Re: how does oracle apply rollback?

On 29 Jul 2003 10:59:26 -0700, rgaffuri_at_cox.net (Ryan Gaffuri) wrote:

>Connor McDonald <connor_mcdonald_at_yahoo.com> wrote in message news:<3F266C6D.35D0_at_yahoo.com>...
>> Ryan Gaffuri wrote:
>> >
>> > Does Oracle do a high level 'update' statement between the rollback
>> > data and the table. I would think that this would not be possible
>> > since rollback is stored differently.
>> >
>> > what about redo? Or what about when you do an 'online' index rebuild,
>> > someone else updates and when complete you need to apply the changes.
>> >
>> > anyone know how the algorithms differ from a generic update statement?
>> >
>> > I dont need to know. Im just curious.
>>
>> I don't think its done with anything reminiscent of SQL. If you dump
>> out undo blocks, they typically contain just the bare bones of what is
>> required to undo a change. eg for an insert, just the rowid; for a
>> delete, all the "row" etc.
>>
>> online rebuild is somewhat different. It appears to be equivalent to a
>> materialised view, namely, you start rebuilding and store a journal
>> table of changes that get applied once the rebuild is complete.
>>
>> hth
>> connor
>
>do you think on a rollback it just applies the undo directly to the
>blocks? sort of like a direct path load? they just use C, grab what
>they need in an array and apply the changes directly to the block?
>
>im assuming the undo and redo are applied the same way or am I off
>base?

1 it does
2 no a direct path load is something after the HWM 3 It is more difficult and they have courses about it, and you will have to read the explanation at least 10 times, to get the idea. It is not THAT simple

Sybrand Bakker, Senior Oracle DBA

To reply remove -verwijderdit from my e-mail address Received on Tue Jul 29 2003 - 13:06:45 CDT