[gridengine users] export of environment variables from start_proc_args

[Dave Love]

William Hay <w.hay at ucl.ac.uk> writes:

> On 6 July 2012 12:43, Mark Dixon <m.c.dixon at leeds.ac.uk> wrote:
...

>> 1) Tightly-integrated openmpi jobs started up orted on slave nodes with a
>> $1 equal to " orted". I added something to strip initial spaces out to
>> fix. Noticed with openmpi 1.4.0.
>
> That looks lime an OpenMPI bug to me.

Oh yes; I'll look into getting it fixed.  I don't understand the code
there on a quick look.  Presumably this is because the starter method
doesn't emulate the builtin supplying -c "..." to the shell.

>>
>> 2) The HP-MPI bundled with Fluent, tightly-integrated with the standard
>> gridengine rsh wrapper, managed to somehow get a $1 equal to "-n". I added
>> something to shift the variables if that happened. The wrapper probably
>> just needs updating to handle to handle multiple "-n"'s or something like
>> that. Noticed with ANSYS 12.0 (or 12.1 - not sure which).

I'm not sure I understand the problem.  Is it specific to using the
starter method?  Fluent 12 apparently works here with the standard
wrapper as "rsh".

>> 3) A command line would sometimes be given to the starter_method via qrsh
>> of the form "FOO=bar; export FOO; <command>".

That's the sort of thing I meant about the exec not being equivalent to
the builtin method (assuming it works with the builtin).  I quite often
do that sort of thing interactively with qrsh.

>> This is when I added the
>> which "if" test to decide on an eval or exec (see my last email).

That's not what I assumed about use of eval.  As far as I remember, when
I was mistakenly trying to emulate the builtin, I found I had to use
eval to evaluate the the shell variable before exec'ing it for reasons I
didn't (try hard to) understand.

>> After I added the third kludge (back in Feb 2010), no more problems have
>> been noticed or reported. We have hundreds of active users, so I hope that
>> this qualifies as being well tested :)

I guess it depends on the specification.  I doubt it will generally
behave the same as without the starter, but it's probably relatively
safe in the absence of Liverpool access to the system.

-- 
Community Grid Engine:  http://arc.liv.ac.uk/SGE/