[petsc-dev]   How to compose a git URL for going through 2 machines

Nystrom, William David wdn at lanl.gov
Thu Nov 6 18:02:19 CST 2014


Just to wrap this up, I was able to get my git clones to work finally by making a git_ssh
script and then point git at it via the GIT_SSH environment variable.  My git_ssh script
looked basically like the following:

$ cat git_ssh
#!/bin/bash
shopt -s extglob
if [[ $1 = @(lo|cj|mp|pi)-fe*(1|2) ]] ; then
        ssh wdn at red-wtrw.lanl.gov ssh "$@"
else
        ssh "$@"
fi

export GIT_SSH=/path/to/git_ssh

After that git clone lo-fe:repo works.

Thanks to all that provided suggestions.  The above was a suggestion I received via
another email thread.  None of the port forwarding or ssh tunneling approaches worked.
So it was nice to discover that the above approach worked.

Dave

________________________________________
From: Nystrom, William David
Sent: Thursday, November 06, 2014 1:39 PM
To: Jed Brown; Åsmund Ervik; petsc-dev at mcs.anl.gov
Cc: Nystrom, William David
Subject: RE: [petsc-dev]   How to compose a git URL for going through 2 machines

Well, I have contacted the consultants at both LANL and LLNL to discuss the problem
and seek a solution.  At the moment, I have not even been able to get the more
traditional solution of port forwarding using the -L option of ssh to work - the most
recent suggestion by Satish.  Very frustrating.

________________________________________
From: Jed Brown [jed at jedbrown.org]
Sent: Thursday, November 06, 2014 1:13 PM
To: Nystrom, William David; Åsmund Ervik; petsc-dev at mcs.anl.gov
Subject: Re: [petsc-dev]   How to compose a git URL for going through 2 machines

"Nystrom, William David" <wdn at lanl.gov> writes:

> Checked in to this a bit and the version of openssh being used in
> my situation is 5.3.x and is not something I can control.

Dave, insistence on restricting the network and refusal to upgrade are
choices made by LANL.  Those choices have a very real cost, but that
cost should be borne by the organization making those decisions, not the
broader PETSc development community.  Use netcat if need be, etc., but
above all, please make the organization feel the cost of those
decisions.  If the cost is not visible, such policies tend to grow
unchecked, ultimately wasting everyone's time.



More information about the petsc-dev mailing list