[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
Re: [ptp-dev] PTP and epp-parallel features; Juno M7 deadline is Monday - feature freeze
|
Greg,
The changes you made to the job submission
specification in the XML seem to do what I want, but I've just tested the
change and there is a problem. When doing the remote job submission,
the command executing, according to "ps xw" on the remote host
is:
tcsh -c /bin/sh -c 'echo "PID=$$
PIID=40" > /dev/pts/5; env -i "MESG2=World" "PATH=/bgsys/drivers/ppcfloor/bin:/usr/local/Modules/3.2.9/bin:/usr/local/qsub/2.1.11/bin:/usr/local/jobmap/1.2/bin:/usr/local/slurm/2.2.7/bin:/bgsys/drivers/ppcfloor/comm/xl/bin:/vlsci/IBM/swail//bin:/vlsci/IBM/swail//local/bin:/vlsci/IBM/swail//bin/aufslib:/vlsci/IBM/swail//Projects/bin:/vlsci/IBM/swail//Palm/bin:/usr/local/bin:/usr/local/sbin:/usr/ucb:/usr/bin:/usr/sbin:/usr/etc:/sbin:/bin:/etc:/usr/bin/X11:."
"MESG=Hello World" "MMCS_SERVER_IP=10.4.0.24" cd /vlsci/IBM/swail
&&env -i "MESG2=World" "PATH=/bgsys/drivers/ppcfloor/bin:/usr/local/Modules/3.2.9/bin:/usr/local/qsub/2.1.11/bin:/usr/local/jobmap/1.2/bin:/usr/local/slurm/2.2.7/bin:/bgsys/drivers/ppcfloor/comm/xl/bin:/vlsci/IBM/swail//bin:/vlsci/IBM/swail//local/bin:/vlsci/IBM/swail//bin/aufslib:/vlsci/IBM/swail//Projects/bin:/vlsci/IBM/swail//Palm/bin:/usr/local/bin:/usr/local/sbin:/usr/ucb:/usr/bin:/usr/sbin:/usr/etc:/sbin:/bin:/etc:/usr/bin/X11:."
"MESG=Hello World" "MMCS_SERVER_IP=10.4.0.24" sbatch
/vlsci/IBM/swail/af43fbae-4f95-47e1-8665-a2e18dcb87admanaged_file_for_script'
The response in PTP is:
sbatch Exited with value: 127
env: cd: No such file or directory
Job Submit Failed
Closely looking at the command, I think
the problem is the "cd /vlsci/IBM/swail" which is executed by
the first "env" command. As "cd" is a shell built-in
and the "env" command is expecting an executable to run, it is
failing. I think then the "sbatch" is then failing because
of the previous failure. When I try to execute the second "env"
command directly on the command line it works fine. I suppose the
solution is to remove the first "env" command that does the "cd"
and just have the "sbatch" execution. Hopefully this should
work.
I've also updated the JAXB XML definition
for the new "sbatch" job submission and it is attached. I
noticed the schema has changed for Juno - do you need updates of the other
XML files, or has that already been done by you?
Sorry for the delay in testing, but
I've been flat out preparing for our BG/Q delivery and go-live!
I'll copy the above into Bugzilla for
this issue to keep the tracking up to date.
Regards,
Simon Wail, Ph.D
|
HPC Specialist
|
| IBM Research Collaboratory
for Life Sciences - Melbourne
|
|
phone:
| +61 3 9035-4341
fax: +61 3 8344-9130
|
address:
| VLSCI, Gnd Floor, 187 Grattan St
|
| Carlton VIC 3010 Australia
|
email:
| simon.wail@xxxxxxxxxxx |
|
From:
Greg Watson <g.watson@xxxxxxxxxxxx>
To:
Parallel Tools Platform
general developers <ptp-dev@xxxxxxxxxxx>,
Date:
09/06/2012 06:58 AM
Subject:
Re: [ptp-dev]
PTP and epp-parallel features; Juno
M7 deadline is Monday - feature freeze
Sent by:
ptp-dev-bounces@xxxxxxxxxxx
Simon,
Are the changes I added adequate for you? Do you have
updates to the SLURM RM? The last possible date to get anything into Juno
is Tuesday.
Greg
On May 7, 2012, at 1:56 AM, Simon Wail wrote:
Greg,
Glad to hear all the contrib RM's will be included in the Juno release,
particularly SLURM for Blue Gene :-)
I will have some changes in the near future (hopefully before the final
Juno build) to support sub-block jobs on the Blue Gene/Q with SLURM. The
reason I can't submit the changes now, is because I don't have a working
SLURM simulator that supports sub-block jobs (it seems to be broken), nor
do I have a real BG/Q running SLURM on which I can test my changes. I
assume my changes can be submitted as a "bug" fix prior to the
Juno final release build.
For support of the BG/P SLURM RM we do need to fix the escape character
problem I emailed about last week. Let me know if I can help on that
issue.
Regards,
Simon Wail, Ph.D
|
HPC Specialist
|
<Mail Attachment.gif>
| IBM Research Collaboratory
for Life Sciences - Melbourne
|
<Mail Attachment.gif>
|
phone:
| +61 3 9035-4341
fax: +61 3 8344-9130
|
address:
| VLSCI, Gnd Floor, 187 Grattan St
|
| Carlton VIC 3010 Australia
|
email:
| simon.wail@xxxxxxxxxxx |
|
From: Greg
Watson <g.watson@xxxxxxxxxxxx>
To: Parallel
Tools Platform general developers <ptp-dev@xxxxxxxxxxx>
Date: 05/05/2012
05:57 AM
Subject: Re:
[ptp-dev] PTP and epp-parallel features; Juno
M7 deadline is Monday - feature freeze
Sent by: ptp-dev-bounces@xxxxxxxxxxx
Yes, I think org.eclipse.ptp.rm.contrib should be included. I'm probably
going to separate the generic configurations out of this plugin and into
jaxb.core so they are included automatically.
Greg
On May 4, 2012, at 2:26 PM, Alameda, Jay wrote:
Beth,
After seeing Greg’s excellent presentation on the RM re-org, I think the
issue about org.exlipseptp.rm.jaxb.contrib
is still relevant – these may be more hidden (ie, not show up until
we make a connection), but we need to decide what goes into the EPP-parallel
and what doesn’t – ie, with the SR2 release, all of the contrib RMs (actually
jaxb xml documents) were not loaded until we did an update from the PTP
update site --
Jay
From: ptp-dev-bounces@xxxxxxxxxxx
[mailto:ptp-dev-bounces@xxxxxxxxxxx]
On Behalf Of Beth Tibbitts
Sent: Friday, May 04, 2012 12:30 PM
To: Parallel Tools Platform general developers
Subject: [ptp-dev] PTP and epp-parallel features; Juno M7 deadline
is Monday - feature freeze
I am finally revisiting this today.
For PTP (part of PLDT) we are adding two new features:
1. openacc
2. openshmem
And I'm looking at features we need to put in the Juno repository, so that
we can build them in the package.
Attaching end of thread which this discussion after the M6 build.
M7 build for us is next Tuesday
Also the discussion in http://dev.eclipse.org/mhonarc/lists/ptp-dev/msg06304.html
of 3/22/12
We were deciding what to do about..
org.eclipse.ptp.cdt.compilers
org.exlipseptp.rm.jaxb.contrib (new RM re-org may make this moot)
org.eclipse.ptp.pldt.fortran...
Any other features we need in the parallel package that aren't in M6?
Note that Tuesday's build is our last chance to add features to Juno.
So we should check in no later than Monday if possible; often they
grab our build early on +2 day.
...Beth
Beth Tibbitts
Eclipse Parallel Tools Platform http://eclipse.org/ptp
IBM STG - High Performance Computing Tools
Mailing Address: IBM Corp., 745 West New Circle Road, Lexington,
KY 40511
<image001.gif>Jeffrey
Overbey ---03/22/2012 05:52:50 PM---> > In order to build parallel
package juno M6 we tried to include these new
<image002.png>
From:
| <image003.png>
Jeffrey Overbey <jeffreyoverbey@xxxxxxx>
|
<image002.png>
To:
| <image003.png>
Parallel Tools Platform general developers <ptp-dev@xxxxxxxxxxx>,
|
<image002.png>
Date:
| <image003.png>
03/22/2012 05:52 PM
|
<image002.png>
Subject:
| <image003.png>
Re: [ptp-dev] PTP Juno repository does not contain features that Parallel
package needs
|
<image002.png>
Sent by:
| <image003.png>
ptp-dev-bounces@xxxxxxxxxxx |
In order to build parallel package juno M6 we tried to include these new
features
...
But they are not in the common Juno Repository
Just to clarify... it's required that every feature in the EPP package
also be in the Juno repository?
Something feels odd about contributing the jaxb.contrib stuff to the common
repo/PTP core... I think because we're filling it with machine-specific
RMs, and the Juno repo targets the world. It would be really nice
if the machine-specific RMs (e.g., for Forge, Lonestar, and Blue Waters)
would only show up when you connect to that specific machine.
We discussed what (not) to do with the cdt.compilers stuff in a previous
thread <http://dev.eclipse.org/mhonarc/lists/ptp-dev/msg05736.html>,
although merging it into an existing feature is fine with me. Same
with pldt.fortran -- the reasons for separating it are in <https://bugs.eclipse.org/349385>,
but merging it into core PLDT is fine with me if you don't mind the Photran
dependency.
Jeff_______________________________________________
ptp-dev mailing list
ptp-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ptp-dev
_______________________________________________
ptp-dev mailing list
ptp-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ptp-dev
_______________________________________________
ptp-dev mailing list
ptp-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ptp-dev
_______________________________________________
ptp-dev mailing list
ptp-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ptp-dev
_______________________________________________
ptp-dev mailing list
ptp-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ptp-dev
Attachment:
Slurm-BGP-Batch.xml
Description: Binary data