Re: log file sync

Thanks for the response. I suppose it is somewhat relevant how long the instance has been up, though I would think that log file syncs representing 56% of the total wait time would be some concern regardless of the length of time the instance has been up. The statistics I posted were taken when the instance had been up for about a day. So the time waited, which was 538 minutes, seems substantial to me. During a recent 6 day period, the log file sync waits totalled 1929 minutes (115727 seconds), or 38% of the total wait events of the instance.

I am primarily tuning the database. The statistics I posted were the top wait events for the instance, which is the first place I look when tuning. Latch free waits are 0.3% of total time waited (and the vast majority of these are shared pool, library cache, and cache buffer chains), so the problem doesn't seem to be with redo-related latches.

It just seems that having log file syncs as such a large percentage of total waits suggests something might be improved, and I'm considering what that something is.

--Terry

> Are you having a performance problem or are you 'just' tuning your ratio
> statistics? Have you heavy waits on redo allocation, redo copy, redo
> writing or the like? The (redo)log_buffer is written to disk when 1/3
> full or 3 seconds have passed since last write or at a commit. And your
> statistics doesnt tell anything if they aren't comparable with how long
> the instance has been running. If it has been up 172 days, then a summed
> wait of about 56 seconds isn't a problem. Or?
>
> /Svend Jensen
>
>
Received on Mon Feb 25 2002 - 16:39:19 CST