Using Slurm - Revision history

Haars0011 at 14:57, 19 May 2025

2025-05-19T14:57:17Z

← Older revision		Revision as of 16:57, 19 May 2025
Line 137:		Line 137:

	=== Using GPU ===		=== Using GPU ===
	There are six GPU nodes, in order to run a job that uses GPU on one of these nodes, you can add
			==== Nvidia GPUs ====
			There are six GPU nodes with Nvidia GPUs, in order to run a job that uses a GPU on one of these nodes, you can add

	<pre>		<pre>
Line 167:		Line 169:
	We have set up the scheduler in such a way that the A100s are chosen first, and then the A6000s, and lastly the V100s		We have set up the scheduler in such a way that the A100s are chosen first, and then the A6000s, and lastly the V100s
	The pricing for all is the same.		The pricing for all is the same.

			==== AMD GPUs ====
			Besides the Nvidia GPUs, the cluster also has 2 nodes with AMD GPUs.

			The way to use them is the same as for the Nvidia GPUs, just use the '''gpu_amd''' partition to select them.

	== Monitoring submitted jobs ==		== Monitoring submitted jobs ==

2025-05-19T14:49:24Z

← Older revision		Revision as of 16:49, 19 May 2025
Line 167:		Line 167:
	We have set up the scheduler in such a way that the A100s are chosen first, and then the A6000s, and lastly the V100s		We have set up the scheduler in such a way that the A100s are chosen first, and then the A6000s, and lastly the V100s
	The pricing for all is the same.		The pricing for all is the same.

	~~Please use the nvidia constraint if your jobs are limited to those, as we will put the AMD GPUs online in the future, which will then probably break your analyses.~~


	== Monitoring submitted jobs ==		== Monitoring submitted jobs ==

2024-05-22T11:21:19Z

Defaults

← Older revision		Revision as of 13:21, 22 May 2024
Line 34:		Line 34:

	The default run time for a job is '''1 hour'''.		The default run time for a job is '''1 hour'''.

			The maximum run time for a job is '''3 weeks'''.

	The default memory limit is '''100MB per node'''.		The default memory limit is '''100MB per node'''.


	== Submitting jobs: sbatch ==		== Submitting jobs: sbatch ==

2024-04-16T08:55:21Z

← Older revision		Revision as of 10:55, 16 April 2024
Line 155:		Line 155:
	To select a certain type, use the flag:		To select a certain type, use the flag:
	<pre>		<pre>
	--constraint		#SBATCH --constraint
	</pre>		</pre>
	Example:		Example:
	<pre>		<pre>
	--constraint nvidia,A100		# This will limit this job to the A100 GPUs
			#SBATCH --constraint='nvidia&A100'
	</pre>		</pre>

2024-04-16T08:06:40Z

← Older revision		Revision as of 10:06, 16 April 2024
Line 156:		Line 156:
	<pre>		<pre>
	--constraint		--constraint
			</pre>
			Example:
			<pre>
			--constraint nvidia,A100
	</pre>		</pre>

	A rough estimate is that the A100/80G cards are about twice as fast as the A6000/48G or the V100/16G.		A rough estimate is that the A100/80G cards are about twice as fast as the A6000/48G or the V100/16G. But this all depends on whether your analyses actually needs the RAM and can completely fill the GPU.
	We have set up the scheduler in such a way that the A100s are chosen first, and then the A6000s, and lastly the V100s		We have set up the scheduler in such a way that the A100s are chosen first, and then the A6000s, and lastly the V100s
	The pricing for all is the same.		The pricing for all is the same.

2024-04-16T07:59:51Z

@@ Line 136: / Line 136: @@
 === Using GPU ===
-There are four GPU nodes, in order to run a job that uses GPU on one of these nodes, you can add
+There are six GPU nodes, in order to run a job that uses GPU on one of these nodes, you can add
 <pre>
@@ Line 143: / Line 143: @@
 </pre>
 To your sbatch script. Without this parameter, your job won't run on one of these nodes.
 == Monitoring submitted jobs ==

2023-09-15T15:57:13Z

two to four

← Older revision		Revision as of 17:57, 15 September 2023
Line 136:		Line 136:

	=== Using GPU ===		=== Using GPU ===
	There are ~~two~~ GPU nodes, in order to run a job that uses GPU on one of these nodes, you can add		There are four GPU nodes, in order to run a job that uses GPU on one of these nodes, you can add

	<pre>		<pre>

2023-09-15T15:56:42Z

Removed GPU flavour, as we only have A100s right now.

← Older revision		Revision as of 17:56, 15 September 2023
Line 140:		Line 140:
	<pre>		<pre>
	#SBATCH --gres=gpu:<num gpus>		#SBATCH --gres=gpu:<num gpus>
	~~#SBATCH --constraint=<gpu flavour e.g. K80, V100>~~
	#SBATCH --partition=gpu		#SBATCH --partition=gpu
	</pre>		</pre>

Prins0891: to

 to fix code blocks

2023-06-15T14:49:26Z

<source> to <pre> to fix code blocks

← Older revision		Revision as of 16:49, 15 June 2023
Line 42:		Line 42:
	=== Example ===		=== Example ===
	Consider this simple python3 script that should calculate Pi to 1 million digits:		Consider this simple python3 script that should calculate Pi to 1 million digits:
	<~~source lang='python'~~>		<pre>
	from decimal import *		from decimal import *
	D=Decimal		D=Decimal

Prins0891: to

 to fix code blocks

2023-06-15T13:37:25Z

<source> to <pre> to fix code blocks