[Swift-user] Swift runs one task despite slots set to higher number

[Ketan Maheshwari]

Hi Mihael,

Swift is submitting one 16 *node* job with this setup. My intention is to
run 10 such 16 node jobs.

Thanks,
Ketan

On Sun, Feb 1, 2015 at 8:57 PM, Mihael Hategan <hategan at mcs.anl.gov> wrote:

> You are asking for 16 cores through granularity, but explicitly setting
> jobType=single. This results PBS only attempting to start one worker for
> each job. However, the coaster scheduler thinks that 16 workers were
> requested and, since that is more than the number of queued jobs, it
> will not start any other ones.
>
> Mihael
>
> On Sun, 2015-02-01 at 20:09 -0600, Ketan Maheshwari wrote:
> > Hi Mihael,
> >
> > The run is not yet finished. Please find attached the log so far.
> >
> > Thanks,
> > Ketan
> >
> > On Sun, Feb 1, 2015 at 8:04 PM, Mihael Hategan <hategan at mcs.anl.gov>
> wrote:
> >
> > > Hi,
> > >
> > > Can you post the swift log please?
> > >
> > > Mihael
> > >
> > > On Sun, 2015-02-01 at 19:58 -0600, Ketan Maheshwari wrote:
> > > > Hi,
> > > >
> > > > Running a workflow, I need to run 10 PBS tasks each with a size of 16
> > > > nodes.
> > > >
> > > > The workflow starts with 4 single task stages and then fans out to a
> > > > foreach loop with 10 tasks.
> > > >
> > > > Here is the sites file I am using:
> > > >
> > > > <?xml version="1.0" encoding="UTF-8"?>
> > > > <config xmlns="http://www.ci.uchicago.edu/swift/SwiftSites">
> > > > <pool handle="blues">
> > > >   <execution provider="coaster" jobmanager="local:pbs"/>
> > > >   <filesystem provider="local" url="none" />
> > > >   <profile namespace="globus" key="maxtime">13400</profile>
> > > >   <profile namespace="globus" key="maxWalltime">03:41:00</profile>
> > > >   <profile namespace="globus" key="jobsPerNode">1</profile>
> > > >   <profile namespace="globus" key="slots">10</profile>
> > > >   <profile namespace="globus" key="ppn">2</profile>
> > > >   <profile namespace="globus" key="jobtype">single</profile>
> > > >   <profile namespace="globus" key="nodeGranularity">16</profile>
> > > >   <profile namespace="globus" key="maxnodes">16</profile>
> > > >   <profile namespace="karajan" key="jobThrottle">2.20</profile>
> > > >   <profile namespace="karajan" key="initialScore">10000</profile>
> > > > <!--  <profile namespace="globus" key="keepSiteDir">true</profile>-->
> > > >   <workdirectory>/home/ketan/swift.workdir</workdirectory>
> > > > </pool>
> > > >
> > > > <pool handle="localhost">
> > > >   <execution provider="coaster" jobmanager="local:local" />
> > > >   <profile namespace="globus" key="slots">1</profile>
> > > >   <profile namespace="karajan" key="jobThrottle">2.06</profile>
> > > >   <profile namespace="karajan" key="initialScore">10000</profile>
> > > >   <profile namespace="globus" key="jobsPerNode">3</profile>
> > > >   <profile namespace="globus" key="maxtime">9700</profile>
> > > >   <profile namespace="globus" key="maxWalltime">02:39:00</profile>
> > > >   <profile namespace="globus" key="walltime">10060</profile>
> > > >   <profile namespace="globus" key="maxnodes">1</profile>
> > > >   <profile namespace="globus" key="nodegranularity">1</profile>
> > > >  <!-- <profile namespace="globus" key="keepSiteDir">true</profile>-->
> > > >
> > > >   <workdirectory>/home/ketan/swift.workdir</workdirectory>
> > > >   <filesystem provider="local"/>
> > > > </pool>
> > > >
> > > > </config>
> > > >
> > > > However, Swift runs only 1 task throughout the workflow even after
> > > reaching
> > > > the foreach loop.
> > > >
> > > > Any suggestions?
> > > >
> > > > Thanks,
> > > > Ketan
> > > > _______________________________________________
> > > > Swift-user mailing list
> > > > Swift-user at ci.uchicago.edu
> > > > https://lists.ci.uchicago.edu/cgi-bin/mailman/listinfo/swift-user
> > >
> > >
> > > _______________________________________________
> > > Swift-user mailing list
> > > Swift-user at ci.uchicago.edu
> > > https://lists.ci.uchicago.edu/cgi-bin/mailman/listinfo/swift-user
> > >
>
>
> _______________________________________________
> Swift-user mailing list
> Swift-user at ci.uchicago.edu
> https://lists.ci.uchicago.edu/cgi-bin/mailman/listinfo/swift-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/swift-user/attachments/20150201/474fbd3b/attachment.html>