Re: PMON

On Apr 11, 1:23 pm, Uwe Weber <uwe.we..._at_teleos-web.de> wrote:
> Hi all,
>
> yesterday some maintenance tech decided that it was save to
> swap one of the interfaces on our firewall. Unfortuntaley,
> it was not. All connections from users to our web application
> were cut off. Which left our main production instance (10.1)
> with about 400 sessions whose clients were gone. First salvo
> of prority one calls hits the helpdesk. So far, so bad.
>
> I would have expected that PMON should find and kill the
> orphaned servers processes ( we use dedicated server ),
> but it did not. When the firewall came back online, the
> number of server processes doubled and used up all
> remaining memory of our AIX LPAR. OS began to swap and
> db response time went through the roof ( next salvo of
> priority one calls, Director of IT operations comes into
> DBA office 'inquiring').
>
> Luckily the systems people had another 4 GIG of RAM to add
> to the 16 GB already allocated to the LPAR and db perfomance
> began to recover.
>
> Today I asked an Oracle Corp onsite consultant why PMON
> did not get rid of the old server processes. He told me that
> he did not know the algorithm that determines PMON's behaviour,
> but that it could take up to nine hours before it kills
> orphaned servers. But there should be an event, whose event
> number he does not remember, that starts the cleanup
> immediately.

I'm wondering if he is confusing it with an event that does that for SMON.
>
> So my questions are: What is number of this event?
> Has anybody already used it in a similar situation
> and did it work as expected?
>
> Any other suggestions, hints or tips to cope with a situation
> like this?

See metalink Note:1020720.102

http://www.dizwell.com/dokuwiki/doku.php?id=nondbakill is pretty bizarre.

Personally, I've always just very carefully killed the processes from unix manually.

jg

--
@home.com is bogus.
http://www.securityfocus.com/bid/23403/