This may be an interim summary if some other ideas come out. The original
question was (briefly):
> When the defragmenter is running, the system slows down
> dramatically, to the
> point where the application starts having problems. I've
> tried nice'ing the
> defragment process, but that had no discernable effect.
>
> Are there any tricks to get the defragmenting done while
> minimizing the
> impact on the rest of the system?
My thanks to all who replied. The suggestions I got came down to these:
* Do the defragmentation when Oracle is down or the load is light. We never
shut it down (at least not for months on end) - the system runs a 24x7
manufacturing process.
* Limit the defragmenter to 1 thread. We only have one file set per domain,
and only do one domain at a time, so by my reading of the man page we're
only using 1 thread anyway. I'll investigate this anyway.
* Defragment only when the file systems get below 95% or so. This isn't a
good option for us, and might even be worse, since we'd never know when the
system might go off on a defragmenting binge.
* Don't worry about it: Oracle doesn't create fragmentation. True enough,
but the file systems do get fragmented from other files (including Oracle
archive files) coming and going. So I think we still need to do this. Note
that the problems occur even though we're defragmenting file domains other
than the ones where the database files are.
* There's little you can do. It's the nature of the defragmenter, and it
does a lot of its work in the kernel, so nice'ing is ineffective.
I'm still open to other ideas...
- Bluejay Adametz
For every credibility gap there is a gullibility fill.
Received on Mon Mar 04 2002 - 20:43:04 NZDT