about Oracle Forums, please email oracle-forums-moderators_us@oracle.com. Technical questions should be asked in the appropriate category. Thank you!

Interested in getting your voice heard by members of the Developer Marketing team at Oracle? Check out this post for AppDev or this post for AI focus group information.

dbwr for redo log buffer.

685267Feb 16 2009 — edited Mar 19 2009
hi,

Is there any situation, when dbwr writes anything in log buffer other than writing into data files.

Comments

Pavan Kumar
Hi,

Can you re-phrase your Question or Query...!! Bit hard to Understand.

- Pavan Kumar N
591309
Never ever.
In computers,things are done only as per programmed.
DBWR has no relation to redo log buffer and LGWR has no relation with buffer cache.
Pavan DBA
hi dbwr process is having only one way action and its job is to write dirty blocks from db buffer cache to db files. log buffer cache has no relation with dbwr.
Aman....
user_7000006 wrote:
hi,

Is there any situation, when dbwr writes anything in log buffer other than writing into data files.
No, the dbwr is not going to read from the log buffer. Yes, it may happen that it can ping to lgwr or vice versa, but it won't read through log buffer and write to the log files.

That said, its always better if you come up with some reference of any site/book/paper where you have read it.At times, things taken out of the context mean differently and that may lead to a different answer for a query.
HTH
Aman....
Timur Akhmadeev
Well, never say never... 
SQL> select sid from v$session s, v$process p where s.program like '%DBW%' and p.addr = s.paddr;
 
       SID
----------
       662
 
SQL> select name, value from v$sesstat s, v$statname n where s.statistic# = n.statistic#
  2  and name like 'redo%' and sid = 662
  3  /
 
NAME                                                                  VALUE
---------------------------------------------------------------- ----------
redo synch writes                                                         0
redo synch time                                                           0
redo blocks read for recovery                                             0
redo entries                                                        3763645
redo size                                                         422412184
redo buffer allocation retries                                        20052
redo wastage                                                              0
redo writer latching time                                                 0
redo writes                                                               0
redo blocks written                                                       0
redo write time                                                           0
redo log space requests                                                3415
redo log space wait time                                                  0
redo log switch interrupts                                                0
redo ordering marks                                                       1
redo subscn max counts                                                    5
 
16 rows selected
26741
The "redo blocks read for recovery" isn't done by DBWR.
Timur Akhmadeev
Yes, but what about 'redo size'? In what DBWR activities redo is being generated? Maybe it is blocks cleanout?
26741
Yes, block cleanout DOES generate redo. But it is LGWR that writes the redo that is generated.
Aman....
Timur Akhmadeev wrote:
Yes, but what about 'redo size'? In what DBWR activities redo is being generated? Maybe it is blocks cleanout?
Yes in the block clean out , there would be redo generation. DBWR does use an IPC connetion with LGWR but that doesn't mean that it will write from the LOg buffer to the log files, for that , there is a never.

Aman....
Timur Akhmadeev
Probably I misunderstood OP's question. Sorry.
Aman....
No probs!
Aman....
Jonathan Lewis
Timur Akhmadeev wrote:
Well, never say never... 
redo entries                                                        3763645
redo size                                                         422412184
redo buffer allocation retries                                        20052
To allow for the newer "two-pass recovery" the database writer writes "block written records" (BWR) to the log buffer. This is just a short note for each block written which comprises the block address and the block SCN. Since these BWRs are not critical data (i.e. if you lose one it slows down recovery, but doesn't make it fail) DBWR doesn't send a synch message to the log writer to get them written to disc, they just go with the next log write.


Regards
Jonathan Lewis
http://jonathanlewis.wordpress.com
http://www.jlcomp.demon.co.uk

"For every expert there is an equal and opposite expert"
Arthur C. Clarke

To post code, statspack/AWR report, execution plans or trace files, start and end the section with the tag {noformat}
{noformat} (lowercase, curly brackets, no spaces) so that the text appears in
fixed format
.
26741
We learn something new every day.
Pavan Kumar
Hi,

Jonathan Can you get little more insight in to that topic. I will be thankful to you. That's why I request to prepare one seminar in INDIA for us. we wil be lucky..!!

@Hemanth, you are right. :-)

- Pavan Kumar N
Timur Akhmadeev
Thank you, Jonathan.

Pavan, a very in-depth Oracle internals insights could be found through [Google patents search|http://www.google.com/patents]. For example, this particular topic is covered in patent #5974425 [Method and apparatus for reapplying changes to a database|http://www.google.com/patents?id=aEIlAAAAEBAJ&dq=%22block+written+records%22] (here is [direct link|http://www.google.com/patents/pdf/Method_and_apparatus_for_reapplying_chan.pdf?id=aEIlAAAAEBAJ&output=pdf&sig=ACfU3U0YNeEcMBiYScNGYiXecJYfQo5icQ] to complete pdf version). "Block written records" are called "confirmation records" in that paper.
Pavan Kumar
Hi

Thanks for sharing the link.. !!

- Pavan Kumar N
Aman....
Pavan,
I shall just attempt to give a short note about it. From 9i onwards , the instance recovery mechanism is changed to incorporate some thing called Two Pass recovery. The traditional mechanism used to check back the entire redo since the last checkpoint would push the redo vectors from that point to the tail of the log to the datafiles. There may be very well chances that some of the redo vector's buffers are already a paprt of the datafile. From 9i , the mechanism is enhanced to add a small bit to the redo content called BWR(Block Write Record) which is a bit added to the content by DBWR mentioning that it has already completed teh write of this vector to the datafile. In the event of instance crash, redo log is read in two phases, One pass would read the entire redo. 2nd pass will skip those contents which are having BWR record associted with it and will prepare the final lot which will be applied to the data file. That's what is called Two Pass recovery.
HTH

Jonathan sir, please correct where ever I missed anything.
Regards
Aman....
Aman....
Timur,
Thanks for sharing the link.

Cheers
Aman....
1 - 18
Locked Post
New comments cannot be posted to this locked post.

Post Details

Locked on Mar 17 2009
Added on Feb 16 2009
#general-database-discussions
18 comments
1,442 views