Re: Redo size vs. archived logs size

Dusan Bolek wrote:
> Hello,
>
> during weekend I have been doing stress testing of an archived logs
> destination on one of our systems here. I have done this by writing a
> PL/SQL package that basically inserts as many rows as specified,
> updated them one by one and deletes them. It also measures time elapsed
> and redo size generated taken from v$mystat. Before and after of each
> run I did Statspack's snaps to know what happened between. There was no
> other significant activity on this system at that time.
> During a test output analyse phase I realised that the amount of
> archived logs generated is about ten times bigger than amount of redo
> size generated taken either from v$mystat or statspack report. Can
> anyone explain this behaviour? My expectation was that the archived

I don't know anything about any of this, of course, but I'll wildly speculate anyways.

Cache fusion allows dirty blocks to be written into the redo stream. I think this can be overridden by setting gc_files_to_locks to non-0, so you might give that a try. See Note:139436.1 on metalink for a detailed description. Also, there are RAC redo bugs, but not enough info is given to know if they could apply to your situation. Bug 4269963, for example, kind of says to me they're still futzing around with this stuff.

Also, post your v$mystat inquiry, maybe some additional eyes might see something silly.

> logs size should be about the same size as redo generated plus some
> overhead. I do not understand why there is such a huge difference.
> Oracle 9.2.0.6, AIX 5.2 64bit, two node RAC with crossarchivation (but
> activity only on "my" node).
>
> --
> Dusan Bolek

jg

--
@home.com is bogus.
Pickle Palace:  http://neopoleon.com/blog/posts/13932.aspx