airavata-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gary E. Gorbet" <gegor...@gmail.com>
Subject Fwd: cluster status function in Airavata
Date Thu, 08 Oct 2015 11:51:27 GMT


> Begin forwarded message:
> 
> From: "Gary E. Gorbet" <gegorbet@gmail.com>
> Subject: cluster status function in Airavata
> Date: October 7, 2015 at 5:30:22 PM CDT
> To: SciGaP Dev <dev@scigap.org>
> 
> The two text files attached illustrate a Gateway function that is currently performed
by direct ssh commands. I would like to see that functionality moved to Airavata for two main
reasons:
> 
> (1) Any refinements for changes on clusters or for new clusters would be centralized;
and
> 
> (2) Gateway servers on multiple hosts would referenced common APIs.
> 
> The first attached file - cluster_status.php-local_status.txt - is the code portion from
a local_status() function within the UltraScan gateway script cluster_status.php. This code
portion shows the ssh commands issued for each of a list of clusters. The parsed command output
reveals the count of running jobs and queued jobs for each cluster. That information is used
on the gateway submit page to hint at likely wait-in-queue-status time. My proposal is that
a Thrift client API would return this information, using code on the Thrift server similar
to that in the sample PHP script.
> 
> The second attached file shows a bash shell script executed remotely on the Jureca cluster
at the PRACE Juelich center in Germany. This special script was found to be necessary because
none of the normal queue status commands ( sinfo, qstat, ... ) were found to return reliable
information. The script shown basically just issues squeue commands and counts the lines returned
as a way of counting PENDING and RUNNING jobs in the “batch” queue on Jureca.
> 
> The attached image file shows how the job count information is used at the submit stage
of the UltraScan gateway.
> 
> - Gary
> 
> 
> 
> 
> 
> 

Mime
View raw message