System - Revision history

GiulioFontana at 10:41, 3 February 2026

2026-02-03T10:41:15Z

← Older revision		Revision as of 10:41, 3 February 2026
Line 224:		Line 224:
	This section provide a brief guide to Mufasa users (especially those who are not experienced in the use of Linux and/or remote servers) about interacting with the system.		This section provide a brief guide to Mufasa users (especially those who are not experienced in the use of Linux and/or remote servers) about interacting with the system.

	== ~~<span style="background:#FFFF00">~~Storage spaces~~</span>~~ ==		== Storage spaces ==

	User jobs require storage of programs and data files. On Mufasa, the space available to users for data storage is the <code>/home/</code> directory. <code>/home/</code> contains two types of directories:		User jobs require storage of programs and data files. On Mufasa, the space available to users for data storage is the <code>/home/</code> directory. <code>/home/</code> contains two types of directories:
Line 252:		Line 252:
	::: '''Important!''' Shared directories are used by several people, so it's important to ''quickly remove from them any file that is not actively in use''.		::: '''Important!''' Shared directories are used by several people, so it's important to ''quickly remove from them any file that is not actively in use''.

	== ~~<span style="background:#FFFF00">~~Disk quotas~~</span>~~ ==		== Disk quotas ==

	In Mufasa, [[#Storage spaces\|Storage spaces]] are subjected to quotas: i.e., the files that are stored in them cannot occupy more than a given amount of disk space. Quotas apply both to personal directories (e.g., <code>/home/userX/</code>) and to shared directories (e.g., <code>/home/shared/ResearchGroupY/</code>).		In Mufasa, [[#Storage spaces\|Storage spaces]] are subjected to quotas: i.e., the files that are stored in them cannot occupy more than a given amount of disk space. Quotas apply both to personal directories (e.g., <code>/home/userX/</code>) and to shared directories (e.g., <code>/home/shared/ResearchGroupY/</code>).
Line 420:		Line 420:
	If you want additional information about file and directory permissions in a Linux system work, [https://www.redhat.com/sysadmin/linux-file-permissions-explained this is a good online guide].		If you want additional information about file and directory permissions in a Linux system work, [https://www.redhat.com/sysadmin/linux-file-permissions-explained this is a good online guide].

	= ~~<span style="background:#ffff00">~~Containers~~</span>~~ =		= Containers =

	[[File:Singularity.png\|right\|262px]]		[[File:Singularity.png\|right\|262px]]

GiulioFontana at 10:39, 3 February 2026

2026-02-03T10:39:59Z

← Older revision		Revision as of 10:39, 3 February 2026
Line 5:		Line 5:
	Remote access to Mufasa is performed using the [[System#Accessing Mufasa\|SSH protocol]] for the execution of commands and the [[System#File transfer\|SFTP protocol]] for the exchange of files. Once logged in, a user interacts with Mufasa via a terminal (text-based) interface.		Remote access to Mufasa is performed using the [[System#Accessing Mufasa\|SSH protocol]] for the execution of commands and the [[System#File transfer\|SFTP protocol]] for the exchange of files. Once logged in, a user interacts with Mufasa via a terminal (text-based) interface.

	= ~~<span style="background:#ffff00">~~Mufasa 2.0~~</span>~~ =		= Mufasa 2.0 =

	At the beginning of November 2025 Mufasa has been subjected to a comprehensive hardware and software overhaul: the new system is sometimes called '''Mufasa 2.0'''; to distinguish it from the old system, the latter is sometimes called "Mufasa 1.0".		At the beginning of November 2025 Mufasa has been subjected to a comprehensive hardware and software overhaul: the new system is sometimes called '''Mufasa 2.0'''; to distinguish it from the old system, the latter is sometimes called "Mufasa 1.0".

	This wiki ~~is currently (November 2025) being~~ updated to reflect the changes. Elements that changed significantly from Mufasa 1.0 to Mufasa 2.0 are <span style="background:#ffff00">highlighted in yellow</span>. When the title of a section is highlighted, it means either that it is a new section, or that the section's contents significantly changed.		This wiki has been updated to reflect the changes.

