Author |
Message
|
harelmoshe1 |
Posted: Wed Jan 11, 2006 3:10 am Post subject: biphttplistener goes wild on CPU usage |
|
|
Apprentice
Joined: 29 Dec 2005 Posts: 41
|
Hey,
I'm using Broker version 5.0 with CSD4 on AIX.
I have a simple flow of the form:
HTTPInput >> Compute >> HTTPReply
In the compute node, the only thing done is database-writing.
Occasionslly, say once in 100 or 150 idenitcal requests, the biphttplistener process gets stuck, taking 90%+ of CPU usage and blocking acception of any further requests. Seldom, it causes the DataFlowEngine of the EG to also take 90%+ of CPU (this is possible cause the server has 6 CPUs). This is a not DB problem because it happens even if the DB writing is node is taken out.
Any ideas ?
I'm totally lost, i can't even isolate the problem - the flow stands loads of some 500 requests sent in a row, but can randomly get stuck after another a single request is sent... |
|
Back to top |
|
 |
elvis_gn |
Posted: Wed Jan 11, 2006 3:50 am Post subject: |
|
|
 Padawan
Joined: 08 Oct 2004 Posts: 1905 Location: Dubai
|
Hi harelmoshe1,
What does the broker log say ? |
|
Back to top |
|
 |
harelmoshe1 |
Posted: Thu Jan 12, 2006 2:17 am Post subject: |
|
|
Apprentice
Joined: 29 Dec 2005 Posts: 41
|
It says nothing unusual but the lines about the execution group being started, there are no other errors realted to the listener or the EG itself... |
|
Back to top |
|
 |
elvis_gn |
Posted: Thu Jan 12, 2006 2:31 am Post subject: |
|
|
 Padawan
Joined: 08 Oct 2004 Posts: 1905 Location: Dubai
|
Hi harelmoshe1,
Dont have much idea if the problem is due to HTTP....
Could you check the message size....I had a problem some time back, when the message had many repeating element occurances, the flow was looping on the single message...and CPU usage went up...that was a STACK problem which you should find in the forum....
You could check it out, but not guranteeing anything.
Regards. |
|
Back to top |
|
 |
nathanw |
Posted: Mon May 15, 2006 7:49 am Post subject: |
|
|
 Knight
Joined: 14 Jul 2004 Posts: 550
|
OK
I am experiencing this on a test system
the production environment is configured the same and it is not happening (ok the production is alot bigger interms of CPU memory etc)
Once i restart the brokers all is fine for approximately 20 minutes then the biphttplistener goes to 90+%
have checked through the logs for the brokers and there is nothing mention
environment
xseries running suze 9 2 cpu 4 gb ram
WBIMB 5 FP4
One thing I can add is that I am getting a lock in the RetainedPubsTableLockSemaphore at the same time as the listener goes haywire |
|
Back to top |
|
 |
poker |
Posted: Mon May 15, 2006 10:23 am Post subject: |
|
|
Novice
Joined: 14 Mar 2006 Posts: 20
|
This is a know problem. Even on our HP-UX machine it was a problem.
Wht CSD level do u have? Better move to CSD 6 |
|
Back to top |
|
 |
nathanw |
Posted: Mon May 15, 2006 11:59 pm Post subject: |
|
|
 Knight
Joined: 14 Jul 2004 Posts: 550
|
|
Back to top |
|
 |
|