On Thu, 2003-09-25 at 16:55, Scot Jenkins wrote:

> you might try iostat (part of sysstat package) to see if disk io is
> really an issue.  Also look at top at the same time to see if the box is
> hitting memory (shift+M to sort my memory) and/or cpu (shift+P to sort by
> cpu) limits when the database performance is a problem.

Thanks Scot. I'll play with this a little and let you know what I find
out.

> another thing to do would be to try see if a specific query is causing
> the problem.  Alot of times some poorly written sql is to blame for poor
> performance.  Joins on large tables or joins on too many tables will
> really slow things down.  you could try adding and "explain" to a query
> to see what it's actually doing in mysql.
> 
> Tom Penney wrote:
> > My co worker is convinced that the ext3 partition hosting our mysql
> > databases needs to be defragmented to improve lookup performance. I
> > understand that fragmentation on ext2/ext3 file systems does not happen
> > nearly as bad as it on fat32 or ntfs file systems. I've been told
> > fragmentation is not a problem at all and ext3 File systems never need
> > to be defragmented. I don't know if I believe that to be the whole truth
> > because fragmentation does occur even if it's not really a big problem.
> > 
> > I don't believe defragging this partition is going to make a noticeable
> > difference but I don't really know for sure. The database, although it
> > pretty big & flat, (~4Gig), it's pretty static. Not a lot is added to
> > it. I have not yet run fsck on this partition to find out what the
> > actual non-contiguous file count is. I have to take the box down to do
> > that.
> > 
> > Googling on the subject gets me to a lot of lug list archives of people
> > voicing conflicting opinions but not a lot of solid info that was not
> > over my head. 
> > 
> > I have these questions for you all.
> > 1. Do you think defragging this partition is going to make a noticeable
> > difference in performance?
> > 2. Can anyone point me to any resources that would convince my coworkers
> > that the problem is not the drive, it's the database? 
> > 3. Is backing up the partition, deleting it, then restoring the only way
> > to deferment a ext3 partition?
> > 4. is there a way to determine the how contiguous or fragmented one
> > particular file is?
> > 
> > Interesting article:
> > http://www.informatik.uni-frankfurt.de/~loizides/reiserfs/agesystem.html
> > 
> > -- 
> > Tom Penney <blots at visi.com>
> > 
> > 
> > _______________________________________________
> > TCLUG Mailing List - Minneapolis/St. Paul, Minnesota
> > http://www.mn-linux.org tclug-list at mn-linux.org
> > https://mailman.real-time.com/mailman/listinfo/tclug-list
-- 
Tom Penney <blots at visi.com>


_______________________________________________
TCLUG Mailing List - Minneapolis/St. Paul, Minnesota
http://www.mn-linux.org tclug-list at mn-linux.org
https://mailman.real-time.com/mailman/listinfo/tclug-list