It really is amazing the difference that things on a software level
can make, and even seemingly trivial settings...
I don't know if you guys are equipped to do this, but the only real
way to find out what is slow is to measure different things and see
what is causing the slowness. This can be done with log messages on
the simplest level, and with a profiler to really get an obscene
amount of information.
Doing measurements is the only way to avoid flailing around and
randomly changing things, which seems to be more or less the stage
your at right now.
Best of luck,
-David
On Aug 19, 2006, at 7:32 AM, Walter Vaughan wrote:
--- Walter Vaughan <[EMAIL PROTECTED]> wrote:
We have noticed a drastic slowness in building pages
that does not seem to be
fixed with following the Production Guide
instructions on setting the caches.
After seeing the drastic slowdown after testing with postgreSQL, we
tested on a vastly underpowered box (Athlon-1.2Ghz,
256Meg,MySQL5.1,JDK1.4.2,FreeBSD6.1) with the same data we had
loaded, and ofBiz works just fine.
The slow server in question is a dual Xeon Intel Server with all
the goodies of two years ago... 4 drive RAID 5, 1 Gig Memory, 100%
Intel, 'cept the drives. The same OS, but the JDK is 1.5 from the
freeBSD foundation. So last night did a Vacuum on the PostgreSQL,
and followed its instructions, tried to restart PostgreSQL and it
failed with a very cryptic "unable to start postmaster". Rebooted
the server, and even it wouldn't restart pgsql. Time to go back to
MySQL. We've more problems than slow query times.
With MySQL we're back to the behavior before, except the page
delays are somewhat random, and only last a second or two. A good
example would be clicking radio buttons in the Party Lookup area
changing from Postal to Telecom.
So, I'm going to dump the native 1.5 JDK and fall back to the Linux
based 1.4.2 and see what that gets us.
Else that or I demo the system using a computer that was picked up
for $20 at the local GoodWill thrift shop vs a $2500 server.
--
Walter
