[coyotos-dev] Safe states for userspace

[Sam Mason]

On Wed, Aug 01, 2007 at 12:04:18PM +0200, Pierre THIERRY wrote:
> As I was presenting my paper on persistence in a workshop of ECOOP, the
> mention of orthogonal persistence triggered a question about the
> relation between checkpointing and userspace applications: the
> checkpoint is guaranteed to save only consistent states of the system,
> but is there a way for an application to declare when it has reached
> such a consistent state itself?
>
> Else, what would happen when the application is in a inconsistent state,
> the system is checkpointed, shut down, and restarted?

You don't seem to mention it explicitly anywhere, but I assume you're
talking about some sort of distributed computation.  When you mention
the application, you're talking about some aggregation of applications
running in multiple independent systems.

This appears to require the checkpointing of these distinct systems
to become synchronised which could be accomplished by giving the
application is question a capability to the system's Checkpoint object
and stopping whatever else causes the checkpointing to routinely occur.

> The problem is, I suppose in some cases it could lead to a crash of the
> application. With non orthogonal persistence, you would still have the
> serialized state at hand, be able to restart the application, inspect
> what goes wrong, and maybe try to tweak the state to get it to some
> consistent state.
> 
> But in Coyotos, if the application crashes because of an inconsistent
> state, won't its storage be reclaimed? In which case, well, game is
> over.

I was under the impression that when a program crashes it just ceases to
execute instructions, it's the processes' keeper that decides what to do
the process when it crashes.


  Sam