Linux Cheat Sheet¶

Why Learn Bash¶

Bash is a powerful command line shell that allows you to interact with our systems efficiently. It enables scripting and automation, job submission, file and directory management, remote access, resource monitoring, environment customization, parallel computing, data preprocessing, version control, error handling, software dependency management, custom workflows, and offers a valuable skill set applicable beyond HPC tasks. In other words, it allows you to do a lot!

In the cheat sheet below, we offer some basics as a reference. We also highly suggest you explore some more comprehensive guides such as the following: Introduction to Linux on HPC

Shell Basics¶

Shortcuts¶

Shortcuts can be used as stand-ins for system locations, avoiding the need for full paths. A few examples:

Shortcut	Description
`~`	Your home directory
`.`	Your working directory
`..`	One directory above your working directory

Commands¶

`ls`	List the contents of a directory
`ls -la`	List the contents of a directory including hidden files
`cd`	Change your working directory
`pwd`	Show the path to your working directory
`mkdir`	Create a new directory
`rm`	Delete a file. Be careful, files deleted this way can't be retrieved
`rm -r`	Delete a directory. Be careful, directories deleted this way can't be retrieved

Hidden Files and Directories¶

Hidden files and directories start with a dot . and won't show up when you do a simple ls. Some of these files are used to configure your environment when you first log in. Be careful when removing or modifying these files since doing so can cause unintended consequences.

Tip

The ~ in the filenames below indicates your home directory. For example, ~/.bashrc is specifying a file in your home called .bashrc.

File/Directory	What It Does	Warnings
`~/.bash_profile`	This file sets your working environment when you first log into HPC. This file sources your `~/.bashrc` (see below).	See the list in the danger block below.
`~/.bashrc`	This file sets your working environment when you first log into HPC.	See the list in the danger block below.
`~/.local`	This is a hidden directory in your home where pip-installed python packages, jupyter kernels, RStudio session information, etc. goes.	If you pip-install python packages when a virtual environment is not active, they will be installed in this directory. These will then be automatically loaded for all future python sessions (version-specific), including in Singularity images. This may cause versioning issues. We recommend always using virtual environments.
`~/.apptainer`	A hidden directory in your home where Apptainer images and cache files are stored by default.	This directory can grow large quickly and fill up your home. You can modify your `~/.bashrc` to set a different cache directory location that has a larger quota.

Danger

When working with hidden configuration files in your account (.bashrc and .bash_profile), be careful of:

Aliasing important Linux commands

For example, the character . is a shortcut for source. If you do something like add alias .=<command> to your bashrc, you will lose basic functionality in the terminal, e.g., access to modules, virtual environments, etc.
Recursively sourcing configuration files

If you add source ~/.bashrc or source ~/.bash_profile to your bashrc, then you will enter an infinite sourcing loop. This means when you try to log in, your terminal will freeze, then your access will be denied.
Using echo

If you use CLI tools for data transfer, e.g. scp or sftp, they may require a "silent" terminal. If you're trying to initiate transfers and are getting the error "Received message too long", check your bashrc to make sure you aren't printing anything to the terminal.
Removing your files

If you delete either your ~/.bashrc or ~/.bash_profile, you will lose access to module commands.

Environment Variables¶

Bash variables control how your environment is configured. For example: what executables, libraries, header files, etc. are findable by default; information about your Slurm job; MPI settings; GPU configuration; etc. To see all the environment variables that are defined for your session, try running the command env.

