Re: "X commits per minute": is there an absolute threshold ?

yong321_at_yahoo.com (Yong Huang) wrote in message news:<b3cb12d6.0210311459.5846a122_at_posting.google.com>...
> groups commits together". It's true LGWR waits 10 centiseconds after
> it's posted by the first commit, to see if there's another commit
> request coming in. If yes, they're combined into one. But I don't

Jonathan Lewis emailed me saying it doesn't sound likely that's the case. The reason is that if that were true, each session would have a lot of log file sync wait time. I think Jonathan is right. I read that from Metalink Note:194107.999. Some Oracle kernel developer is saying that. But I think a simple test like this

SQL> set serverout on
SQL> declare
  2 t1 number;
  3 t2 number;
  4 begin
  5 select hsecs into t1 from v$timer;   6 dbms_output.put_line(t1);
  7 execute immediate 'grant select on q to public';   8 select hsecs into t2 from v$timer;   9 dbms_output.put_line(t2);
 10 end;
 11 /
583316080
583316082

PL/SQL procedure successfully completed.

SQL> /
583316408
583316409

PL/SQL procedure successfully completed.

SQL> /
583316519
583316521

PL/SQL procedure successfully completed.

can prove that wrong. (Here I have a table Q. GRANT is a DDL which commits) Also, tracing lgwr process (truss -t pwrite -p [pid] on Solaris) while I press / in sqlplus seems to show lgwr writing happens immediately; 10 centiseconds would be a noticeable delay, I think.

As Jonathan pointed out, a group commit occurs when LGWR is writing while other sessions commit. Those commits will be combined into one log write.

Yong Huang Received on Wed Nov 06 2002 - 17:37:41 CST