Cluster Usage: Difference between revisions

From DISI
Jump to navigation Jump to search
(adding some information. almost nothing here.)
No edit summary
 
(5 intermediate revisions by 3 users not shown)
Line 2: Line 2:
Note this information is only relevant if you have ssh access to the cluster.
Note this information is only relevant if you have ssh access to the cluster.


*To request a shell on a cluster node, type  
To request a shell on a cluster node, type  
  qrsh  
  qrsh  
Or for a shell on a specific set of nodes (say any from the node-5-* or node-6-* sets):
Or for a shell on a specific set of nodes (say any from the node-5-* or node-6-* sets):
  qrsh -l hostname="node-[56]*"
  qrsh -q all.q -l hostname="node-[56]*" -now no


Back to [[Portal:Lab]]
The -now no parameter is necessary if the cluster is otherwise full and you'll have to wait to get a job.
 
If you're interested in writing your own cluster scripts, the nice way of doing this is to use the following lines:
 
#$ -q all.q
 
The first line uses the correct queue, all.q for basically any job.
 
#$ -t 1-500
 
This line (modify the 500 to mean the max) should be used in concert with $SGE_TASK_ID to write array job scripts, instead of scripts that only run single jobs.
 
Back to [[:Category:Cluster]]




[[Category:Internal]]
[[Category:Internal]]
[[Category:Cluster]]

Latest revision as of 16:51, 14 May 2014

Information on how to use the Shoichet Lab Cluster. Note this information is only relevant if you have ssh access to the cluster.

To request a shell on a cluster node, type

qrsh 

Or for a shell on a specific set of nodes (say any from the node-5-* or node-6-* sets):

qrsh -q all.q -l hostname="node-[56]*" -now no

The -now no parameter is necessary if the cluster is otherwise full and you'll have to wait to get a job.

If you're interested in writing your own cluster scripts, the nice way of doing this is to use the following lines:

#$ -q all.q

The first line uses the correct queue, all.q for basically any job.

#$ -t 1-500

This line (modify the 500 to mean the max) should be used in concert with $SGE_TASK_ID to write array job scripts, instead of scripts that only run single jobs.

Back to Category:Cluster