An example of an important environment variable is PATH. This is set to a :-delimited list of paths that it uses to search for executables. You can see what it's set to by running echo $PATH. Any time you run a command, for example ls, that list is searched in order. To add new executables to your environment, you can add a path to PATH. For example:

export PATH=/path/to/new/directory:$PATH

This is what modules do: they update your environment variables to put new software and libraries into your environment for use. Some examples of some important variables are listed below:

Variable	Function
`PATH`	A colon-delimited list of paths used to search for executables. When you run a command, such as `ls`, the directories listed in `PATH` are searched in order.
`LD_LIBRARY_PATH`	A colon-delimited list of directories in which the linker should search for shared libraries at runtime. Useful for specifying additional library directories for dynamically linked programs. For example: `export LD_LIBRARY_PATH=/path/to/library:$LD_LIBRARY_PATH`
`CFLAGS`, `CXXFLAGS`, `LDFLAGS`	Environment variables used to specify compiler flags for compiling C, C++, and linking respectively. These variables are often used to customize the build process of software. For example: `export CFLAGS="-O2 -march=native $CFLAGS"`
`CC`, `CXX`, `FC`	Environment variables specifying the C, C++, and Fortran compilers, respectively. These variables allow users to specify which compiler should be used during the build process of software. For example: `export CC=gcc`
`OMP_NUM_THREADS`	Specifies the number of OpenMP threads to use for parallelized programs. This variable controls the number of threads used in parallel regions of OpenMP-enabled programs. For example: `export OMP_NUM_THREADS=4`

Slurm also sets its own environment variables.

Linux File Permissions¶

Linux file permissions control who can access files and directories and what they are able to do with them.

HPC users may find it useful or necessary to share files and/or directories with collaborators. To do this effectively, it may become necessary to familiarize yourself with the linux file permission system in order to give collaborators access to your files.

Types of Permissions¶

All files and directories on a Linux system have permissions defined for them. There are three types:

Permission	Symbol	Directory meaning	File meaning
Read	`r`	Users can view the contents of the directory	Users can read the contents of a file
Write	`w`	Users can modify the contents of a directory	Users can write to a file
Execute	`x`	Users can `cd` into a directory	Users can directly execute a file

When a permission is missing, you will see a -

Viewing Permissions¶

To see a file or directory's file permissions, run the command ls -l. This will show you output that looks like the following:

-rwxr-x--- 1 username groupname 0 Feb 27 09:54 file

String	Access Group	Meaning
`-`	Tells you whether this item is a regular file (`-`), directory (`d`), or link `(l)`	In this example, we're looking at a regular file
`rwx`	This describes the permissions that are set at the user level. In this case, they apply to the user with the username `username`. Your username is your NetID	In this example, the file's owner can read, modify, and execute this file.
`r-x`	This describes the permissions that are set at the group level. In this case, they apply to anyone who is a member of the group `groupname`. To see your groups you're a member of, run the command `groups`.	In this example, group members are allowed to see and execute the contents of this file, but they cannot modify it.
`---`	This describes the permissions that are set for anyone else on the system.	In this example, the rest of the HPC community can't see or edit the contents of this file and can't execute it

Changing Permissions¶

Permissions changes limitations

Only the owner of a file or directory can change its permissions.

To change the permissions of a file or directory, you can use the command chmod. This command accepts two types of input: symbolic and octal.

Symbolic¶

Symbolic notation is the most intuitive to understand. It involves a comma-delimited list of permissions symbols, each representing a specific permission type (read, write, execute) for a particular user or group. Here's a breakdown of the symbols used:

Symbol	Meaning
`u`	"User". Refers to the user who owns the file.
`g`	"Group". Refers to the group that the file belongs to.
`o`	"Other". Refers to other users who are not the owner or in the group.
`a`	"All". Refers to all users (`u`, `g`, and `o`).

For each of these, you can use + to add a permission, - to remove a permission, or = to set the permissions explicitly. The permission symbols are: r,w, and x and match those described under Types of Permissions above.

