Problem forming clustur of laptops via router. (MPI)

jayeshsharma52 · 02-23-2013, 07:11 AM

I am trying to form a cluster of two laptops connected through router, I have assigned static ips to both the laptops.
-I have generated key using "ssh-keygen -t rsa"
-copied "**_rsa.pub to "authorized keys"
-compiled mpi_hello.c with mpicc
-but when i give "mpiexec -n 8 -f /home/mpiuse/machinefile ./mpi_hello" command it gives me error saying---
"ssh_askpass: exec(/usr/bin/ssh-askpass): No such file or directory
Host key verification failed"
I followed steps given in this link.
Please suggest me how to move ahead.
Thanks

tronayne · 02-23-2013, 07:54 AM

Thake a look at this article: http://www.linuxjournal.com/article/6602 (11 SSH Tricks), might prove interesting.

What you probably want to do to generate keys is execute ssh_keygen and, when prompted for the passphrase, simply hit the enter key. Do this separately on each machine.

Copy the id_rsa.pub on machine "A" to a file in /.ssh nemed "A" (really, use the actual system name). Do the same thing, individually, on the other system(s), copying the id_rsa.pub file to the name of that machine in the ~/.ssh directory.

Now you're ready to -- on machine A -- get the file(s) from the other machine(s). Use scp to do that; you copy the "B" file (which is the public key for "B") to ~/.ssh, then append that file to ~/.ssh/authorized_keys (you can append a file easily with cat filename >> authorized_keys). Repeat that for any additional machines.

Then get on machine "B" and do the above, getting the public key file from "A."

Once that's complete, you'll be able to ssh, scp, sftp, etc. between the machines without a password, relying on the public-private keys your generated.

Hope this helps some.

jayeshsharma52 · 02-23-2013, 08:55 AM

I have followed exactly same steps on bothe machines(node1 and node0)
-steps followed.
-ssh-keygen -t rsa
-cd .ssh
-cat id_rsa.pub >> authorized_keys
-mpicc mpi_hello.c -o mpi_hell
-mpiexec -n 8 -f /home/mpiuse/machinefile ./mpi_hello
---------------------------------------------------------------
my "machinefile" looks like:
---------------------------
node1:2 #node1 is remote machine
node0:2 #node0 is master machine
-----------------------

tronayne · 02-23-2013, 10:10 AM

You're not coping the public key file on the machine it is generated on into authorized_keys are you? If you are, that's the wrong place to put it -- it goes on the other machines you want to connect to without a password (and the public key file form those machines goes in this machine's authorized_keys file.

Here's a an example. Let's say that I have two machines, fubar and snafu and I want them to connect without a password in both directions.

If you've goofed up your .ssh directories, you can simply rm -rf .ssh and start again.

On fubar

Code:

log in as user_name
ssh-keygen
<enter key twice when prompted for a password>
cd .ssh
cp id_rsa.pub fubar

On sanfu

Code:

log in as user name
ssh-keygen
<enter key twice when prompted for a passphrase>
cd .ssh
cp id_rsa.pub snafu
scp fubar:/home/user_name/.ssh/fubar .
cat fubar >> authorized_keys

On fubar

Code:

log in as user_name
cd .ssh
scp snafu:/home/user_name.ssh/snafu .
cat snafu >> authorized_keys

That's how to do that; you're allowing an external machine to log in in this machine without a password by having its public key on this machine in authorized_keys.

Hope this helps some.

jayeshsharma52 · 02-24-2013, 07:15 AM

I am very thankful for your suggestions as your advice does help me a lot ,and now I am able to connect to other machine directly without password.
But now I am facing some other problem.
As I want to run application on two machines so when I tries -"mpiexec -n 8 -f /home/mpiuse/machinefile ./mpi_hello" terminal simply hangs waiting.
And trying some brute force I figured out that when I remove commented line which is in "vim /etc/hosts" :
#127.1.1.0 jayesh (my name is "jayesh")(on removing comment)
The command "mpiexec -n 8 -f /home/mpiuse/machinefile ./mpi_hello" runs 4 processes and gives error for :unable to run other 4 process
(only 4 processes are running because of define node0:2 means 2 process on node0 and other two on node1- node1:2).
If possible please help providing a right track.
Thanks for showing interest in my case.

tronayne · 02-24-2013, 10:19 AM

I'm not real sure about this, but the manual page for mpiexec state:

Quote:

If you are simply looking for how to run an MPI application, you probably want to
use a command line of the following form:

% mpirun [ -np X ] [ --hostfile <filename> ] <program>

This will run X copies of <program> in your current run-time environment (if run-
ning under a supported resource manager, Open MPI's mpirun will usually automati-
cally use the corresponding resource manager process starter, as opposed to, for
example, rsh or ssh, which require the use of a hostfile, or will default to run-
ning all X copies on the localhost), scheduling (by default) in a round-robin fash-
ion by CPU slot. See the rest of this page for more details.

Which looks to me like you would be running "all X copies on the localhost;" i.e., not a remote host? Again, not sure about that (don't know about mpiexec), but it seems like that might be the case and that you more likely want to use ssh to execute a command on a remote host.

Hope this helps some.

jayeshsharma52 · 02-24-2013, 10:32 AM

Yeah, for local host this command is working perfectcly but when I am trying to merge it with ssh its giving some above dictated errors.
Thanks

shivalig · 03-19-2014, 02:38 AM

Thanks a lot solution worked for me