[Paraview] trouble with mpi 2, mpdboot

nick nickfzx at aol.com
Mon Nov 8 06:47:07 EST 2004


Hi

I am trying to setup MPICH 2 correctly.

I am having trouble getting mpdboot to work as it is supposed to.  I am
able to ssh all 7 nodes in our small cluster.  And all nodes are setup
in an identicle way. But when I run: mpdboot -n 7, the terminal returns
these error messages:

[root at 192 root]# mpdboot -n 7
mpdboot_rank_0 (mpdboot 269): starting local mpd on 192.168.1.206
mpdboot_rank_0 (mpdboot 321): starting remote mpd on 192.168.001.207
mpdboot_rank_0 (mpdboot 335): starting remote mpd on 192.168.001.208
mpdboot_rank_2 (mpdboot 269): starting local mpd on 192.168.001.208
mpdboot_rank_2 (mpdboot 321): starting remote mpd on 192.168.001.211
mpdboot_rank_2 (mpdboot 335): starting remote mpd on 192.168.001.212
Host key verification failed.
mpdboot_rank_1 (mpdboot 269): starting local mpd on 192.168.001.207
Host key verification failed.
mpdboot_rank_1 (mpdboot 321): starting remote mpd on 192.168.001.209
mpdboot_rank_1 (mpdboot 335): starting remote mpd on 192.168.001.210
Host key verification failed.
Host key verification failed.


When I run mpdboot with 3 processes or less it works find:

[root at 192 root]# mpdboot -n 3
mpdboot_rank_0 (mpdboot 269): starting local mpd on 192.168.1.206
mpdboot_rank_0 (mpdboot 321): starting remote mpd on 192.168.001.207
mpdboot_rank_0 (mpdboot 335): starting remote mpd on 192.168.001.208
mpdboot_rank_2 (mpdboot 269): starting local mpd on 192.168.001.208
mpdboot_rank_1 (mpdboot 269): starting local mpd on 192.168.001.207
[root at 192 root]#

But when I go to over 4 or more it breaks, I can see that the ranks are
being used multiple times and are therefore returning host key errors. 
But I am not sure why the same rank is being used for different machines.

I have tried doing mpdboot -n 3 with a different 3 nodes and it works 
just fine.

My host file looks like this:

192.168.1.207
192.168.1.208
192.168.1.209
192.168.1.210
192.168.1.211
192.168.1.212

I have not included the localhost in this list.  Is this the right
format for mpich 2 machine files…is there someway of specifying for each
node?

I have spent a long time problem solving this one but I am out of ideas.
Thanks for the help

----------------------------------------
Nick Franklin
Computer vision startup
Tel: (650) 363 1540
Email: nickfranklin at gmail.com
----------------------------------------



More information about the ParaView mailing list