The general syntax is

chmod [who][operator][permissions]

Here are some examples of how you might use symbolic notation with the chmod command:

To give the owner read and write permissions: chmod u+rw file.txt
To remove execute permissions for the group: chmod g-x file.txt
To give all users read and execute permissions without write permissions: chmod a=rx file.txt

Octal¶

Octal notation is a more compact way of representing permissions using numbers. Each permission type (read, write, execute) is assigned a numeric value:

Octal notation	Permission
`4`	read
`2`	write
`1`	execute

To set permissions, you add these values together for each permission type. For example:

Read and write permissions: 4 (read) + 2 (write) = 6
Execute permission only: 1 (execute)

You then represent the permissions for the user, group, and others as a three-digit number. For instance:

600 gives read and write permissions to the owner and no permissions to the group and others.
755 gives read, write, and execute permissions to the owner, and read and execute permissions to the group and others.

To use octal notation with the chmod command, you specify the permissions directly as numbers. For example:

chmod 600 file.txt
chmod 755 file.txt

These commands will set the permissions of file.txt accordingly.

Changing Group Ownership¶

Ownership changes limitations

Only the owner of a file or directory can change its group ownership

To change the group ownership of a file or directory, you can use chgrp. For example:

chgrp groupname /path/to/file

Changing User Ownership¶

Changing the user ownership of a file requires root permissions, even if the user owns the file they are trying to modify. If ownership modifications are needed, contact our consulting team for assistance.

File Management in Shared Spaces¶

ACLs unavailable

Using something like setfacl is possible on some systems to set the default permissions for specific directories and their contents. Unfortunately, ACLs are not available on our systems due to software limitations on our file servers. Some alternatives are detailed below.

If you're working in shared storage and want your files to be accessible to collaborators, sometimes it can be challenging to ensure those files are created with the correct group ownership and file permissions to allow access. In those cases, it might be helpful to look into setting the SGID (Set Group ID) and using the umask command described below.

Default Group Ownership¶

The SGID (Set Group ID) bit can be set on a directory to ensure that files and directories created within that parent directory inherit its group ownership. This ensures that those files can be made accessible (with the proper permissions) to other members of that group.

To set the SGID bit on a directory, use the chmod command with the g+s option:

chmod g+s /path/to/directory

You'll notice that when a directory does not have the SGID bit set, its group permissions will look something like rwx. When the SGID bit is set, the x will become either an s (if group execute permissions are set) or S (if group execute permissions are disabled). For example:

(puma) [netid@wentletrap EXAMPLES]$ ls -ld parent_directory/
drwxr-xr-x. 2 netid hpcteam 0 Sep  4 12:50 parent_directory/
(puma) [netid@wentletrap EXAMPLES]$ chmod g+s parent_directory/
(puma) [netid@wentletrap EXAMPLES]$ ls -ld parent_directory/
drwxr-sr-x. 2 netid hpcteam 0 Sep  4 12:50 parent_directory/

Default Permissions¶

The umask command controls the default file permissions for all new files and directories created at a user level. It defines which permissions should not be set by default, and thus indirectly determines the default permissions.

By setting an appropriate umask, you can ensure that files and directories are created with the permissions that are appropriate for your collaborative environment.

To check your active umask value, you can run the command without arguments. For example, the default on our system is:

(puma) [netid@wentletrap EXAMPLES]$ umask
0022

To make sense of this output output, take the last three digits and subtract them from either 777 (for directories) or 666 (for files). So for example, in the output shown above, 0022 means new directories will always be created with the permissions 777 - 022 = 755 which is the same as rwxr-xr-x. Similarly, for files: 666 - 022 = 644 which is the same as r-wr--r--. If you're unfamiliar with octal formatting, see octal file permissions above for more information on how these numbers correspond to file permissions.

