[torqueusers] poll_jobs server parameter setting behaviour

Garrick Staples garrick at usc.edu
Thu May 29 11:42:19 MDT 2008


On Thu, May 29, 2008 at 03:30:17AM -0700, Virendra Pratap. Singh alleged:
> 
> Hi All,
> 
>   I want to understand the behaviour of "poll_jobs" torque server
> parameter. The description of this from the documentation
> http://www.clusterresources.com/torquedocs21/a.bserverparameters.shtml
> states:
> 
> poll_jobs	<BOOLEAN> 	TRUE (FALSE in TORQUE 1.2.0p5 and
> earlier)	if set to TRUE, pbs_server will poll job info from mom's
> over time and will not block on handling requests which require this job
> information.  If not set, no polling will occur and if job information
> is requested which is stale, pbs_server may block while it attempts to
> update this information.  NOTE:  For large systems, this value should be
> set to TRUE.
> 
> So this is the situation in which I want to understand its behavior. I
> have this parameter set to TRUE. I do a "qstat" to enquire about my
> submitted job. I believe if the server has a NON stale data for this job
> with itself it will return that. Incase the job data is stale it will
> make a non blocking call to the respective processing nodes mom for the
> fresh info on this job. Since the call is non blocking, due to any
> reason the mom/server communication doesn't goes through, then WHAT will
> be the info returned by the server for this qstat request.
>     Will the server return the STALE info in this case or what?

If not set, pbs_server doesn't request job info from MOMs until someone does a
qstat.  Then pbs_server sends *blocking* job status requests to each MOM if the
in-memory information is not current.  This is OK on smaller clusters.

If set, pbs_server regularly polls for job status info on a set time interval
When someone does a qstat, there is no need to block because the in-memory
information is always relatively current.

-- 
Garrick Staples, GNU/Linux HPCC SysAdmin
University of Southern California

Please avoid sending me Word or PowerPoint attachments.
See http://www.gnu.org/philosophy/no-word-attachments.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://www.supercluster.org/pipermail/torqueusers/attachments/20080529/94c98494/attachment.bin


More information about the torqueusers mailing list