	= Hardware =		= Hardware =
Line 21:		Line 21:
	* 2 AMD Epyc 7542 32-core processors (64 CPU cores total)		* 2 AMD Epyc 7542 32-core processors (64 CPU cores total)
	* 1 TB RAM		* 1 TB RAM
	* 7 TB of SSDs ~~<span style="background:#ffff00">~~(fast temporary repository for datasets actively being used - see [[#Storage spaces\|Storage spaces]])~~</span>~~		* 7 TB of SSDs (fast temporary repository for datasets actively being used - see [[#Storage spaces\|Storage spaces]])
	* 25 TB of HDDs (for user <code>/home</code> directories)		* 25 TB of HDDs (for user <code>/home</code> directories)
	* ~~<span style="background:#ffff00">~~8 Nvidia A100 GPUs~~</span>~~ [based on the ''Ampere'' architecture]		* 8 Nvidia A100 GPUs [based on the ''Ampere'' architecture]
	* [https://ubuntu.com/ Ubuntu Linux 24.04 LTS server] operating system		* [https://ubuntu.com/ Ubuntu Linux 24.04 LTS server] operating system

Line 36:		Line 36:
	All in all, the GPU complement of Mufasa comprises the following devices:		All in all, the GPU complement of Mufasa comprises the following devices:

	:; ~~<span style="background:#FFFF00">~~5 GPUs with 20GB of RAM and 3 compute units~~</span>~~		:; 5 GPUs with 20GB of RAM and 3 compute units
	:; ~~<span style="background:#FFFF00">~~5 GPUs with 20GB of RAM and 4 compute units~~</span>~~		:; 5 GPUs with 20GB of RAM and 4 compute units
	:; ~~<span style="background:#FFFF00">~~3 GPUs with 40 GB of RAM~~</span>~~		:; 3 GPUs with 40 GB of RAM

	Thanks to MIG, users can use all the GPUs listed above as if they were all physical devices installed on Mufasa, without having to worry (or even know) which actually are and which instead are virtual GPUs. How these devices are made available to Mufasa users is explained in [[SLURM]].		Thanks to MIG, users can use all the GPUs listed above as if they were all physical devices installed on Mufasa, without having to worry (or even know) which actually are and which instead are virtual GPUs. How these devices are made available to Mufasa users is explained in [[SLURM]].
Line 52:		Line 52:
	= Accessing Mufasa =		= Accessing Mufasa =

	== ~~<span style="background:#ffff00">~~Login server~~</span>~~ ==		== Login server ==

	Differently from Mufasa 1.0, Mufasa 2.0 employs a ''login server'' to manage user connections. The login server is a Linux virtual machine with very limited resources (no GPUs, few CPUs, small RAM). Its task is only to provide users with a way to log into the system and launch [[User Jobs]] with SLURM. Jobs launched via SLURM run on Mufasa 2.0's physical hardware (not on the virtual hardware of the login server) and therefore can access to the hardware resources of Mufasa, such as the GPUs.		Differently from Mufasa 1.0, Mufasa 2.0 employs a ''login server'' to manage user connections. The login server is a Linux virtual machine with very limited resources (no GPUs, few CPUs, small RAM). Its task is only to provide users with a way to log into the system and launch [[User Jobs]] with SLURM. Jobs launched via SLURM run on Mufasa 2.0's physical hardware (not on the virtual hardware of the login server) and therefore can access to the hardware resources of Mufasa, such as the GPUs.
Line 60:		Line 60:
	Please note that if you try to run computationally heavy processes in the login server you can easily overwhelm its scarce resources, making it unavailable to all users and thus making Mufasa unreachable by anyone. The login server has safety mechanisms to prevent processes from hogging too much of its resources... by killing such processes. However, please avoid checking if these mechanisms work.		Please note that if you try to run computationally heavy processes in the login server you can easily overwhelm its scarce resources, making it unavailable to all users and thus making Mufasa unreachable by anyone. The login server has safety mechanisms to prevent processes from hogging too much of its resources... by killing such processes. However, please avoid checking if these mechanisms work.

	== ~~<span style="background:#ffff00">~~Logging into the login server~~</span>~~ ==		== Logging into the login server ==

	User access to Mufasa is always remote and exploits the ''SSH'' (''Secure SHell'') protocol.		User access to Mufasa is always remote and exploits the ''SSH'' (''Secure SHell'') protocol.
Line 100:		Line 100:
	from the command prompt of the remote shell.		from the command prompt of the remote shell.

