online index rebuilds 2006-01-26 - By Mark W. Farnham
You mean other than having to do an at best order n log n operation to
create the index instead of reading the index in order so the order of the
operation is merely n, and the overhead of absorbing dynamic changes to the
table and resultant changes into the index being built?
As to the "WHY?" question, I'm not sure, but my guess would be that they
believed the process flow of managing the changes while the new indexes was
being built were more reliable that way.
Circa May 1991 the level of difficulty of getting the code provably correct
to read the existing index to create the new index without locks was
referred to by a guy who thought he might have to write it as "Jesus code,"
meaning no disrespect, but the notion that it would require a divine being
to get it right.
Some alternate psuedo code to read the index to build the new index while
locking only handfulls of blocks worth of rows at a time must not have made
the cut, or was lost in the shuffle.
Probably the advantage of the full table scan is getting a handle on which
rows need fixup in the new index image from the read consistent read to fix
it up to the current committed image before you swap it in place.
I'm just guessing here.
mwf
-- --Original Message-- --
From: oracle-l-bounce@(protected) [mailto:oracle-l-bounce@(protected)]On
Behalf Of Roger Xu
Sent: Wednesday, January 25, 2006 3:05 PM
To: joseph@(protected); Bobak, Mark; ORACLE-L
Cc: tanel.poder.003@(protected)
Subject: online index rebuilds
Any other differences besides: 1) does not lock the table; 2) does a full
table scan instead of full index scan; ?
-- --Original Message-- --
From: oracle-l-bounce@(protected)
[mailto:oracle-l-bounce@(protected)]On Behalf Of Joseph Amalraj
Sent: Tuesday, January 24, 2006 9:48 AM
To: Bobak, Mark; joseph@(protected); ORACLE-L
Subject: RE: ORA-01652 (See ORA-01652.ora-code.com): unable to extend temp segment
Thanks,
Had read the article, but I am interested as to why Oracle is doing a
full table scan instead of full index scan for online index rebuilds.
<snip>
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META http-equiv=Content-Type content="text/html; charset=iso-8859-1">
<META content="MSHTML 6.00.2800.1528" name=GENERATOR></HEAD>
<BODY>
<DIV><FONT face=Arial color=#0000ff size=2></FONT> </DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff size=2>You
mean other than having to do an at best order n log n operation to create
the index instead of reading the index in order so the order of the operation
is
merely n, and the overhead of absorbing dynamic changes to the table and
resultant changes into the index being built?</FONT></SPAN></DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff size=2>As to
the "WHY?" question, I'm not sure, but my guess would be that they believed the
process flow of managing the changes while the new indexes was being built were
more reliable that way.</FONT></SPAN></DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff size=2>Circa
May 1991 the level of difficulty of getting the code provably correct to read
the existing index to create the new index without locks was referred to
by a guy who thought he might have to write it as "Jesus code," meaning no
disrespect, but the notion that it would require a divine being to get it
right.</FONT></SPAN></DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff size=2>Some
alternate psuedo code to read the index to build the new index while locking
only handfulls of blocks worth of rows at a time must not have made the cut, or
was lost in the shuffle.</FONT></SPAN></DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff
size=2>Probably the advantage of the full table scan is getting a handle on
which rows need fixup in the new index image from the read consistent read to
fix it up to the current committed image before you swap it in
place.</FONT></SPAN></DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff size=2>I'm
just guessing here.</FONT></SPAN></DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=592125806-26012006><FONT face=Arial color=#0000ff
size=2>mwf</FONT></SPAN></DIV>
<DIV><SPAN class=592125806-26012006></SPAN><FONT face=Tahoma><FONT size=2><SPAN
class=592125806-26012006><FONT face=Arial
color=#0000ff> </FONT></SPAN></FONT></FONT></DIV>
<DIV><FONT face=Tahoma><FONT size=2><SPAN
class=592125806-26012006> </SPAN>-- --Original Message-- --<BR><B>From:</B
>
oracle-l-bounce@(protected) [mailto:oracle-l-bounce@(protected)]<B>On
Behalf
Of </B>Roger Xu<BR><B>Sent:</B> Wednesday, January 25, 2006 3:05
PM<BR><B>To:</B> joseph@(protected); Bobak, Mark; ORACLE-L<BR><B>Cc:</B>
tanel.poder.003@(protected)<BR><B>Subject:</B> online index
rebuilds<BR><BR></DIV></FONT></FONT>
<BLOCKQUOTE>
<DIV><FONT face=Arial color=#0000ff size=2><SPAN class=468280220-25012006>Any
other differences besides: 1) does not lock the table; 2) does a full table
scan instead of full index scan; ?</SPAN></FONT></DIV>
<BLOCKQUOTE>
<DIV class=OutlookMessageHeader dir=ltr align=left><FONT face=Tahoma
size=2>-- --Original Message-- --<BR><B>From:</B>
oracle-l-bounce@(protected) [mailto:oracle-l-bounce@(protected)]<B>On
Behalf Of </B>Joseph Amalraj<BR><B>Sent:</B> Tuesday, January 24, 2006 9:48
AM<BR><B>To:</B> Bobak, Mark; joseph@(protected);
ORACLE-L<BR><B>Subject:</B> RE: ORA-01652 (See ORA-01652.ora-code.com): unable to extend temp
segment<BR><BR></FONT></DIV>
<DIV>Thanks, </DIV>
<DIV> </DIV>
<DIV>Had read the article, but I am interested as to <U>why</U> Oracle is
doing a full table scan instead of full index scan for online index
rebuilds.<BR><BR><SPAN class=592125806-26012006><FONT face=Arial
color=#0000ff
size=2> <snip> </FONT></SPAN></DIV></BLOCKQUOTE></BLOCKQUOTE><
/BODY></HTML>
|
|
