[DRMAA-WG] Pending work on DRMAAv2 job monitoring
Peter Tröger
peter at troeger.eu
Tue Jun 2 09:49:04 CDT 2009
Dear all,
please remember to work actively on the set of new monitoring attributes for
DRMAAv2. The table is still open for everybody:
http://spreadsheets.google.com/ccc?key=rrAIK9utkSoDQXF8kasYCLQ
I recently added the job monitoring possibilities from the viewpoint of Condor.
Best regards,
Peter.
Daniel Templeton schrieb:
> 26 May 2009 Meeting Minutes
>
> ---------------------------
>
> o Minutes from previous meeting accepted
> o Results of research into JSDL requestable resources in Google Docs
> - http://spreadsheets.google.com/ccc?key=rrAIK9utkSoDQXF8kasYCLQ
> - The set of monitorable attributes is not the same as the set of the
> requestable attributes
> - Easiest way to handle monitoring resources would be like the job
> template
> . Some required, some optional, openly extensible
> - Should be able to monitor jobs that aren't in your session, just
> like drmaa_job_ps()
> . Means that we need to be able to report attributes we'd normally
> expect to have known fixed values
> . Also need to consider that job attributes can be altered after
> submission, by user or admin
> - Two ways to pick resources: find the intersection of all
> DRM-supported attributes or find out if all support the ones we want
> . Easier to check if all support the ones we think are important
> . host name - the name of the host on which the job (or master task)
> is running
> . slave names - the names of the hosts on which the job's slave
> tasks are running
> . anything that can be set in the job template
> . submission host - the name of the host from which the job was
> submitted
> . job owner - the name of the user who submitted the job
> . wallclock time - amount of time since the job was started
> . CPU time - amount of CPU seconds consumed
> . submission time, dispatch time, start time, finish time
> . wallclock limit - amount of time remaining until the job will be
> terminated
> . memory usage and limits
> ~ What to do about parallel jobs? Want to know what every slave is
> doing; maybe a separate attribute
> . data? Not reported by all DRMs
> . host information - OS, architecture, etc. - maybe let the user
> figure it out from the host name
> ~ Expand the API to allow for querying host information
> o Need to complete list of attribute proposals and then check if it's
> mappable to all DRMs
> o DRMAA face-to-face at ISC will be June 23rd and 24th in Hamburg
>
> Daniel
> --
> drmaa-wg mailing list
> drmaa-wg at ogf.org
> http://www.ogf.org/mailman/listinfo/drmaa-wg
More information about the drmaa-wg
mailing list