Getting Started¶

The classes in this module are intended to provide easy access to common infrastructural patterns in use at Thunderbird. This should…

help you set up a Pulumi project,
reduce most infrastructure configuration to values in a YAML file,
simplify the process of building complete infrastructural patterns.

As such, it is somewhat opinionated, requires certain usage patterns, and strongly suggests some usage conventions.

Prerequisites¶

To use this module, you’ll need to get through this checklist first:

Ensure Python 3.12 or greater is installed on your system.
Install Pulumi.
Understand the basic concepts of Pulumi, particularly Resources and Component Resources.
Provide an awscli configuration with your credentials and default region. (You do not have to install awscli, though you can read how to here. Some of these docs refer to helpful awscli commands.) The Pulumi AWS provider relies on the same configuration, though, so you must create the config file.
Optionally, set up an S3 bucket to store your Pulumi state in.

The Troubleshooting section has some details on how to work through some issues related to setup.

Quickstart¶

After ensuring you meet the above prerequisites, run the quickstart.sh script, adjusting the following command to refer to your particular project details:

./quickstart.sh \
    /path/to/project/root \ # The root of your code project where you want to set up a pulumi project
    pulumi_login_url      \ # URL to use with `pulumi login`; use "https://api.pulumi.com" for Pulumi Cloud
    project_name, \         # Name of your project as it will be known to pulumi
    stack_name, \           # Name of the first stack you want to create
    [code_version]          # Code version (git branch) that you want to pin. Optional; defaults to "main"

This will…

run you through some prompts where you can enter further project details,
install a simple Pulumi program intended to set up a basic networking landscape,
run a pulumi preview command to finish setting up the environment and confirm the project is working.

If you are using an S3 bucket to privately store your state, you’ll need to make sure you have configured your AWSCLI tool with an account that has permission to manipulate that bucket. Prefix your bucket name with s3:// to use as your pulumi_login_url value (e.g.,: s3://acme-awesomeapi-pulumi). If you will use Pulumi Cloud, use https://api.pulumi.com. If you have a self-hosted Pulumi Cloud API, you may specify your custom URL here.

The output should look something like this:

Previewing update (mystack):
     Type                              Name                                 Plan
 +   pulumi:pulumi:Stack               myproject-mystack                    create
 +   ├─ tb:network:MultiCidrVpc        myproject-mystack-vpc                create
 +   │  ├─ aws:ec2:Vpc                 myproject-mystack-vpc                create
 +   │  ├─ aws:ec2:Subnet              myproject-mystack-vpc-subnet-0       create
 +   │  ├─ aws:ec2:Subnet              myproject-mystack-vpc-subnet-1       create
 +   │  └─ aws:ec2:Subnet              myproject-mystack-vpc-subnet-2       create
 +   ├─ aws:ec2:RouteTableAssociation  myproject-mystack-vpc-subnetassoc-0  create
 +   ├─ aws:ec2:RouteTableAssociation  myproject-mystack-vpc-subnetassoc-1  create
 +   └─ aws:ec2:RouteTableAssociation  myproject-mystack-vpc-subnetassoc-2  create

Resources:
    + 9 to create

Manual Setup¶

If you want to do everything the Quickstart script does manually (or just understand this project framework better), follow this guide.

S3 bucket¶

Note

This step is optional. If you do not set up an S3 bucket, you can use Pulumi Cloud instead by specifying https://api.pulumi.com or a custom self-hosted URL when you run pulumi login in the next step.

Create an S3 bucket in which to store state for the project. You must have one bucket devoted to your project, but you can store multiple stacks’ state files in that one bucket. The bucket should not be public (treat these files as sensitive), and it’s usually a good idea to turn on versioning.

The name of an S3 bucket is used as part of a global domain, and so your bucket name must be globally unique. A good way to handle this is to include an organization name in your bucket name. As a template, you may use:

$ORG-$PROJECT_NAME-pulumi

Repo setup¶

You probably already have a repository with your application code in it. If not, create one now.

Create a subdirectory called pulumi and create a new project and stack in it. You’ll need the name of the S3 bucket or cloud host from the previous step here. If you are operating in an AWS region other than what is set as your default for AWSCLI, be sure to export AWS_REGION=us-east-1 or whatever else you may need to do to override that.

cd /path/to/pulumi/code
pulumi login s3://s3-bucket-name
pulumi new aws-python

Follow the prompts to get everything named.

