My Setup to New Linux Server Account

In this article, I list some procedures I do when setting up new Linux server account. They’ll be helpful in the later development stages.

1. Change the default shell

This step is optional. Sometimes the default shell in the machine is merely sh. We can change it to other better shells like bash, zsh, for easier use.

1.1 Show the current shells

You can display the current shell name by either the following commands:

echo $0
# OUTPUT
-bash

echo $SHELL
# OUTPUT
/bin/bash

1.2 Get all available shells

To check what shells are installed, type the following commands:

cat /etc/shells
# OUTPUT

# /etc/shells: valid login shells
/bin/sh
/bin/bash
/bin/rbash
/bin/dash
/usr/bin/tmux
/usr/bin/screen
/bin/zsh
/usr/bin/zsh

1.3 Change the default shell

Use the chsh (change shell) command to change the shell, with -s flag:

chsh -s /bin/bash
Password:

Then type the password of the login account (The password is hidden, just type it, and press Enter). Finally, you can quit the shell, and restart it once. You’ll see the new shell.

2. Install conda

Python environment is very essential in my workflow, especially in deep learning. Please refer to the official docs:

https://docs.conda.io/projects/conda/en/latest/user-guide/install/linux.html

3. Setup ssh key (password-free login)

Modified from the blog here.

(Replace {KeyName} to any name you like)

3.1. Generate SSH Key Pair on Your Local Machine

(macOS Users please do the following)

## Execute the following commands on your LOCAL machine
cd ~/.ssh
ssh-keygen -t rsa -b 1024 -f "{KeyName}" -C "{Put Any Comment You Like}"
ssh-add -K ./{KeyName}

To check if you make it right, type the following command and you should see a string as the output.

cat ~/.ssh/{KeyName}.pub

(Windows Users please do the following)

## Execute the following commands on your local machine
cd C:\Users\{UserName}\.ssh
ssh-keygen -t rsa -b 1024 -f "{KeyName}" -C "{Put Any Comment You Like}"

To check if you make it right, double click the key file (C:\Users\{UserName}\.ssh\{KeyName}.pub) to view it and you should see a string.

Notes:

• The commands above are executed on your own computer, instead of the server.
• It is fine to use the SSH key file generated before (eg. id_rsa.pub), if you have.
• ssh-keygen will ask you to set a paraphrase, which improves the security of using SSH key authentication. Type “Enter” for no paraphrase.

Parameters for ssh-keygen command explanation:

• -t: type for ssh key generation. Here we use rsa
• -b: bits
• -f: name of your ssh key file. You are recommended to set this parameter in case it
  is conflict with the ssh key file you generated before.
• -C: comments to distinguish the ssh key file from others

3.2. Transfer Your SSH Public Key File to the Server

We use the scp command to transfer the public key generated in the former step to the server.

~/.ssh/{KeyName}.pub is the public key you just generated. {Your Account Name}@{Server ip} is your server account name and ip address.

## Execute the following commands on your LOCAL machine
# (macOS)
scp ~/.ssh/{KeyName}.pub {Your Account Name}@{Server ip}:~

# (windows)
scp C:\Users\{UserName}\.ssh\{KeyName}.pub {Your Account Name}@{Server ip}:~

Notes:

• This command will ask you for password for data transfer. Please make sure that you type in the correct password. If you are Windows user and you want to copy & paste the password for convenience, try “Ctrl + Shift + V” if you fail to paste the password with “Ctrl + V”.

3.3. Configure authorized_keys on the Server

## Login the server with password. Do this step in locally to establish a ssh connection
ssh {Your Account Name}@{Server ip}
## All the following commands should be executed on the ssh connection just established. (i.e., execute them on the server)
mkdir -p ~/.ssh
cat ./{KeyName}.pub >> ~/.ssh/authorized_keys
rm -f ~/{KeyName}.pub
chmod 700 ~/.ssh
chmod 600 ~/.ssh/authorized_keys

To check if you make it right, execute the following command and you should see a string as the output that is the same as in Step-1.

cat ~/.ssh/authorized_keys

3.4. Prepare for SSH Connection with Key Authentication

Add the following content to ~/.ssh/config (In windows, it’s C:\Users\{UserName}\.ssh\config) file on your LOCAL machine:

Host {Any Name You Like}
    HostName {ip}
    IdentityFile ~/.ssh/{KeyName}
    User {Your Account Name}

Notes:

• It is recommended to configure with the Remote SSH extension on VS Code. Please refer to Remote Development using SSH for more information.

3.5. Login the Server Password-free with SSH key Authentication

ssh {Your Account Name}@{Server ip}

or

ssh {Your Account Name}@{Any Name You Like} # above in Sec. 4

or

ssh {Any Name You Like} # above in Sec. 4

4. Set some alias

Alias convenient us typing some common commands.

Add the following to the end of ~/.bashrc:

alias cf="ls -l | grep "^-" | wc -l"
alias ns="nvidia-smi"
alias py="python"
alias act="conda activate"

The first alias, cf means count files, can count number of visible files (excluding directories) under the current working directory.

The second alias, ns is short for nvidia-smi to check for the GPU information.

The third alias py is short for python.

The fourth alias act can be used like act py39 (activate the conda py39 environment)

5. Configure the CUDA compiler version

This step is optional. If you want to compile some PyTorch operators locally, you should set the nvcc version same to the PyTorch runtime version.

Add the following to the end of ~/.bashrc:

export PATH=/usr/local/cuda-11.1/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-11.1/lib64:$LD_LIBRARY_PATH

Change the CUDA directory (/usr/local/cuda-11.1 above) to the one you actually want to use. For example, the PyTorch is compiled under cuda-11.3, you probably want to use nvcc with cuda-11.x, but not cuda-10.2.

6. ~/.bashrc template

# path to some executable
export PATH=<Your-Path-to-Some-EXE>:$PATH  

# path to cuda, change version to yours
export PATH=/usr/local/cuda-11.1/bin:$PATH 

# path to cuda libraries, change version to yours
export LD_LIBRARY_PATH=/usr/local/cuda-11.1/lib64:$LD_LIBRARY_PATH 

# path to self-defined temp directory
export TMPDIR=/<Path-to-Temp-Dir>/tmp
# path to some common visited places (shortcut)
export workdir=/<Path-to-Common-Dir>

# Shortcuts
alias countf="ls -l | grep "^-" | wc -l" # count files
alias countd="ls -l | grep "^d" | wc -l" # count directories
alias sized="du -h --max-depth=1" # size of each subdirectory
alias ns="nvidia-smi" 
alias py="python"
alias act="conda activate"
alias pys="python setup.py install"