Oracle spinning in tight loop

[Marc Bohnert]

Hello All,

My setup is:
HPUX-10.20
Oracle 7.3.2

I have a server process that connects to Oracle, forks children that connect to Oracle and then sits in a select waiting for children to die. The children do all the work in the database. Once a day the parent gets into a situation where it consumes 100% of the CPU. I noticed that it has received 72 million signals.

Does anyone know if there is a bug in HPUX and or Oracle that would cause a process to get this many signals and consume that much CPU.

Thank you
Marc Bohnert

[Joseph Loo]

Does it stay at 100% or more and just hanged there after? Are you also running any Oracle Apps or Oracle Web apps in the server?

I need to know more to answer you.

[Alexander M. Ermes]

Hi there.
DO you work with Oracle Applications ?
Check fur TWO_TASK parameter in the environment abd disable it.
Rgds
Alexander M. Ermes

[Marc Bohnert]

We are connecting using bequeath. We were actually going to try to set TWO_TASK to see if it helps. It is my understanding that TWO_TASK will cause the connection to use the listener. Is that true? Why should we not set TWO_TASK?

[Steve Slade]

Hi Marc,

What do you mean by the parent process sitting in a select? What is it selecting?

[Philip Chan_1]

Hi Marc,

What is the timeout setting for the select call? I presume the "select" must be enclosed by a loop, and, if the "select" call was not blocked then the loop iteration may cause your CPU to reach 100% utilization.

Rgds,
Philip

[Philip Chan_1]

Hi Marc

My wild guess is that the 72 million signals were causing the select call to return immediately, therefore the loop contained the select became an unblocked loop instead, that WILL draw the CPU to peak for sure.

Perhaps you should find out where the hell these 72 million signals came from, then stop them.

Rgds,
Philip