Set up this module¶

Ensure your pulumi code directory contains a requirements.txt file with at least this repo listed:

tb_pulumi @ git+https://github.com/thunderbird/pulumi.git

You can pin your code to a specific version of this module by appending @branch_name to that. For example:

tb_pulumi @ git+https://github.com/thunderbird/pulumi.git@v0.0.13

Pulumi will need these requirements installed. On your first run of a pulumi preview command (or some others), Pulumi will attempt to set up its working environment. If this fails, or you need to make adjustments later, you can activate Pulumi’s virtual environment to perform pip changes. Assuming Pulumi’s virtual environment lives at venv, run:

source ./venv/bin/activate
pip install -U -r requirements.txt

You can now develop Python Pulumi code in that directory, as shown in the following section.

Use this module¶

When you issue pulumi commands (like “up” and “preview” and so on), it looks for a __main__.py file in your current directory and executes the code in that file.

__main__.py imports and uses the tb_pulumi library:

import tb_pulumi

# ...or you can import specific modules...

from tb_pulumi import (ec2, fargate, secrets)

Create a config file¶

Create a config file for each stack, i.e., config.$STACK.yaml (where $STACK maps to a Pulumi stack/application environment). This file maps parameters for tb_pulumi resources to their desired values. Currently, only the resources setting is formally recognized.

Note

When you run pulumi stack select $STACK on a tb_pulumi project, these two files become active in the Pulumi run: Pulumi.$STACK.yaml and config.$STACK.yaml. The former configures Pulumi for your stack (in addition to Pulumi.yaml) while the latter configures your tb_pulumi project.

Let’s look at an example tb_pulumi configuration file.

resources:
  tb:network:MultiCidrVpc:
    vpc:
      cidr_block: 10.0.0.0/16
      egress_via_internet_gateway: True
      enable_dns_hostnames: True
      enable_internet_gateway: True
      endpoint_interfaces:
        - ecr.api
        - ecr.dkr
        - logs
        - secretsmanager
      subnets:
        us-east-2a:
          - 10.0.101.0/24
        us-east-2b:
          - 10.0.102.0/24
        us-east-2c:
          - 10.0.103.0/24

At the top-level is the resources key. Nested inside are configurations for kinds of resources. This resource uses the tb_pulumi.network.MultiCidrVpc class.

Note

We recommend using resource key names that are named after the Pulumi Types for each resource. These are documented alongside each class in the tb_pulumi module. This is, however, completely optional convention.

The Pulumi Type for a MultiCidrVpc is tb:network:MultiCidrVpc, so we have chosen that as a name under which we define our MultiCidrVpc configs. We call this one particular config vpc (you normally need only one, though this convention allows for as many as you like).

These resources must still be defined in code (more on that later), but that code will largely just establish relationships between resource patterns (like using the ID of a VPC built by a MultiCidrVpc pattern as an input to a SecurityGroupWithRules pattern) and pass the YAML configuration through to those patterns. This simple relationship between the __main__.py code and the tb_pulumi YAML config is one core function of this project’s conventions.

The full listing of values supported by each pattern can be found by browsing the detailed tb_pulumi documentation. The barebones config example used in the quickstart can be found in our sample config.

Define a ThunderbirdPulumiProject¶

In your __main__.py file, start with a simple skeleton (or use our __main__.py example to start):

import tb_pulumi
project = tb_pulumi.ThunderbirdPulumiProject()

If you have followed the conventions outlined above, project.config is now a dict representation of the YAML file (see tb_pulumi.ThunderbirdPulumiProject.config). You can use this in the next step to feed parameters into resource declarations.

Moreover, as you create resources with this library, the project will track them, making them available to you later to act on as a group. This is explained in more detail on the Monitoring Resources page.

Declare ThunderbirdComponentResources¶

A Pulumi ComponentResource is a collection of related resources. In an effort to follow consistent patterns across infrastructure projects, the patterns available in this module all extend a custom class called a tb_pulumi.ThunderbirdComponentResource. If you have followed the conventions outlined so far, it should be easy to stamp out infrastructure with them by passing project.config config options into the constructors for these classes.

