Hey all,


I want to do sonme testing of the nodes on a cluster I am using, testing to see if they run mpi. So I need to check every node by itself first.

which means that I need a machinefile that lists one node at a time:

node2

for example

I want to write a script to test each node in order: in pseudocode:

for i in 2:n
open machinefile
sed s/i-1/i/ #replace the previous number with the new one, will this work?
execute mpi call
do something with the return value
end


So: can someone show me how to do this? I am completely new to shell scripts, so the syntax is what is slowing me down rather than the logic. In the end, this should do something like:


Edit machinefile to contain

node2

run mpi on node 2, and tell me if it worked


Edit machinefile to contain

node3

run mpi on node 3, and tell me if it worked

etc

But I really only need help with the editing the file part of it

#!/bin/bash
for i in {2..71}
do
echo node2 | sed s/[0-9]*/$i/
done





That just appends the value of i before the node2... how can I make is replace the 2 with i?


EDIT:

This works:

#!/bin/bash
for i in {2..71}
do
echo node2 | sed s/\[0-9]\/$i/
done

but when I try to make it act on the file:

#!/bin/bash
for i in {2..71}
do
sed s/\[0-9]\/$i/ <machinefile >machinefile
done

I just get an empty file at the end. Ideas?

Yup. Can't redirect from a file to the same file.
Either use a tmp file, or sed with -i.


Cheers,
Tink

P.S.: Use code tags