azure_scripts

CAUTION

NOTING HERE IS TESTED

Application:

Install packages on Azure linux Virtual Machine instance. Install CUDA toolkit because you'll use this typically on a CUDA-Machine Save money by deallocating the VM instance after installation.

Run a [python] script with psub and forget. psub will deallocate resources once script is over.

Installation:

Pull to local machine

• On your real local machine pull:

git pull https://github.com/pradyparanjpe/azure_scripts.git

Modifications

1. See Package_Requirements

2. Add your packages in the packages folder's copy inside your local machine

Copy

• Start Azure virtual machine and copy the contents of this modified git folder to the Azure virtual machine using a command such as

rsync -auvz ./azure_scripts uname@azure.ip:~/.

Initiate azurecli Requirements

• See Requirements section

Uninstallation

• Delete the folder azure_scripts

rm -rf ${HOME}/azure_scripts/

• remove line containing "azure_scripts" from the file ~/.bashrc

sed -i -e 's/^.*azure_scripts.*$//g' ${HOME}/.bashrc

Usage

1. Run psub command of the form

nohup psub [-e <PACKAGE_FOLDER_NAME>] <command> &

- NOTE THE TRAILING '&'

2. Kill a submitted process, and all child-process created by it.

p9kill -15 <PID>

Install Packages

• On Azure virtual machine,

psub linux_pypkg_install.sh [<PACKAGE_FOLDER_NAME>] 2>&1 >> installation.log &

Requirements

After copying azure_scripts to the Azure VM,

1. Read ALL related documentation

2. Run linux_init_azurecli.sh with modifications

bash ~/azure_scripts/bin/linux_init_azurecli.sh -n <VMN> -g <VMG> -p <AZUREGITPATH>

• Login to azure virtual machine:

az login

• Follow instuctions (Open a link in web-browser and supply the displayed KEY)

• Now, the AzureCLI API has been granted access to manage your Azure virtual machines.

3. Update linux environment

source ~/.bashrc

Package_Requirements

1. Python:
   • Package should be pip3 compatible.
   • setup.py must be present in the parent package folder, I do not look recursively.
   • Only pip3 dependencies will be installed
2. C/Cpp:
   • Package should be generally installable by issuing ./configure; make; make install
   • This feature is planned, and may be added in future.