[torqueusers] parallel shells

Daniel Widyono widyono at seas.upenn.edu
Sun Jul 24 23:01:52 MDT 2005


Forgive me if I should have seen this somewhere, but:

1) search through my records of this list turned up empty,

2) google search did not turn up what I need (see below),
   only partial solution

I would like to use a parallel shell for sysadmin purposes.  That is, I want
to access, as root, all nodes to which I can login via ssh.  pbsdsh only runs
within a job.  I need to get all nodes, not just those assigned to a
particular job, and I need to be able to do this even while other jobs are
running.

I see a few references to using pdsh like so: "pdsh -a".  However, I
downloaded what I believe to be the latest version of pdsh source (2.5-1, as
src.rpm), from ftp://ftp.llnl.gov/pub/linux/pdsh/ (dated 7/21/2005).  That
version does not support a "-a" option.

Can someone illuminate my troubles?  Other desired responses are: what do
*you* use to run simple administrative tasks in parallel, without taking up
many system resources on the main node (e.g. I can easily do "for node in
...; do ssh $node blah & done" but that doesn't scale with regards to open
sockets, sane error reporting, tight integration with resource manager, etc.)

With regards,
Dan W.


More information about the torqueusers mailing list