[gomd-devel] Re: Fwd: <DAEMON> chpox error while restoring proc.

[Gian Paolo Ghilardi]

Hi Eugen.

Thanks a lot for your answer.
You're right: omdiscd (openMosix discovery daemon) uses sockets and this
causes the error.

Your answer is really useful: I'll add a check to verify dumpfile integrity
after creation.

Thanks.

Gian Paolo Ghilardi


----- Original Message -----
From: "Eugen Meshcheryakov" <address@hidden>
To: <address@hidden>
Cc: "csaa" <address@hidden>
Sent: Monday, November 17, 2003 3:43 PM
Subject: Re: Fwd: <DAEMON> chpox error while restoring proc.


> Hello Gian Paolo Ghilardi!
>
> > This is a forwarded message
> > From: Gian Paolo Ghilardi <address@hidden>
> > To: "Olexander Sudakov" <address@hidden>, "gomd developers mailing
list" <address@hidden>
> > Date: Sunday, November 16, 2003, 10:46:14 PM
> > Subject: <DAEMON> chpox error while restoring proc.
> >
> > [CHP]: restoring the process...
> > [CHP]: invoking "ld-chpox /tmp/omdiscd_20031115_113908_1939.dump"
> > CHPOX_IOCTL_CKINIT: Exec format error
> > ...
> > <<<ERROR_LOG>>>
> >
> > In attachment the full log.
> > I've followed exactly the commands reported on:
> > http://www.unicc.kiev.ua/tasks/chpx_eng.html
> > Any idea?
> >
>     This looks like dump file created marked as 'bad'. To check this
> you can do one of the following:
>    - see content of '/proc/chpox/info' file. If checkpoint was
> successfuly completed then letter O should be in square brackets:
>      8151:31:9 [O|1] -> /tmp/tt [1069079436.9082]
>                 ^^^
> if checkpoint failed you will see entry like this:
>      8384:31:9 [E|1] -> /tmp/tt [0.0]
>                 ^^^
>
>    - try to execute (chpox.magic file can be found in chpox dist.):
>      file -m chpox.magic <dump file>
> you will see:
>      /tmp/tt: CHPOX image version 0.1 i386 CORRUPTED
>                                            ^^^^^^^^^
> or:
>      /tmp/tt: CHPOX image version 0.1 i386 complete 1 child process(es)
>                                            ^^^^^^^^
>     The reason of marking file as corrupted may be that program
> you try to checkpoint uses some unsupported things like sockets.
> To see more info you can configure chpox with flag --enable-debug
> (this does not work with gcc-2.95) and see messages on the console
> or in the log file.
>
> > I'd like to know if it's possible to invoke restoring without using
> > ld-chpox.
> >
>     It is possible, but restoring API can be changed in future versions.
> See source code of ld-chpox for details.
>
> Regards,
> Eugeniy Meshcheryakov
>
>
>