RE: Oracle and Apache Web server bottleneck

Don:

Ok...so now the question is 'why are the http processes choking the system?'. I loaded the Oracle Apache web server on a little one-lung HP C110 workstation, and it ran fine with the default number of http processes, which I think was 8 or 10. I did throttle it back to four because this is a sandbox machine and 10 http process was way overkill for what I needed.

IMHO, it's a contention issue of some kind. The CPU's peg because they're all waiting on a resource and can't continue until they get it and do something.

Are you running the Oracle server, App server, and http server all on the same box? When we were running PeopleSoft, having the App server and DB server on the same box proved to be our biggest heartache. We moved the app server to a second machine and much of the problem went away...except for the I/O.

Some more questions to ask are:
Does Glance show what the http processes are doing? (what mode: running, waiting, sleeping, etc)
Are your kernel parameters set up to handle many processes? How's your shared memory?
Are the semaphores in contention between processes? Does anything else peak when the CPU's peak? I/O? Memory? Network? Are there any relevant messages in the syslog? alert_SID.log? Apache log?

Keep Thinking!
Mike

-----Original Message-----
Sent: Friday, February 15, 2002 8:33 AM
To: Multiple recipients of list ORACLE-L

Yes, that is autoraid. The db is not all that big yet, so the array stats show that db stuff is staying in RAID +1.

Glance is showing that the httpd processes are choking the CPUs, forcing other jobs to go to "sleep".

At 2/14/02, you wrote:
>Hm...a 12x36 disk array. Is that an AutoRAID?
>
>I had one of those hooked up to a 6-way K570 running PeopleSoft
>over Oracle 8.0.5. Yuck. P**s-poor performance all the way
>around. Especially when the disks started to fill up. The
>problem was that the array got so busy internally swapping
>stuff from the RAID 5 area to the RAID 0 area that it could
>not service the I/O requests from the database. We went to
>a fiber array and most of the problems disappeared.
>
>If you are using an AutoRAID, check your internal statistics
>to see if you're swapping like mad in there.
>
>HTH,
>Mike
>
>-----Original Message-----
>Sent: Wednesday, February 13, 2002 6:38 PM
>To: Multiple recipients of list ORACLE-L
>
>
>We have a single HP 9000 with 2-360Mhz CPU, 4G ram and a 12 x 36G disk
>array.
>
>We were running Oracle Apps 11.0.3 on Oracle 8.0.5, but recently upgraded
>to Oracle 8.1.7.2 so that we could provide web access to purchasing
>(iProcurement) and later to HR (employee self service).
>
>Oh, but do we have a problem.
>
>In testing, with a single instance, we found that the Apache web server was
>able to service 1-10 users very well, 10 - 14 was poor with 1-3 minute
>response time, and above 16, the web browser sessions timed out after 10
>minutes. The CPUs peg at 100% around the 12 user and stay there. With 150
>Purchashing users, we have delayed go-live until we get this Apache
>bottleneck solved.
>
>The boss is talking to HP about getting a second server. Server A will
>have the DB and server B will have apps, Oracle web server and Apache web
>servers.
>
>I'm a thinking that we need to get that Apache web server onto a different
>box of some kind.
>
>Any ideas on dealing with the Apache server problem?
>
>
>Thx.
>

-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.com
-- 
Author: Vergara, Michael (TEM)
  INET: mvergara_at_guidant.com

Fat City Network Services    -- (858) 538-5051  FAX: (858) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).