Messages in this thread | | | Date | Wed, 22 Oct 2008 10:29:07 -0400 | From | Daniel Jacobowitz <> | Subject | Re: [RFC v7][PATCH 2/9] General infrastructure for checkpoint restart |
| |
On Tue, Oct 21, 2008 at 08:02:43PM -0700, Dave Hansen wrote: > Let's say you have a process you want to checkpoint. If it uses a > completely discrete IPC namespace, you *know* that nothing else depends > on those IPC ids. We don't even have to worry about who might have been > using them and when. > > Also think about pids. Without containers, how can you guarantee a > restarted process that it can regain the same pid?
OK, that makes sense. In a lot of simple cases you can get by without regaining the same pid; there's an implementation of checkpointing in GDB that works by injecting fork calls into the child, and it is useful for a reasonable selection of single-threaded programs.
-- Daniel Jacobowitz CodeSourcery
| |