IWETHEY v. 0.3.0 | TODO
1,095 registered users | 0 active users | 0 LpH | Statistics
Login | Create New User
IWETHEY Banner

Welcome to IWETHEY!

New Oracle bug
Server: Oracle 9.2.?, Solaris 8
Client: Oracle 9.0.1, Red Hat Linux 7.2/3? - 2.4.7-10smp

select file1, field2, 'hardcode crap' field3,
field4, 'more hardcode' field5, ' ' filler1,
field6, ' ' filler2
from table etc.

Note: The above filler fields have spaces
in them, your browser may not show them.

The final filler causes a CPU leak on the Linux box!!!!

Initially it will pull about 5000 RPS, using 30% of
CPU. After about 2 minutes, the CPU starts to creep up.
At 80% of CPU, the RPS starts to degrade, until I kill
it. It was running at about 100 RPS toward the end.

This is 100% Oracle, since it is repeatable in SQLPLUS
session. It was initally in PERL-DBI, which allowed
the fine grained timings on groups of records.

Get rid of the final filler2, and it works fine.

Wow.

It would take a LOOONNNNGGGGGG time to export 60 million
records at 100 RPS.

I'll have the new client up in a day or so, and will report
back if it has the same problem.
New But, but, it's unbreakable! :)
Alex

"Television: chewing gum for the eyes." -- Frank Lloyd Wright
New Actually... maybe not...
I'd like you to try it on RH7.3 and link with GCC 3.x.

We had a similar problem with 9.something... on AIX and the your same client on RH7.2 with the same kernel. We changed to a vanilla 2.4.16 with a "RedHat" config compiled with GCC 3.04 (at the time) and updated Libraries(2.2.4something)... then re-linked Oracle with the installer.

The problem appeared to disappear. Hasn't shown backup yet.

Now, I know you and I differ on things of this sort... but if you could try it... It may just work and then you could tell Oracle to pay you.... for once.

greg - Grand-Master Artist in IT,
curley95@attbi.com -- REMEMBER ED CURRY!!!
New Hrmph
Differ? On what? The effort required to get
to the fix? Maybe.

I've given up DBAing and sysadmining these boxes
unless there is an emergency. So I've sworn
no compiler or library upgrades, so I won't be
trying.

I almost did, though. And as I caught myself
type 'su', I realized it was a nono. Imagine,
as the Godlike sysadmin you are, there is one
power user. He hired YOU. And still has the
root passwords, but has sworn not to use them
unless he can't get a hold of you, and only for
production emergencies. You'd be REALLY unhappy
if he started upgrading chunks of the box, for
development, right?

I'll forward you message to my DBA and sysadmin,
but I'd really not want them to spend the time.

The new boxen that just showed up are the real targets,
(dual 2.4Ghz Xeons, yay!!!) and they will have 9.2 to
start with. If it still doesn't work, then they play.

But actually, I'm happy with my "fix", so I'd suggest
they ignore it. We have REAL work to do here.

I've also just reread your message, and realized
you told be to do it on a different RH version,
and relink from the installer from scratch. Yuk.
Any idea what version Dell is shipping nowadays
with their dual Xeons? I'd hate to reinstall,
with the vanilla kernel, since they do a bit
of patching so their raid controller and
hyperthreading work.

Also, on the Oracle $$ side. Yeah right! Why
would they give me any $$. Is there a bug fee
I'm unaware of? Unless there is real $$, I
don't have time or spare equipement at the
office. I could do it at home, but my cable
modem is too slow to duplicate the problem,
and I don't have Solaris server locally.
New (Voice pitch=deep) BEAUTIFUL! do-oo-bownp-bownp (/Voice)
At least I get... REAL equipment... RS/6000s and Netfinity servers. Not those Sun pieces and Dell desktop(ish) servers... *grin*

Only reason I said do RH7.3 was that those probs never showed up on a fresh build of 7.3. And the newer stuff is already there. Also RH7.3 doesn't include gcc3 by default so...

And I actually compiled the "special" hardware modules to work with the 2.4.16 kernel.

Yeah... you are right if my Manager just delved in and started updating large chunks I'd ... I'd ...I'd... well not be happy that's for sure.

Other than that... if they need to get ahold of me... you know how to...

edu.dot.grcc@gfolkert

There yah be sir.


greg - Grand-Master Artist in IT,
curley95@attbi.com -- REMEMBER ED CURRY!!!
New Nice try
Too bad you are wrong.

[link|http://configure.us.dell.com/dellstore/config.asp?customer_id=555|http://configure.us...tomer_id=555]&order_code=PE2650PAD&s=biz

Imagine a STACK of these, fastest CPUs, 2GB of RAM each, filling a rack.

Them's my compute servers. Nothing desktop about them.

I tried the IBM way. Too slow. Too expensive for too little speed.
But then, as an IBM bigot, you'd be used to that.
New Bwahahaha... ME???
I am no IBM Bigot...

I admit I like the hardware... BUT when it comes to RAW power... I used to prefer ALPHAs.

For Intel... I Loved ALR before they got bought by Gateway...

I also Loved NetFRAME before Micron bought them and killed the Tech they had...

Right now... I prefer (Personally I might add though) [link|http://www.microway.com/|Microway].

I might add: [link|http://www.microway.com/2000dual.html|These puppies!]

Now thems some horse power...

greg - Grand-Master Artist in IT,
curley95@attbi.com -- REMEMBER ED CURRY!!!
New ALR?
I had the original dual CPU SystemPro rip-off.
I sent it back. It refused to boot about
1/2 the time. I then ended up with an
Everex StepMP
New Went away on the new Dell box
Linux rip 2.4.18-4smp #1 SMP Thu May 2 18:32:34 EDT 2002 i686 unknown

Oracle 9.2 client

All better
     Oracle bug - (broomberg) - (8)
         But, but, it's unbreakable! :) -NT - (a6l6e6x)
         Actually... maybe not... - (folkert) - (6)
             Hrmph - (broomberg) - (4)
                 (Voice pitch=deep) BEAUTIFUL! do-oo-bownp-bownp (/Voice) - (folkert) - (3)
                     Nice try - (broomberg) - (2)
                         Bwahahaha... ME??? - (folkert) - (1)
                             ALR? - (broomberg)
             Went away on the new Dell box - (broomberg)

You can put makeup on a pig. It's still a pig.
118 ms