Re: deadlocks and redologs interaction

You try to link deadlocks to the redologs, but there is no connection.

A deadlock is just a 'clash' of transactions where on gets committed, the other rolledback:

so let's rewrite:

>My suspicion is that an extensive # of deadlocks can interfere with
>redolog rotation;

to:
My suspicion is that an extensive # of transactions that gets committed or rolled back can interfere with redolog rotation;

True, although the words 'can interfere with' need to be replaced by 'will cause more'.

And your posting:

>What I am looking for, though, is an
>explanation of the impact an extensive # of deadlocks can have on the
>ability of oracle to put redologs in an INACTIVE state in v$log.

The INACTIVE state means that the log isn't needed for crash recovery, the file
is ready for use. If 9 out of 10 logs are in the ACTIVE state, that's not a good position.
Increase the size of the logs in this case.

> I suspect a redolog cannot be freed up until a deadlocked transaction is
> completed or rolled back.

Forget the word 'deadlocked' in that sentence and read it again. The redo don't treat a deadlock differently - for simplification purpose - the redologfile is a journal of DML commands for recoverypurpose.

H. Received on Wed Sep 12 2007 - 04:32:52 CDT