Re: High Redo Copy Contention

I see a dogpile of responses starting to gather on this thread, so maybe it is time to step in.

Why are you tuning from just one element of the instance? That is, what makes you think you have a redo copy problem in the first place? Look at things this way -

1. Look at the top 5 or so wait events. Is "latch free" one of them? Ignore things like "pmon timer" or "sqlnet message from client" - those are idle waits that have nothing to do with performance. If latch free is a top non-idle waiter, then look at the latch sleeps breakdown of willing-to-wait latches in your estat report
2. Organize the latch statistics by "sleeps". Is redo copy one of the top sleeps? Forget about hit/miss ratio. Sleeps = wait time. Wait time means "work not getting done because of a bottleneck".

If you are having trouble analyzing all of this, zip up your estat report and post it here. I will look and comment. Make sure the report is taken for a reasonably short period - say 1/2 hour or even 15 minutes, during a period when you think the performance is worst, presumably when the instance is busy. Zip it up and post it here.

• ricky

Anurag Minocha wrote:

> Hi,
> The numbers for the actual gets and misses are also high. Below is the query
> I ran to generate the numbers.
>
> Please Help.
>
> select name, gets, misses,
> 2 immediate_gets, immediate_misses, sleeps
> 3 from v$latch
> 4 where name in ('redo allocation', 'redo copy') ;
>
> Tablespace GETS MISSES IMMED GETS IMMED MISS SLEEPS
> --------------- -------- -------- ---------- ---------- -------
> redo allocation ######## 13176 0 0 814
> redo copy 27093 26585 ######## 56439 442
>
> Thanks
> Anurag
>
> --
> Posted from jshemail.herold.com [63.105.8.12]
> via Mailgate.ORG Server - http://www.Mailgate.ORG
Received on Fri Apr 06 2001 - 09:45:49 CDT