If lets say my log switch is happening after 20 mins.. And my value for fast_Start_mttr_target is 200... So normally checkpoint should happen at log switch that means when a log group is filled then... but 200 seconds are smaller than 20 minutes.. So will oracle perform a checkpoint every 200 seconds even if the log group is half filled lets say... So my question is in this case.. Will oracle perform a checkpoint before a log group is filled as in the above case?
I think it will
Checkpoint automatically occures , at a log switch, forced by dba, when dbf goes offline and when you specify the parameter fast_start_mttr_target=<No of Seconds>.
You should be able to test it don't you?
Edited by: pgoel on Mar 12, 2011 3:08 PM
Edited by: pgoel on Mar 12, 2011 3:09 PM
Always start your question by this style :
Oracle Version :
Then just before writing your question; please think...
1.Have i searched in the Docs ?
2.Have i searched in this forum?
3.What about google's response on my question?
If you are at this stage then :
<<<Now your question text>> and etc.
Like i found couple of links to answer you : what should be the redlog size in OLTP environment like banks?(10g) FAST_START_MTTR_TARGET
(Reply by Aman)
I hope you will try to remeber by small suggestations :-)
About my question, there was no relation of this question to operating system and oracle version.. I simply gave all details that fast_start_mttr_target is set to 200 and lets say log switch is happening every 20 minutes.. So i wanted just a yes or no....Its was a basic oracle arhitecture common for all versions 9i onwards... So the answer seems to me is that now dbwr will work according to this parameter by writing dirty buffers from buffer cache to datafiles and indicating this in the log buffer that these buffers wont be required for next instance recovery.
After reading all your links, i think setting this parameter to high value means your frequency of checkpoints is low and instance recovery may take long time and if this parameter is low then this may lead to database contention for disk i/o.
So overall answer is yes, dbwr does keeps buffer cache with datafiles, if fast_start_mttr_target is set to a lower value.
The bottomline should be that you if you set this parameter to 200 lets say as above mentioned, then instance recovery time will not be 200 seconds, infact we are setting a limit that if beyond this value no dirty buffer should remain in database buffer cache.....
The value of this parameter basically determines the Fast Start Fault Recovery or Incremental Checkpointing of Oracle. This has nothing to do with the log switch. And from 10.2 onwards, the parameter setting is going to be automatically tuned by Oracle. Edit:
I am trying to find a document link on OTN which explains the behavior of this parameter from 10g onwards very clearly but seems there is some trouble and I can't access the OTN. If you can, search for "fast start fault recovery" in the htdocs over OTN. If I can find the link,I shall post .
Edited by: Aman.... on Mar 13, 2011 10:06 AM