Comments on: Index Quiz 2

By: IQ2 – Answers « Oracle Scratchpad

IQ2 – Answers « Oracle Scratchpad — Sun, 04 Oct 2009 07:11:29 +0000

[...] files — Jonathan Lewis @ 9:04 pm UTC Jul 24,2009 I was planning to supply the answers to Index Quiz 2 as a comment – but there’s a lot of block dumps involved, and it’s easier to do [...]

By: Log Buffer #156: a Carnival of the Vanities for DBAs | Pythian Group Blog

Log Buffer #156: a Carnival of the Vanities for DBAs | Pythian Group Blog — Fri, 31 Jul 2009 22:42:36 +0000

[...] Jonathan Lewis with his Index Quiz 2 , which is easier, he promises, than Index Quiz [...]

By: Index Quiz 1 « Oracle Scratchpad

Index Quiz 1 « Oracle Scratchpad — Fri, 24 Jul 2009 21:08:47 +0000

[...] by Index Quiz 2 « Oracle Scratchpad — July 23, 2009 @ 7:55 am UTC Jul 23,2009 | [...]

By: Jonathan Lewis

Jonathan Lewis — Fri, 24 Jul 2009 16:23:37 +0000

lascoltodelvenerdi,
True – freelists is ignored as well (and freelist groups) – if you’re using ASSM.

By: lascoltodelvenerdi

lascoltodelvenerdi — Fri, 24 Jul 2009 12:03:34 +0000

@Jonathan
PCTUSED and FREELISTS, am I rigth?

By: Jonathan Lewis

Jonathan Lewis — Fri, 24 Jul 2009 11:03:54 +0000

Franck,
That’s just about the perfect answer.

I’m hoping to find time to write up my own answer some time this evening, but that will just add a couple of extra details.

By: Jonathan Lewis

Jonathan Lewis — Fri, 24 Jul 2009 11:01:52 +0000

lascoltodelvenerdi,
The behaviour of initrans isn’t affected by switching to ASSM – the only parameter affected (viz: ignored) by that switch is pctused.

By: Franck Pachot

Franck Pachot — Fri, 24 Jul 2009 07:02:55 +0000

Hi,

Here is my guess:
There is no need to reserve a number of ITL because it is always possible to split a block in order to have free space. So initrans is ignored.
However, the transaction that splits the block must have an ITL entry for itself, so I guess that oracle always keep one free ITL entry in order to be sure that a transaction can split the block.

So here is what I guess that happens:
1- the insert needs 1 ITL entry, and adds 1 free ITL free, so result is 2 ITL entries
2- when the block was split, ITL entries were copied in the two resulting blocks, so both blocks have 2 ITL entries
3- each concurrent insert need their own ITL (so 10) and keep 1 free entry, so result is 11.
After the block split, they are copied in both blocks.

Regards,
Franck.

By: lascoltodelvenerdi

lascoltodelvenerdi — Thu, 23 Jul 2009 14:48:41 +0000

@Coskan

It looks like Oracle is ignoring the initial initrans.

We did not ask one thing: the index is created in an ASSM tablespace?

If this is true, the initrans parameter (and others) is ignored.

By: coskan

coskan — Thu, 23 Jul 2009 14:05:00 +0000

1- itc=2
2-
Branch Block itc=1
Leaf Block1 itc=2
Leaf Block2 itc=2
3-
Branch Block itc=1
Leaf Block-1 itc=11
Leaf Block-2 itc=11

but the 11 one is because I ran the multiple insert from the 11th session I think I am not sure

Initran for index behaviour is also covered in this ML: 729445.1