[mpich-discuss] mpdboot fails

[Rajeev Thakur]

Can you try the latest release of MPICH2, 1.3. It uses a different process manager, called hydra, by default. You don't need to run mpdboot.

Rajeev

On Oct 25, 2010, at 10:55 AM, Matt Thiffault wrote:

> So, I have a number of machines I'm trying to set up with MPICH2, but I'm starting off with just 2. One is behind an NAT router and the other has a public ip, but is protected by an iptables firewall. 
> 
> I can ssh back and forth between them, and I have them keyed so that you don't need a password to do so after starting an ssh-agent and giving the password to the RSA key. 
> 
> I have MPICH_PORT_RANGE and MPIEXEC_PORT_RANGE set in /etc/profile on both systems to 10000:10100 and both firewalls allow traffic through on those ports (I've tested this with netcat). 
> 
> mpdboot fails with the following output:
> mthiffau at foehammer ~ (255)% mpdboot -n 2 -v -f mpd.hosts
> running mpdallexit on foehammer
> LAUNCHED mpd on foehammer  via  
> RUNNING: mpd on foehammer
> LAUNCHED mpd on halo.mthiffau.ca  via  foehammer
> mpdboot_foehammer (handle_mpd_output 415): failed to connect to mpd on halo.mthiffau.ca
> 
> So, mpd starts locally, and on the other machine, but afterwards comunication seems to break down. Does the ssh-ing back and forth have to be completely passwordless? Is there something else I have to do to ensure the correct port range gets used?
> 
> I suspect it might be the latter, as even with the environment variables set, mpdcheck -s starts listening on random ports that are outside the range specified.
> 
> Any help would be appreciated, thanks a bunch.
> 
> Matt Thiffault
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss