Skip to main content
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [ptp-user] IBM LoadLeveler (Eclipse Iuno RC3 Build id: 20120606-2254)
Christoph,

We've only just started testing the LL support, so I'd expect a few bugs in this version.

Would you be able to open a bug for the STDOUT and STDERR settings?

I've never seen any problems with the application executable like you describe. Would you please open a bug with instructions on how to replicate this?

The monitoring is not working because LL 5.1 changed the default output from the llstatus command from 'machine' to 'cluster'. We should have a fix for this in 6.0.1. In the mean time, you can edit the file '.eclipsesettings/rms/LL/da_nodes_info_LML.pl' and change the line 'open(IN," $cmd -l |");' to 'open(IN," $cmd -L machine -l |");'.

Regards,
Greg


On Jun 19, 2012, at 11:52 AM, Christoph Pospiech wrote:

Hi,

 

I just tried to launch a parallel program using run configurations->Parallel Applications->target system configuration "IBM LoadLeveler".

 

I have found a few issues with that.

 

Once you fill in "IBM LoadLeveler" for target system configuration and "SuperMUC" for the connection name, Eclipse appears to correctly query the local LoadL configuration. In particular, the pull down menu for the classes is found correctly.

 

But the fields for STDOUT and STDERR are empty by default. If you leave them empty, you get what you asked for - no output. This might be confusing. Would it be possible to add a non-zero default, say job_name.step_name.[out|err] in the top level directory of the ptp-sync replication ?

 

Specifying the application is a real challenge. The browse button lets you find the correct binary, but in my case at least this led to the following error message.
"Application file does not exist".
Retrying to set the application via the browse button now generates a wrong path like this.
<project_root>/bin/serial_mesh_test/serial_mesh_test
The correct path would have been
<project_root>/bin/serial_mesh_test
Manually correcting the path does not always work. Sometimes you have to submit ignorantly with the wrong path, ignore the expected error message and then retry setting the application path manually. Then the job submits correctly.
To me this looks like a bug.

 

Now the monitoring perspective pops up, but it looks strange - as if there was only the login node, not a whole linux cluster. Is something missing here ?

 

--

 

Mit freundlichen Grüßen / Kind regards

 

Dr. Christoph Pospiech
High Performance & Parallel Computing
Phone: +49-351 86269826
Mobile: +49-171-765 5871
-------------------------------------------------------------------------------------------------------------------------------------------
IBM Deutschland GmbH / Vorsitzender des Aufsichtsrats: Martin Jetter
Geschäftsführung: Martina Koederitz (Vorsitzende), Reinhard Reschke, Dieter Scholz, Gregor Pillen, Joachim Heel, Christian Noll
Sitz der Gesellschaft: Ehningen / Registergericht: Amtsgericht Stuttgart, HRB 14562 / WEEE-Reg.-Nr. DE 99369940

 

_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ptp-user

Back to the top