Re: Does a commit cause a checkpoint?

Ryan wrote:

> hmmm... figured it was something like that. So Im assuming there is a latch
> that lgwr aquires which tells dbwr do not write xyz to the datafile, until
> Im finished. DBWR polls the latch. Is there a latch and if so what is the
> name? Or does it just use inter-session communication? like synchronization
> etc...
>
> Is this the log synch wait I see when I run statspack? dbwr is waiting for
> LGWR to write the blocks to the redo log before DBWR can go? Im assuming it
> can also be caused if you have your datafiles on the same mount point as
> your redo logs and you can only write to either datafiles or redo logs at
> the same time.

Sorry to reference you back to yet another doc -

"Oracle9i Database Performance Tuning Guide and Reference, Release 2 (9.2), Part Number A96533-02"
- Chapter 22, "Instance Tuning"
- Section "Table of Wait Events and Potential Causes"

has this to say about log file sync:

General area: I/O, over- committing
Probable cause: Slow disks that store the online logs Look for/examine: Un-batched commits & Check the disks that house the online redo logs for resource contention. Check the number of transactions (commits + rollbacks) each second, from V$SYSSTAT. Received on Thu Jun 19 2003 - 19:27:26 CDT