I also made this observation. The problem with the master running
continuously only appears to happen if the slave that dies happens to be
on the same node as the master. If the slave that exits is on a
different node then the master does not appear to start eating cpu
cycles constantly.
> > A few clarifications:
> >
> > - Are you running on SMPs? I.e., is your master running on the same
> > node as at least one slave, and have slaves running on other nodes,
> > too?
> > - What RPI are you using?
> > - Can you verify that your master is stuck in the 200 loop? I.e., it's
> > looping around the MPI_Waitany and the processing that follows it?
> >
> >
|
