Bug 4778 - WS-Fork job manager doesn't set environment up for mpi jobs
: WS-Fork job manager doesn't set environment up for mpi jobs
Status: RESOLVED WONTFIX
: GRAM
wsrf scheduler interface
: unspecified
: IA64 All
: P3 normal
: ---
Assigned To:
:
:
:
: 5484
  Show dependency treegraph
 
Reported: 2006-10-13 16:27 by
Modified: 2012-09-05 11:43 (History)


Attachments


Note

You need to log in before you can comment on or make changes to this bug.


Description From 2006-10-13 16:27:11
The PBS job manager, when running mpi jobs, constructs a script for the
executable that contains the relevant environment variables, and then runs that
script with mpirun (or mpiexec).  The Fork job manager evidently does not
construct such a script, and just runs the executable with mpirun.

This was reported by Brian Toonen, but he doesn't like entering bugs in
bugzilla.
The fix here may just be as simple as cutting and pasting some code from PBS.pm
to Fork.pm.
------- Comment #1 From 2006-10-13 17:32:13 -------
As a note, it is clear that mpich-gm's mpirun doesn't pass through env
variables by default, but with --gm-copy-env, it will.

So for Brian's purposes right now (running the test suite for MPIG), we've got
mpirun set up with --gm-copy-env for fork.  Thus, urgency on this is not really
so high.
------- Comment #2 From 2006-10-31 15:59:02 -------
It sounds like this is not necessarily a fork job manager issue but relates to
how mpi is deployed. If you have suggestions for reliably detecting that this
option is needed, we can add them to the setup scripts. Otherwise, it might be
best to just document this issue in the release notes.

joe
------- Comment #3 From 2008-02-05 15:18:07 -------
We should make sure the doc is clear about

1) what the Fork environment is
2) How MPI is run and any known issues like the mpich-gm env copy issue in this
bug
------- Comment #4 From 2012-09-05 11:43:24 -------
Doing some bugzilla cleanup...  Resolving old GRAM3 and GRAM4 issues that are
no longer relevant since we've moved on to GRAM5.  Also, we're now tracking
issue in jira.  Any new issues should be added here:

http://jira.globus.org/secure/VersionBoard.jspa?selectedProjectId=10363