Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [ptp-user] Eclipse PTP on K - monitoring

Hi,

In order to add monitoring feature to Eclipse PTP on “K” computer I am working on a customised LML-layout. For tests I use a smaller sibling of “K” named “FX10”.  

I customised the perl scripts and layout description file (samples/layout_default_PJM.xml), and now in the monitoring perspective I can see a table with the jobs running on the computer. Nodes layout that I created with nodedisplaylayout tag is displayed correctly. 

What I can’t figure out is how I can map jobs to the nodes layout. There is information on the nodes used by every job in the jobs table, but I’m not sure if its format is correct.

Here is what I see:
Job 75612 is not displayed on the nodes layout. 


In temporary files directory in datastep_LML2LML.xml file I have:

<nodedisplay id="nd_1" title="system: fx02p08">
<scheme>
    <el1 tagname="chassis" min="0" max="6" mask="0xFF0100%01d">
        <el2 tagname="node" min="0" max="16" mask="%01x">
        </el2>
    </el1>
</scheme>
<data>
    <el1 oid="empty" min="0" max="6">
    </el1>
</data>
</nodedisplay>


<data> is empty and so no job is displayed in nodes layout. 


By the way, nodes are numerated with hexadecimal numbers. Is it OK to use mask “%01x” ?

Kind regards,
Peter





On 19 Jan, 2015, at 18:13, Carsten Karbach <c.karbach@xxxxxxxxxxxxx> wrote:

Dear Peter,

thanks for the hint on the broken link. It was moved to
http://llview.fz-juelich.de/LML/OnlineDocumentation/lmldoc.html. I have
updated the documentation page with the new link. On the LML
documentation page you can find a section about LML layout files, which
is located at
http://llview.fz-juelich.de/LML/OnlineDocumentation/layouts.html.

There are also some layout examples integrated into PTP. You can find
them at
http://git.eclipse.org/c/ptp/org.eclipse.ptp.git/tree/rms/org.eclipse.ptp.rm.jaxb.contrib/data.
E.g. take a look at the configuration files
de.fz-juelich.judge.torque.batch.xml, de.fz-juelich.juqueen.ll_bg.xml or
de.fz-juelich.juropa.torque.batch.xml. They all contain <monitor-data>
elements, which represent the layout definitions for each site.

See also
https://wiki.eclipse.org/images/7/7e/Carsten-Karbach-31july2013-Slides.pdf
for an introduction to creating your own monitoring layouts.

Regarding your first question: A summary on the monitoring architecture
is given in the presentation here:
https://wiki.eclipse.org/images/d/d0/PTPUserDev2012_Monitoring_Karbach_Frings.pdf.
When you need to develop your own batch system adapter, you basically
have to write scripts for gathering three types of data: jobs, nodes and
global system information. You have to write one script for each of
these types, where each script generates an LML file. All subsequent
steps of LML_da should be handled automatically. Find examples for these
scripts for all supported batch systems at
http://git.eclipse.org/c/ptp/org.eclipse.ptp.git/tree/rms/org.eclipse.ptp.rm.lml.da/rms.

Best regards,

Carsten

On 01/19/15 09:05, Peter Bryzgalov wrote:
Hi,

I work on adopting Eclipse PTP to “K” computer and its PJM batch system
made by Fujitsu. I have a basic job running and profiling with TAU
features working. Now I'm working on monitoring.
I have very basic understanding of workflow and LML. Where can I find
specifications and examples of what output files should look like on
every step of the workflow? I also need instructions on creating a
layout definition file.

There is a link on wiki.eclipse.org/PTP/designs/scalability
<http://wiki.eclipse.org/PTP/designs/scalability> to LML specification.
Unfortunately http://llview.zam.kfa-juelich.de site is not working.

Kind regards,
Peter Bryzgalov

RIKEN AICS
HPC Usability Research Team
Research Scientist


_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/ptp-user




------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------
Forschungszentrum Juelich GmbH
52425 Juelich
Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher
Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender),
Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
Prof. Dr. Sebastian M. Schmidt
------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------

_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/ptp-user


Back to the top