Need script to run on user login for Ubuntu server 18.04 built via Packer

Question

Bit of an odd one, so bear with me.

I am building an Azure VM image using Packer. Part of the install requires some Python libraries to be installed. I can execute this via a Shell script in the Packer run list, however when the image is sealed and ready to be used those libraries are not installed.

I have found that the libraries aren't installed because it's installed to a user profile, and part of the Packer process is to delete said user profile upon 'resealing' the image for use.

I have Googled a number of ways, but I have't found anything that allows me to run as script on login for a user that doesn't exist yet, as when we first use the image and log in, the account is created.

Does anyone know how I could run this script on logon on first use?

#!/bin/bash

cd /


echo "Testing to make sure that script performed as expected, and basic scenarios work"
for cmd in python pip; do
    if ! command -v $cmd; then
        echo "$cmd was not installed or not found on PATH"
        exit 1
    fi
done


list1=(
    pandevice
    pan-python
    requests
    requests_toolbelt
    requests[security]
)

for i in "${list1[@]}"; do
    pip install $i
done


list2=(
    asn1crypto
    certifi
    cffi
    chardet
    cryptography
    enum34
    idna
    ipaddress
    pan-python
    pandevice
    pycparser
    pyOpenSSL
    requests
    requests-toolbelt
    six
    urllib3
)

for x in "${list2[@]}"; do
    if ! pip freeze | grep $x; then
        echo "$x was not installed or not found on PATH"
        exit 1
    fi
done

Just so you know, I have thought about using sudo pip install, however; What are the risks of running 'sudo pip'?

Answer 1

I am curious how your packer file looks like. I have run some scripts using root as follows.

packer file example:

{
    "variables": {
        "image_class": "centos7-base",
        "build_number": ""
    },
    "builders": [
        {
            "type": "googlecompute",
            "source_image_family": "{{ user `google_base_image_family` }}",
            "account_file": "{{ user `gce_service_account` }}",
            "image_family": "{{ user `_image_family` }}-{{ user `image_class` }}",
            "image_name": "wp-{{ user `image_class` }}-b{{ user `build_number` }}-{{ timestamp }}",
            "project_id": "{{ user `gce_project_id` }}",
            "ssh_username": "{{ user `ssh_username` }}",
            "subnetwork": "{{ user `gce_subnetwork` }}",
            "network": "{{ user `gce_network` }}",
            "zone": "{{ user `gce_zone` }}",
            "omit_external_ip": "true",
            "use_internal_ip": "true",
            "disk_size": 20

        }
    ],
    "provisioners": [
        {
            "type": "shell",
            "only": ["googlecompute"],
            "script": "base-image.sh",
            "skip_clean": true,
            "execute_command": "sudo chmod +x {{ .Path }}; sudo {{ .Vars }} {{ .Path }}"
        }
    ],
    "post-processors": [
    ]
}

This file is used for google cloud. The important section is provisioner: shell

"provisioners": [
        {
            "type": "shell",
            "only": ["googlecompute"],
            "script": "{{ user `provisioner_root` }}/shell/base-image.sh",
            "skip_clean": true,
            "execute_command": "sudo chmod +x {{ .Path }}; sudo {{ .Vars }} {{ .Path }}"

The install that happens in base_image is persistent. In general, I don't think it is a good idea of doing installs on running servers. That goes against the idea of immutable servers.