colab miscellaneous utils

Miscellaneous utils mainly intended for use in colab

Installation

pip install clmutils  # clm: colab-misc
# pip install clmutils -U to update

Or clone the repo https://github.com/ffreemt/colab-misc-utils.git in Colab/jupyter:

!git clone https://github.com/ffreemt/colab-misc-utils.git
%cd colab-misc-utils
!pip install -r requirements.txt

Usage: with a setting file dotenv/.env

1. setup_git

   from clmutils import setup_git, Settings
   config = Settings()
   setup_git(
     user_name=config.user_name,
     user_email=config.user_email,
     priv_key=config.gh_key
   )

   Do the usual git pull, amend codes and git push stuff.

   For more on Connecting to GitHub with SSH, refer to https://docs.github.com/cn/free-pro-team@latest/github/authenticating-to-github/connecting-to-github-with-ssh

2. setup_ssh_tunnel

   • Run these lines in Colab

   from clmutils import setup_ssh_tunnel, Settings
   config = Settings()
   setup_ssh_tunnel(
     remote_host=config.remote_host,
     remote_user=config.remote_user,
     priv_key=config.cl_key,
     remote_pubkey=config.remote_pubkey
   )

   • Append the value of cl_key_pub in dotenv/.env (below) in remote computer's ~/.ssh/authorized_keys.

   • In the remote computer:

   ssh colab

clmutils.Setting will look for dotenv or .env in /content/drive/MyDrive if google drive is momunted; otherwise it looks for .env in the current dir and parents dir.

dotenv/.env basically contains the necessary information to setup git for github or ssh tunnel:

• For git:
  • gh_key: private key
  • user_name/user_email;
• For reverse ssh tunnel:
  • cl_key/cl_key_pub: private key/public key of Colab
  • remote_pubkey: public_key of the remote computer
  • remote_user: login name

dotenv/.env is read by clmutils.Settin as follows

from clmutils import Settings
config = Settings()
# config = Setting(_env_file=file_path)

config will then have attributes: config.gh_key, user_name, user_email, cl_key, cl_key_pub, remote_pubkey, remote_user. If a particular attribute is not assigned a value in dotenv/.env, it defaults to empty string ("").

dotenv or .env has the following info and format (lines starting with a # are comments):

# for git
user_email = "...@......"
user_name = "......"
gh_key = "-----BEGIN EC PRIVATE KEY-----
...............................................................9
...............................................................J
K9ztlJBRRAOHh5sPhQ4QpdZH1v1rWeDWIQ==
-----END EC PRIVATE KEY-----
"

# for ssh tunnel
remote_host = "168.138.222.163"
remote_user = "ubuntu"

# colab's private key, to be put in ~/.ssh/id_ed25519
# or as specified in ~/.ssh/config
cl_key = "-----BEGIN OPENSSH PRIVATE KEY-----
.....................................................................W
QyNTUxOQAAACCClZGt/9ibAd9oxuWcfuSjnw0ERuY68/1QiirdrrtngQAAAJDlRQSF5UUE
hQAAAAtzc2gtZWQyNTUxOQAAACCClZGt/9ibAd9oxuWcfuSjnw0ERuY68/1QiirdrrtngQ
.....................................................................f
DQRG5jrz/VCKKt2uu2eBAAAACWZvci1jb2xhYgECAwQ=
-----END OPENSSH PRIVATE KEY-----
"

# to be put in remote computer's ~/.ssh/autorized_keys
cl_key_pub = "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIIKVka3/2JsB32jG5Zx+5KOfDQRG5jrz/VCKKt2uu2eB colab-key"

# to be put in colab's ~/.ssh/authorized_keys
remote_pubkey = "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIE60sowZ4M0MA5nTGIH1RN54zERTuWSddFKqyeWZzaKv for-colab"

Usage: withtout dotenv/.env

Set up github with ssh using clmutils.setup_git

For manually setting up github with ssh, refer to https://docs.github.com/en/free-pro-team@latest/github/authenticating-to-github/connecting-to-github-with-ssh

Assume you configure git as follows:

git config --global user.email your-email-address
git config --global user.name your-github-username

With clmutils, you'd do:

from clmutils import setup_git`

user_email = "your-email-address"
user_name = "your-github-username"
gh_key = \
"""
-----BEGIN EC PRIVATE KEY-----
MH.............................................................9
AwEHoUQDQgAEoLlGQRzIVHYw3gvC/QFw3Ru45zGawaBaCq6jTqdyH2Kp8zIB3TdJ
K9ztlJBRRAOHh5sPhQ4QpdZH1v1rWeDWIQ==
-----END EC PRIVATE KEY-----
""".strip() + "\n"

setup_git(user_email=user_email, user_name=user_name, priv_key=gh_key)

You then upload the public key for gh_key to https://github.com/settings/keys.

Refer to Step 2 https://support.cloudways.com/using-git-command-line-ssh/ for how to generate a private/public key pair. You can also use clmutils.gen_keypair to do that in Python.

Alternatively, set up github with ssh in 4 steps

1. Write a private key to ~/.ssh/gh-key

from clmutils import create_file
gh_key = \
"""
-----BEGIN EC PRIVATE KEY-----
MH.............................................................9
AwEHoUQDQgAEoLlGQRzIVHYw3gvC/QFw3Ru45zGawaBaCq6jTqdyH2Kp8zIB3TdJ
K9ztlJBRRAOHh5sPhQ4QpdZH1v1rWeDWIQ==
-----END EC PRIVATE KEY-----
""".strip() + "\n"
# Do not remove .strip() + "\n"
# the private key is very picky about format

create_file(gh_key, dest="~/.ssh/gh-key")

2. Set up github.com config for git push

from clmutils import append_content
config_github_entry = \
"""
Host github.com
   HostName github.com
   User git
   IdentityFile ~/.ssh/gh-key
"""
append_content(config_github_entry, dest="~/.ssh/config")

3. Verify that everything is OK, from a cell

!ssh -o StrictHostKeyChecking=no -T git@github.com

If you see something similar to

Hi your-name! You've successfully authenticated, but GitHub does not provide shell access.

you are good to go.

4. git config --global You can now set up git config global from a cell, e.g.

!git config --global user.email your-email-address
!git config --global user.name your-github-username
# !ssh-keyscan github.com >> ~/.ssh/known_hosts

You are ready to clone your own repo, run your app and generate new data, update the repo and push to github.

Utils planned

• ✅ setup_git sets up git for github.com

• ✅ reverse_ssh_tunnel sets up a reverse ssh tunnel to a remote host with via autossh

• Auxiliary utils

  • ✅ create_file creates a file with given mode, e.g. for .ssh/id_rsa or IdentityFile in .ssh/config

  • ✅ apppend_content appends some content to a file, e.g., for appended a public key to .ssh/authorized_keys

  • ✅ chmod600 chmod of a file

  • ✅ gen_keypair generates private/public key pair.

  • ✅ run_cmd wraps subprocess.check_output

  • ✅ run_cmd1 wraps subprocess.Popen

  • ✅ check_running shows running processes with names containings a certain patter; based on psutil.Proceess()'s cmdlineandstatus`

Demo: notebooks in Colab

git push from Colab: one line (setup_git())

!pip install clmutils
from clmutils import setup_git
gh_key = """..."""
user_name = "..."
user_email = "..."
setup_git(setup_git(user_email=user_email, user_name=user_name, priv_key=gh_key)

git push from Colab in several steps

(in Chinese, shouldn't be too difficult to follow without knowing any Chinese though, just click through 😃)

Reverse ssh tunnel for ssh to Colab VM

in English (I may provide a Chinese version later)