Differences

This shows you the differences between two versions of the page.

--- users:partitions [2016/03/22 15:44] – created ecalore
+++ users:partitions [2018/02/20 12:35] (current) – ecalore
@@ Line 1: / Line 1: @@
 ====== Select a Partition ======
-In SLURM, queues are called partitions. In the COKA cluster at the moment are available the following partitions:
+In SLURM, queues are called partitions. A partition can be selected with the **--partition** or **-p** argument, to know how to do it in your job submission scripts see: [[users:slurm#meaning_of_the_most_common_options|job submission script options]].
-===== Partitions =====
+<WRAP center round tip 60%>
+To query the availability of partitions end nodes on the cluster you can use the following SLURM command:
+<code>
+sinfo -l
+</code>
+</WRAP>
-A partition can be selected with the **--partition** or **-p** argument to srun
+In the COKA cluster at the moment are available the following partitions:
-==== interactive ====
+===== interactive =====
-The Interactive partition is meant for short and/or [[users:interactive|interactive jobs]], such us debugging and compilation of programs.
+The //interactive// partition is meant for **short** and/or [[users:interactive|interactive jobs]], such us debugging and compilation of programs.
 Jobs submitted to this partition:
   * are limited to 30 minutes
-  * multiple jobs can share the same CPU
+  * multiple jobs can share the same CPUs
-  * default memory per requested
+  * default memory per requested task is 1GB
-==== shortrun ====
+===== shortrun =====
-==== longrun ====
+The //shortrun// partition is meant for jobs expected to complete in less than **one hour**.
+Jobs submitted to this partition:
+  * are limited to 60 minutes
+  * multiple jobs can NOT share the same CPUs
+  * default memory per requested task is 1GB
+===== longrun =====
+The //longrun// partition is meant for jobs expected to complete in less than **eight hour**.
+Jobs submitted to this partition:
+  * are limited to 480 minutes
+  * multiple jobs can NOT share the same CPUs
+  * default memory per requested task is 1GB