To set a new umask, you can include arguments either in octal or symbolic format. Symbolic notation is a little more straightforward and intuitive than octal and uses the syntax

umask u=<user permissions>,g=<group permissions>,o=<other permissions>

For example:

(puma) [netid@junonia umask]$ touch test.txt && mkdir test_dir
(puma) [netid@junonia umask]$ ls -l test.txt && ls -ld test_dir/
-rw-r--r--. 1 netid hpcteam 0 Mar 30 10:24 test.txt
drwxr-sr-x. 2 netid hpcteam 0 Mar 30 10:24 test_dir/
(puma) [netid@junonia umask]$ umask u=rwx,g=rwx,o=
(puma) [netid@junonia umask]$ rm -r test.txt test_dir/
(puma) [netid@junonia umask]$ touch test.txt && mkdir test_dir
(puma) [netid@junonia umask]$ ls -l test.txt && ls -ld test_dir/
-rw-rw----. 1 netid hpcteam 0 Mar 30 10:24 test.txt
drwxrws---. 2 netid hpcteam 0 Mar 30 10:24 test_dir/

Note how once umask is set by the user, the default permissions applied to the new test file and directory allow group edit access and remove "other" access completely. The octal notation to accomplish the same result would be umask 007.

Some things to keep in mind:

umask only affects your active terminal session and does not propagate to future sessions. This means if you log out and log back in, your umask will be reset to the system default. If you'd like your default file permissions to be permanently changed, you can add your umask command to your ~/.bashrc. For more information on this file, see hidden files and directories above.
umask applies to all new files and directories you create, so you'll want to make sure you are not inadvertently giving unwanted access to your data.

Sticky Bits¶

In shared environments with group write permissions, it's possible for users to accidentally or intentionally delete files or directories created by others. The Sticky Bit is a security feature that can help manage this issue by restricting file deletion within a directory.

The Sticky Bit, when set on a directory, ensures that only the file's or directory's owner can delete or rename them. This can be particularly useful in shared directories where multiple users need to collaborate but should not interfere with each other's files. To set the Sticky Bit on a directory, use the chmod command with the +t option:

chmod +t /path/to/directory

The change will be displayed at the end of the permissions string either as a t (if execute permissions are set for "other") or a T (if execute permissions are disabled for "other"). For example:

(puma) [netid1@wentletrap EXAMPLES]$ ls -ld parent_directory/
drwxrws--T. 3 netid2 hpcteam 512 Sep  4 13:36 parent_directory/
(puma) [netid1@wentletrap EXAMPLES]$ ls -l parent_directory/
total 4
drwxrws---. 2 netid2 hpcteam 0 Sep  4 13:36 other_user
(puma) [netid1@wentletrap EXAMPLES]$ rm -r parent_directory/other_user/
rm: cannot remove ‘parent_directory/other_user/’: Operation not permitted

Note that although netid1 has full rwx permissions set for both parent_directory and other_user, they cannot delete other_user because the sticky bit is set on the parent directory. Note that if netid2 were to try to delete it, they would be successful.

One thing to keep in mind is that the owner of the parent directory will still be able to delete the contents even if the sticky bit is set and they do not own the directory they are deleting.

Compression and Archiving¶

Tip

For very large directories, use the file transfer node. Hostname: filexfer.hpc.arizona.edu

Are you planning on transferring files to or from HPC? Do you have a lot of them? Then archiving is for you!

Archiving files is the process of consolidating one or more files or directories into a single, compressed package or archive file. This simplifies data management, reduces storage space, and streamlines file transfer and backup operations. Transferring a single archived file to an external backup location my result in transfer speeds that are an order of magnitude faster than transferring the same data as an uncompressed directory with thousands (or sometimes millions) of files.

As an example, we'll use the archiving tool tar. The syntax to create an archive is

tar <options> <output_archive_name> <contents>

We'll use the options czvf which stands for:

c: Create a new archive
z: Filter the archive through gzip
v: Verbosely print the output of the archiving process. Optional
f: User archive file

Archiving a directory dir and its contents into a single file called dir.tar.gz then looks like:

(puma) [user@wentletrap archiving_example]$ tar czvf dir.tar.gz dir
dir/
dir/subdir1/
dir/subdir1/file.txt
dir/file1.txt
dir/file3.txt
dir/file2.txt
(puma) [user@wentletrap archiving_example]$ ls
dir  dir.tar.gz

To unpack this archive, change the c to an x which stands for "extract":

tar xzvf dir.tar.gz