[SimGrid-user] MSG_TRANSFER_FAILURE errors

Martin Quinson martin.quinson at loria.fr
Mon Nov 25 20:24:15 CET 2013


On Sun, Nov 24, 2013 at 11:47:36AM +0100, Christophe ENDERLIN wrote:
> 
> While I was working at building an example, I found an error in my code that was causing this strange behavior. So now it works right.

I'm glad!

> I’ll just keep on monitoring all this, since things aren’t absolutely clear for me : when a recipient host stops, a TRANSFER_FAILURE error is raised, ok. But what if host is ok, but if the recipient process running on it dies ?
> I have to make this all clear for me …

Well. We could dig into the source code to see what it's supposed to
do in this case, but I think that it's easier (and safer) to actually
test it.

My guess is that if the receiver process quits during the message
reception, bad things will happen. It may be as bad as segfaulting
the simulator. We've fighted for months to make it work in this case,
but that's a difficult one. If you find a case where it segfaults,
please send it to us so that we can fix it. One day, it will only
raise a TRANSFER_FAILURE on the sender side, but I'm not sure of
whether we are there yet.

Bye, Mt.

-- 
Chaque génération, sans doute, se croit vouée à refaire le monde. La
mienne sait pourtant qu'elle ne le refera pas. Mais sa tâche est
peut-être plus grande. Elle consiste à empêcher que le monde se défasse.
          --- Albert Camus, Prix Nobel de littérature 1957.



More information about the Simgrid-user mailing list