Re: Monitoring questionable long running sessions

From: VC <boston103_at_hotmail.com>
Date: Mon, 12 Jul 2004 23:54:42 GMT
Message-ID: <6xFIc.77090$Oq2.65625@attbi_s52>

"Rick Denoire" <100.17706_at_germanynet.de> wrote in message news:bp46f055ee84m0tc77mvq68t6730qhu90k_at_4ax.com...
> "VC" <boston103_at_hotmail.com> wrote:
>
>
> >> I read something about mad "spinning" CPU.
> >
> >Never heard about anything like that.
>
> I will try to give a rough explanation. When some kind of resources
> are accessed, a latch is needed. If the latch is not acquired, Oracle
> can decide to "wait", and in the meanwhile, the CPU can be used for
> something else, or it can decide to "spin", which means that it goes
> through a loop and then does a retry. Due to some error, the CPU gets
> caught spinning indefinitely.

The only process with this problem metalink mentions is lgwr althought it's unclear how exactly 100% CPU consumption occurs since a typcal 'spin' is like this:

get a latch ->wait->retry-> etc. Due to waits, there cannot be 100% CPU usage by a 'spinning' process.

When you observed your situation, was it lgwr or some other process ?

>
> You could find related issues at Metalink.
>
> Bye
> Rick Denoire
>
Received on Mon Jul 12 2004 - 18:54:42 CDT