	== ~~<span style="background:#ffff00">~~Direct access to Mufasa (for users with running SLURM jobs)~~</span>~~ ==		== Direct access to Mufasa (for users with running SLURM jobs) ==

	While a user has SLURM jobs in execution, such user may also log into Mufasa directly, without passing through the login server. This allows the user to interact with the running jobs, e.g. to monitor their progress. Direct access to Mufasa is done via SSH with command		While a user has SLURM jobs in execution, such user may also log into Mufasa directly, without passing through the login server. This allows the user to interact with the running jobs, e.g. to monitor their progress. Direct access to Mufasa is done via SSH with command

GiulioFontana: /* Disk quotas */

2026-02-03T10:17:51Z

Disk quotas

← Older revision		Revision as of 10:17, 3 February 2026
Line 262:		Line 262:
	</pre>		</pre>

	When <code>df</code> is run from the is [[#Logging into the login server\|login server]], its output is similar to the following:		When <code>df</code> is run from the is [[#Logging into the login server\|login server]], its output is similar to the following, respectively for a home directory and a shared space:

	<pre style="color: lightgrey; background: black;">		<pre style="color: lightgrey; background: black;">
Line 268:		Line 268:
	Filesystem Size Used Avail Use% Mounted on		Filesystem Size Used Avail Use% Mounted on
	192.168.1.1:/home 200G 161G 40G 81% /home		192.168.1.1:/home 200G 161G 40G 81% /home
			</pre>

			<pre style="color: lightgrey; background: black;">
			gfontana@mufasa2-login:~$ df -h /home/shared/nearmrs/
			Filesystem Size Used Avail Use% Mounted on
			192.168.1.1:/mnt/shared 1.3T 1008G 293G 78% /mnt/shared
	</pre>		</pre>

GiulioFontana: /* The SLURM job scheduling system */

2025-11-27T08:28:32Z

The SLURM job scheduling system

← Older revision		Revision as of 08:28, 27 November 2025
Line 462:		Line 462:
	Jobs run via SLURM have access to all the resources of Mufasa. Jobs run outside SLURM are executed by the [[#Login server\|login server]] virtual machine, which has minimal resources and no GPUs. Using SLURM is therefore the only way to execute resource-heavy jobs on Mufasa. This is a key difference between Mufasa 1.0 and [[#Mufasa 2.0\|Mufasa 2.0]].		Jobs run via SLURM have access to all the resources of Mufasa. Jobs run outside SLURM are executed by the [[#Login server\|login server]] virtual machine, which has minimal resources and no GPUs. Using SLURM is therefore the only way to execute resource-heavy jobs on Mufasa. This is a key difference between Mufasa 1.0 and [[#Mufasa 2.0\|Mufasa 2.0]].

	'''IMPORTANT!''' The priority (and thus the wait before execution) of ~~computing jobs~~ run via SLURM depends on ~~their~~ '''priority'''. [[SLURM#How_to_maximise_the_priority_of_your_jobs\|Learn how to maximise the priority of your jobs]].		'''IMPORTANT!''' The priority (and thus the wait before execution) of any job run via SLURM depends on its '''priority'''. [[SLURM#How_to_maximise_the_priority_of_your_jobs\|Learn how to maximise the priority of your jobs]].

	SLURM is capable of managing complex computing systems composed of multiple ''clusters'' (i.e. sets) of servers, each comprising one ''node'' (i.e. machine) or more. The case of Mufasa is the simplest of all: Mufasa is in fact the one node of a SLURM computing cluster composed of a single machine.		SLURM is capable of managing complex computing systems composed of multiple ''clusters'' (i.e. sets) of servers, each comprising one ''node'' (i.e. machine) or more. The case of Mufasa is the simplest of all: Mufasa is in fact the one node of a SLURM computing cluster composed of a single machine.

