My inital impression when I first started Oracle was that the redo
logs contained SQL statements. Later, I heard that they contain
updates on the block level, which is why you can recover one file
without recovering others. According to Quest, their live re-org tool
can operate on just one table even if it is enmeshed with 10 other
tables because the redo logs separate their information by object, so
that each table is completely atomic. What's the real story?
This is how live re-org is supposed to work..
I decided I want to do a live re-org of a table.
I need space to rebuild that table.
Tool goes out and makes a copy of the table A calling it A' for
example WO taking it offline, and begins putting entries in a queue
that it mines from the redo logs that affect only that table. When
the table has completed construction, there is a "brief" (naturally)
interruption whereby the original table is dropped, the new table is
renamed, the redo entries specific for that table are applied, and the
table is brought back on line. miraculous isn't it. I'm not quite
falling for the pitch, but my question is more along the lines as to
how this is possible. If my table A for example was updated by a
cascade delete from another table that would be part of the same SQL
statment and transaction, how are the redo logs mined for just the
entries that affect table A so that they can be applied to A`. If
anyone wants to respond that either tools are over-rated or that they
have used live reorg to good affect, I'd like to hear that as well.
Primarily I am looking for as much esoteric detail as possible as to
how the redo-logs are really constructed.
Received on Sun May 14 2000 - 00:00:00 CDT
