Hello everyone!
We are implementing a script to automatically kill off idle application
sessions. The way we identify whether it's idle is by the timestamps
obtained by "onstat -g ntt", and we will use "kill -15" to kill off the
"fglgo" process if the sesion is idle for more than a configured length
of time.
We have around 300 users, each day 50 or so of such sessions may be
killed.

Does any one see any problem with this approach? Would this de-stablise
the engine in any way?

OS: Solaris 5.8
IDS 9.40.UC6
IBM INFORMIX-4GL Version 7.32.UC2 Pcode Version 732


Thanks in advance!

Regards,
Carl Y. Wu

Since you already decided to use "onstat -g ntt" the you can also use
session id (sid) of idle session and sent it "onmode -z" to attempt more
gracefull kill.

"Carl Wu" <carl.wu@optus.com.au> wrote in message
news:1161737413.972264.49550@m73g2000cwd.googlegro ups.com...
>
> Hello everyone!
> We are implementing a script to automatically kill off idle application
> sessions. The way we identify whether it's idle is by the timestamps
> obtained by "onstat -g ntt", and we will use "kill -15" to kill off the
> "fglgo" process if the sesion is idle for more than a configured length
> of time.
> We have around 300 users, each day 50 or so of such sessions may be
> killed.
>
> Does any one see any problem with this approach? Would this de-stablise
> the engine in any way?
>
> OS: Solaris 5.8
> IDS 9.40.UC6
> IBM INFORMIX-4GL Version 7.32.UC2 Pcode Version 732
>
>
> Thanks in advance!
>
> Regards,
> Carl Y. Wu
>

Davorin Kremenjas wrote:
> Since you already decided to use "onstat -g ntt" the you can also use
> session id (sid) of idle session and sent it "onmode -z" to attempt more
> gracefull kill.
>

Thanks Davorin.
The application processes are run by a separate user id, and the script
to kill idle processes are also run by that user id. This is an
advantage as that user id does not have DBA permission. "onmode -z"
will have to be run as user "informix", which would be undesirable.