Hi managers,
a addon to this question: Has anybody got his hand on the checkpoint code from CRAY Unicos
or written a version for dUnix?
Regards,
Rudi Gabler
-----Urspr|ngliche Nachricht-----
Von: Rene Demian Taboada Arias [SMTP:rta_at_aixulsa.ulsa.mx]
Gesendet am: Mittwoch, 2. September 1998 18:00
An: alpha-osf-managers
Betreff: [SUMMARY] : How to stop and restart process in between boot...
>ORIGINAL POST:
>
>
>HI MANAGERS!! I wonder if any of you can recommend me (if its
>possible to do) a safe method for stopping a process, have a reboot, and
>restart the process exactly at the point where it has stopped. (some kind
>of reg's dump and reload, who knows?) Any help will be greatly apreciated
>TIA.........
>
>
>SUMMARY
>
>Thanks to all who toke a minute to respond me. The answer's what was
>expected :
>
> There's no answer!
>
>Anyway, here is some advice if you try programming:
>
>***************************************************************************
*
>*****************
>FROM Dr. BLINN
>There is no simple way to do this in full generality. There is nothing in
>the DIGITAL UNIX operating system at the present time to facilitate this.
>You could write an application that periodically saves enough state to be
>able to resume a computation (including re-opening any input and output
>files, etc.) and then restart it from the last checkpoint, or you could
even
>have the application accept a signal telling it that it's time to write out
>a checkpoint (just before shutting the system down).
>But there is no general support for this. It's entirely "roll your own" at
>this point.
>Tom
>***************************************************************************
*
>****************
>
>FROM alan_at_nabeth.cxo.dec.com
>
>I don't believe the system supports a way of doing check-point
>restart. So, if you'd have to do it by hand. Select a signal
>for the program to recognize and arrange for signal handler.
>When the signal is received, save all the program state. Then
>when the programs starts again, it checks to see if there is
>saved state, restores it and continues.
>You may want to take care that the check-point doesn't occur
>in the middle of something inconvient. In this case, the
>signal handler, would set a flag that could be check at each
>iteration of the program and when set, the state is saved
>then.
>
>
Received on Wed Sep 02 1998 - 05:50:34 NZST