This is an old revision of the document!
I worked with Danny and Andy to check that the steps I used with MPI are properly documented in the Lab. Turns out there are a couple of places where things go wrong:
catcommand, you can cut/paste from the lab but check the
»symbol. The shell will only recognize two greater-than signs. Anything else will cause this step to fail.
catcommand two more times for the other two computers you want in your cluster
It likely worked if the first time it asks for your password and then just returns. The next two should just return, without asking for the password.
The other thing to fix is your
host_file. You want not just the name of the computer, but the full domain name of the computer. For example, instead of
farheen, you would write
farheen.sewanee.edu. For example my host_file looks like this:
farheen.sewanee.edu eneva.sewanee.edu hive.sewanee.edu oscar.sewanee.edu
Error I received when trying to ssh into another computer: signing failed: agent refused operation Solution: I received this error because I had put a passphrase on my private key.
Fix it by doing ssh-add