Great suggestions!   And yes - I agree this probably isn't worth automating
- that's why I was looking for a 'native' solution.

But this has all given me some ideas about how to provide the Linux teams
administering the Linux servers a way to easily restart servers:

Example:  Give a guest used by Linux support authority to do CP SEND ..
then they can issue 'vmcp send cp guestid ipl 200 clear'  on Linux to
restart the guestid server..   I know I was looking for a complete restart..
but a way to reIPL the guests might be useful more of the time.  (this has
some security issues, I know ;-)

Thanks!   Scott

On Thu, Oct 30, 2008 at 12:01 AM, Bruce Hayden <[EMAIL PROTECTED]> wrote:

> Change your step that says "go into a loop" to say to wait for a
> message from the *VMEVENT system service telling you the userid has
> logged off.  Once you get that message, then you can autolog..
>
> I think an improvement would be to tell the "RECYCLVM" the Linux id
> that is going to be recycled (it could be from Linux itself) and then
> have the Linux id issue a signal shutdown to itself and the service
> machine would wait for the VMEVENT information as before.  That way,
> you wouldn't have to set up some kind of authorization scheme so
> unauthorized people could recycle your Linux servers.  Since the Linux
> server shuts itself down, you presume a root user did it.
>
> Pipelines can connect to *VMEVENT, but a better option would be to use
> the support in VM operations manager to do it and have it manage the
> whole thing.  But, if you have the operations manager, you could just
> have it watch the console for a certain message to appear (Linux
> executed echo "$$$RECYCLE ME$$$" > /dev/console) and then just do the
> shutdown and restart in an action routine...
>
> Personally, I think this is a rare enough event that it isn't worth
> automating.  You can dynamically add about anything now (CPUs, disks,
> memory in the future) so the need for a recycle because of a directory
> change would be more and more rare.
>
> On Thu, Oct 30, 2008 at 12:51 AM, Scott Rohling <[EMAIL PROTECTED]>
> wrote:
> > I didn't explain something very well in my initial post..
> >
> > In the 2nd paragraph, where I talk about a CMS guest..
> >
> > I meant to say that I could have a guest running CMS who could captures
> MSGs
> > or SMSGs sent to it -- and have a RECYCLE EXEC it ran if it received a
> > message of 'recycle' from a Linux (or any other type of) guest..   The
> > RECYCLE EXEC would:
> >
> > -  Accept an argument of the userid to be recycled
> > -  Issue a CP SIGNAL SHUTDOWN userid      (userid = the userid who sent
> the
> > msg)
> > -  Go into a loop and wait for CP Q USERID userid   to get an RC45
> > (either the Linux guest will logoff or the SIGNAL timeout will log it
> off)
> > -  CP XAUTOLOG userid once the above loop ends
> >
> > Let's say we set up a user called RECYCLVM who is a CMS guest who waits
> for
> > a MSG of 'RECYCLE'... (via WAKEUP) ..   it would call RECYCLE EXEC
> passing
> > the userid of the user who sent the message.
> >
> > So now a Linux guest can issue:   'vmcp msg recyclvm recycle'
> >
> > I know I can do that fairly easily - but I think z/VM itself could do it
> > fairly easily too.  Whether it's worth the development effort -- umm.....
> > Maybe the process above is cheapest  ;-)
> >
> > Thanks - Scott
>
> --
> Bruce Hayden
> Linux on System z Advanced Technical Support
> IBM, Endicott, NY
>

Reply via email to