GiulioFontana: /* The SLURM job scheduling system */

2025-11-27T08:28:06Z

The SLURM job scheduling system

← Older revision		Revision as of 08:28, 27 November 2025
Line 461:		Line 461:

	Jobs run via SLURM have access to all the resources of Mufasa. Jobs run outside SLURM are executed by the [[#Login server\|login server]] virtual machine, which has minimal resources and no GPUs. Using SLURM is therefore the only way to execute resource-heavy jobs on Mufasa. This is a key difference between Mufasa 1.0 and [[#Mufasa 2.0\|Mufasa 2.0]].		Jobs run via SLURM have access to all the resources of Mufasa. Jobs run outside SLURM are executed by the [[#Login server\|login server]] virtual machine, which has minimal resources and no GPUs. Using SLURM is therefore the only way to execute resource-heavy jobs on Mufasa. This is a key difference between Mufasa 1.0 and [[#Mufasa 2.0\|Mufasa 2.0]].

			'''IMPORTANT!''' The priority (and thus the wait before execution) of computing jobs run via SLURM depends on their '''priority'''. [[SLURM#How_to_maximise_the_priority_of_your_jobs\|Learn how to maximise the priority of your jobs]].

	SLURM is capable of managing complex computing systems composed of multiple ''clusters'' (i.e. sets) of servers, each comprising one ''node'' (i.e. machine) or more. The case of Mufasa is the simplest of all: Mufasa is in fact the one node of a SLURM computing cluster composed of a single machine.		SLURM is capable of managing complex computing systems composed of multiple ''clusters'' (i.e. sets) of servers, each comprising one ''node'' (i.e. machine) or more. The case of Mufasa is the simplest of all: Mufasa is in fact the one node of a SLURM computing cluster composed of a single machine.

GiulioFontana: /* Direct access to Mufasa (for users with running SLURM jobs) */

2025-11-25T12:27:52Z

Direct access to Mufasa (for users with running SLURM jobs)

← Older revision		Revision as of 12:27, 25 November 2025
Line 108:		Line 108:
	</pre>		</pre>

	where both username and password are the same used for the [[#Login server\|login server]].		where both <code>username</code> and password are the same used for the [[#Login server\|login server]].
	For example, user <code>mrossi</code> may access Mufasa with command		For example, user <code>mrossi</code> may access Mufasa with command

GiulioFontana: /* Direct access to Mufasa (for users with running SLURM jobs) */

2025-11-25T12:26:09Z

Direct access to Mufasa (for users with running SLURM jobs)

← Older revision		Revision as of 12:26, 25 November 2025
Line 108:		Line 108:
	</pre>		</pre>

	where both username and password are the same used for the [[Login server\|login server]].		where both username and password are the same used for the [[#Login server\|login server]].
	For example, user <code>mrossi</code> may access Mufasa with command		For example, user <code>mrossi</code> may access Mufasa with command

GiulioFontana: /* Direct access to Mufasa (for users with running SLURM jobs) */

2025-11-25T12:25:32Z

Direct access to Mufasa (for users with running SLURM jobs)

← Older revision		Revision as of 12:25, 25 November 2025
Line 108:		Line 108:
	</pre>		</pre>

			where both username and password are the same used for the [[Login server\|login server]].
	For example, user <code>mrossi</code> may access Mufasa with command		For example, user <code>mrossi</code> may access Mufasa with command

GiulioFontana: /* Disk quotas */

2025-11-25T12:23:52Z

Disk quotas

← Older revision		Revision as of 12:23, 25 November 2025
Line 266:		Line 266:
	gfontana@mufasa2-login:~$ df -h /home/gfontana/		gfontana@mufasa2-login:~$ df -h /home/gfontana/
	Filesystem Size Used Avail Use% Mounted on		Filesystem Size Used Avail Use% Mounted on
	192.168.1.1:/home 200G 161G 40G 81% /~~hom~~		192.168.1.1:/home 200G 161G 40G 81% /home
	</pre>		</pre>

GiulioFontana: /* Disk quotas */

2025-11-25T12:23:27Z

Disk quotas

← Older revision		Revision as of 12:23, 25 November 2025
Line 264:		Line 264:

	<pre style="color: lightgrey; background: black;">		<pre style="color: lightgrey; background: black;">
	gfontana@mufasa2-login:~~/home/shared~~$ df -h /home/gfontana/		gfontana@mufasa2-login:~$ df -h /home/gfontana/
	Filesystem Size Used Avail Use% Mounted on		Filesystem Size Used Avail Use% Mounted on
	192.168.1.1:/home ~~150G~~ ~~11G~~ ~~139G 7~~% /~~home~~		192.168.1.1:/home 200G 161G 40G 81% /hom
	</pre>		</pre>