Kimik: Difference between revisions

← Older edit

Latest revision as of 13:21, 15 September 2010

go back to Main_Page, Computational Resources, Clusters, Local Clusters

Kimik is a cluster (currently) containing 94 nodes.

It is of general use of the Institute (computational groups and some experimentalists too).

The disk is shared with Kimik2

Queues/jobs:

Each user has a quota of slots for jobs. The cluster is divided into three queues as follows...

nanYar.short

         Two single processor machines (nodes 001 and 002).
         Allows all job types, but has a 1 hour time limit.

nanYar

         Older single processor machines from nodes 003 to 045. 
         Works with all job types.
         Currently disabled due to problems vs. the SIRIUS server.

nanYar.l

         Single processor machines from nodes 046 to 094.
         Works with all jobs types.

Quota/file server space

The $HOME of each user has a quota of 23 GB.

There is softquota, that will announce by a message that you are close to exceed the quota limit.

After reaching the maximum, the user will not be able to write in the $HOME until deleting/compressing some files. That also affects the calculations, if there is not enough space to write the output, they will unfortunately die.

Kimik "How to ..."[edit]

How to connect to kimik

From a linux PC:

    ssh -X kimik.iciq.es

from outside ICIQ look at Connecting from Outside

How to submit calculations

     Use the qs command.
     If you type the command “qs” you get a help menu that shows you how to use the 
     command and some specific tips for submission of gaussian jobs.
 
     qs numproc queue typejob nameinput.in
   
     With qs you specify the number of nodes you want, which one of the four available queues to access 
     (short queue (s),long-queue (l)), which program to use (adf, 
     gaussian, etc.), and then the input file name (usually with the *.in extension)

How to view the status of your jobs

     Use the command qstat to give a list of jobs running on the system.
     For more information on the jobs, use the -f key.
     For information on a specific user, use the key -u username.

How to kill jobs

     Use the command qdel and the job identifier.

How to check the quota

You can check the disk space you are using by looking at how to check the disk space used

How to print from kimik

     You can run the next command from command line:
              
                  lpr -PLEXMARK file_to_print

@@ Line 1: / Line 1: @@
 go back to [[Main_Page]], [[Computational Resources]], [[Clusters]], [[Local Clusters]]
-[[Kimik's Quads]]
+Kimik is a cluster (currently) containing 94 nodes.
+It is of general use of the Institute (computational groups and some experimentalists too).
-Kimik is a (by now) 137 nodes's cluster.
+The disk is shared with [[Kimik2]]
 * '''Queues/jobs:'''
-Each user  can use up to '''15''' slots for jobs. There are 4 types of queues in it.
+Each user has a quota of slots for jobs. The cluster is divided into three queues as follows...
 nanYar.short
-           queue with 2 nodes that allows you to run jobs for only 1 hour.
+           Two single processor machines (nodes 001 and 002).
+          Allows all job types, but has a 1 hour time limit.
 nanYar
-           queue from node kimik003.iciq.es to kimik045.iciq.es. For running mostly adf jobs (adf has preferences
+           Older single processor machines from nodes 003 to 045.
-           because licenses are for kimik001 to kimik45 and kimik050) although you can run others one.
+          Works with all job types.
+           Currently disabled due to problems vs. the SIRIUS server.
 nanYar.l
-           queue from node kimik046.iciq.es to kimik094.iciq.es. Send all kind of (installed) jobs, so for
+           Single processor machines from nodes 046 to 094.
-          Gaussian or others please use these queue instead of nanYar (unless there are many free nodes on
+          Works with all jobs types.
-          nanYar queue).
+* '''Quota/file server space'''
-nanYar.quad
+The $HOME of each user has a quota of '''23 GB'''.
-           For "Quad Core 2" nodes.
-           Right now, mainly for running Gaussian jobs, as you cannot get a good perfomance with Jaguar jobs.
-* '''How to submit calculations'''
+There is softquota, that will announce by a message that you are close to exceed the quota limit.
-      Use the qs command. If you type the command “qs” you get a help menu that shows you how to use the
-      command.
-      The name of the job in the queue system is that of the input file starting from the end (up to 10
-      letters).
-      In qs you can choose queues, there are three in the cluster:
-      short queue (s): two machines: “kimik 001” and “kimik 002” in which jobs can last up to one hour, after
-      which the system kills the jobs.
-      normal queue (n): it accepts ADF jobs and it goes from nodes 003 to 045. It has no time restrictions.
-      L-queue (l): goes from nodes 046 to 094, it does not allow ADF calculations and has no time restrictions.
-* '''How to kill jobs'''
+After reaching the maximum, the user will not be able to write in the $HOME until deleting/compressing some files. That also affects the calculations, if there is not enough space to write the output, they will unfortunately die.
-      Use the command qdel and the job identifier.
-* '''How to check the state of a queue'''
+== Kimik "How to ..." ==
-      Use qstat command or go to the webpage: kimik.iciq.es/ganglia
+* '''How to connect to kimik'''
-* '''How to print from kimik'''
+From a linux PC:
-      Laserjet printer: command lpr. The queues are:
-      bn1c: prints on one page (1 side, vertically)
-      bn1c2: prints two pages on one side horizontally (booklet style)
-      bn1ch: prints one page on one side horizontally
-      bn2c1: prints on two pages (both sides) vertically
-      bn2c2: prints two pages on both sides (horizontally, booklet style)
-      bn2ch: prints one page on both sides horizontally
-      Hp printer: command lpr. The queue is called “petita”
-      Hp color printer: command lpr. The queue is called “color”
-* '''How to look at you quota/file server space'''
+     ssh -X kimik.iciq.es
-The $HOME of each user has a quota of '''23 GB'''.
+from outside ICIQ look at [[Connecting from Outside]]
-There is softquota, that will announce by a message that you are close to exceed the quota limit.
+* '''How to submit calculations'''
+      Use the qs command.
+      If you type the command “qs” you get a help menu that shows you how to use the
+      command and some specific tips for submission of gaussian jobs.
+      qs numproc queue typejob nameinput.in
+      With qs you specify the number of nodes you want, which one of the four available queues to access
+      (short queue (s),long-queue (l)), which program to use (adf,
+      gaussian, etc.), and then the input file name (usually with the *.in extension)
-After reaching the maximum, the user will not be able to write in the $HOME until deleting/compressing some files. That also affects the calculations, if there is not enough space to write the output, they will unfortunately die.
+* '''How to view the status of your jobs'''
+      Use the command qstat to give a list of jobs running on the system.
+      For more information on the jobs, use the -f key.
+      For information on a specific user, use the key -u ''username''.
-To look at your quota go to your $HOME (cd /home/username/) and use the keywords:
+* '''How to kill jobs'''
+      Use the command qdel and the job identifier.
- du -s -B MB
-that will give your quota in Megabytes.
+* '''How to check the quota'''
+You can check the disk space you are using by looking at [[how to check the disk space used]]
-There are several ways to compress files, you might want to use: gzip -9 filename
+* '''How to print from kimik'''
+      You can run the next command from command line:
+                   lpr -PLEXMARK file_to_print

Kimik: Difference between revisions

Latest revision as of 13:21, 15 September 2010

Kimik "How to ..."[edit]

Navigation menu

Search