Re: Killing an user from the user

Ed Prochak <ed.prochak_at_magicinterface.com> wrote in message news:<sKU1c.1588$xL.475_at_fe03.usenetserver.com>...
> Yong Feei Shun wrote:
>
> > Dear all,
> >
> > I am having a question here. hope you guys could help. Whenever
> > i kill a user, i cant kill the oracle statement there, eventhough the user
> > had been killed. What should be the problem? Does this mean that at the
> > background that session still running? I encountered system slow down
> > problem recently and cause me to kill the user whoever using the sesssion.
> > Please help. Thanks
> >
> >
> >
> > Rgrds
> > Feei Shun
> >
> >
> If you are on a UNIX system I hope you didn't use kill -9
>
> Please try using other signals first. I usually try, in order,
> -1, -2, -3, -4, -15, and only then do I use -9
> the other signals give the programs a chance to clean up after themselves.

That seems overkill. -15 is the default for a reason. Kill, then kill -9 ought to be more than enough. If you kill a parent process with an async child, you might make a zombie either way. So a better way might be to use ps to determine parents and children, kill children first. I usually cut and paste off the ps onto a single kill line so I don't typo them (if I haven't written a shell script for it based on an app's characteristics).

I got the idea the OP was talking about alter system or it's OEM equivalent. That gets cleaned up by SMON, which might never get around to it. Unix kill is generally handled by PMON, which normally gets right to it. So I strongly prefer the unix kill.

>
> And keep in mind a simple guideline: even when the session is gently stopped,
> I found it can take almost as much time for ORACLE to rollback the work as it
> did to start with. IOW your SQL query ran for 3 hours without showing results
> and slowing the system down in the meantime. If you kill it, don't be
> surprised if the system remains slow for an equivalent amount of time (an hour
> or more). While tearing down a connection seems like less work, there is real
> work that must be down to free the resources used.

Too true. If it is some huge update, you might not want to be killing it unless you know the consequences. More often it's just some hog report and you just want it to go away so online users can work.

jg

--
@home.com is bogus.
We all know who our friends are
http://news.com.com/2100-1008_3-5170109.html