[CRIU] Network unlock issue when checkpointing with -R

Cyrill Gorcunov gorcunov at gmail.com
Thu Oct 31 01:26:23 PDT 2013


On Wed, Oct 30, 2013 at 06:33:00PM -0500, Frederico Araujo wrote:
>    Thanks for the answers!
>    Regarding removing -R, I believe that, from the user's perspective, it
>    would be nice to keep it as option. 
>    Why couldn't we just have CRIU perform a normal checkpoint and then, when
>    -R is passed, it does the restore as post-dump step? No need to change
>    anything on the checkpoint process, right?
>    I might be missing something, but if the problem is only related to the
>    network lock/unlock, I believe this would do.

It sounds like a weird scenario -- you've dumped the task, then "restore" it
immediately? What for? Usually "post-dump" step is supposed to serve task
migration on a new node, thus if migration successed the former task should
be killed, if migration failed -- we bring up dumpee back to life. So I
personally don't see much point in -R option. Again, it's not the problem
to "unlock" network if -R passed, but I think it's wrong.

P.S. I would like to hear a word from Pavel.


More information about the CRIU mailing list