To start, for convenience, let’s pull the resources dict into a variable:

# Pull the "resources" config mapping
resources = project.config.get('resources')

Continuing the MultiCidrVpc example, let’s now pull the config for our vpc resource:

vpc_opts = resources['tb:network:MultiCidrVpc']['vpc']

And then define the MultiCidrVpc:

vpc = tb_pulumi.network.MultiCidrVpc(
    name=f'{project.name_prefix}-vpc',
    project=project,
    **vpc_opts)

The tb_pulumi.ThunderbirdPulumiProject.name_prefix value combines the project and stack name to form a convenient identifier to give your resources useful names. Here, we add -vpc to it, giving us something like myproject-stage-vpc.

Passing in the project created beforehand ensures the resources created by the MultiCidrVpc get tracked and become accessible to later aggregate functions. Skipping this will still result in the creation of these resources, but things like the tb_pulumi.monitoring.MonitoringGroup will not be able to detect it.

In Python, the double-star (**variable) notation unpacks a dict’s top level keys and values into named function parameters (called “keyword arguments” and often referred to as “kwargs”). In this case, all of the key/value pairs in the YAML configuration for the MultiCidrVpc called “vpc” get passed in as arguments to the function.

As a demonstration of this (and as a demonstration of code you should not write when using tb_pulumi), here is the equivalent function call without the YAML conversion:

vpc = tb_pulumi.network.MultiCidrVpc(
    name=f'{project.name_prefix}-vpc',
    project=project,
    cidr_block='10.0.0.0/16',
    subnets={
      'us-east-1a': '10.0.101.0/24',
      'us-east-1b': '10.0.102.0/24',
      'us-east-1c': '10.0.103.0/24',
    },
)

You may note some disadvantages to this:

Making configuration changes to an environment means editing code as opposed to adjusting YAML. We find the YAML to be more legible, and we find that after an environment is initially built, the infrastructural patterns do not often change. Rather, we adjust the details; we scale out new servers or use a larger instance type or allow a new IP address access to a system. These are easier to adjust when we can just find an entry in a sensibly organized config file and tweak it.
Reusing the same broad infrastructural definitions becomes much harder here. Suppose we want our staging environment to use different IP space than our production environment. If code is written this explicitly, we must introduce conditionals and break Pulumi’s comprehension of stacks to accomodate each environment’s distinguishing characteristics. Instead, we can apply different YAML configs to the same code to achieve environments that work the same way, but at different scales, against different sets of resources, etc.

Implementing ThunderbirdComponentResources¶

So you want to develop a new pattern to stamp out? Here’s what you’ll need to do:

Determine the best place to put the code. Is there an existing module that fits the bill?
Determine the Pulumi type string for it. This goes: org:module:class. The org should be unique to your organization. For Thunderbird projects, it should be tb. The module will be the Python submodule you’re placing the new class in (e.g., classes in network.py should use network here). The class is whatever you’ve called the class.
Design the class following these guidelines:
- The constructor should always accept, before any other arguments, the following positional options:
  
  name: The internal name of the resource as Pulumi tracks it.
  
  project: The ThunderbirdPulumiProject these resources belong to.
- The constructor should always accept the following keyword arguments:
  
  opts: A pulumi.ResourceOptions object which will get merged into the default set of arguments managed by the project.
- The constructor should explicitly define only those arguments that you intend to have default values which differ from the default values the provider will set, or which imply larger patterns.
- The constructor may accept a final **kwargs argument with arbitrary meaning. Because the nature of a component resource is to compile many other resources into one class, it is not implicitly clear what “everything else” should apply to. If this is implemented, its function should be clearly documented in the class. If this isn’t passed into the superconstructor, you will need to implement all superconstructor arguments into your constructor.
- The class should extend tb_pulumi.ThunderbirdComponentResource.
- The class should make an appropriate call to its superconstructor, which ensures the resources can be properly tracked in the project (among other things).
- Any resources you create must have the parent=self pulumi.ResourceOption set. Set an appropriate depends_on value when necessary.
- At the end of the __init__ function, you must call self.finish(), passing in a dictionary of resources (see tb_pulumi.ThunderbirdComponentResource.finish()). For tb_pulumi.monitoring.MonitoringGroup derivatives, call this at the end of the tb_pulumi.monitoring.MonitoringGroup.monitor() function instead.

Troubleshooting¶

Pythonic problems¶

This Pulumi code is developed against Python 3.12 or later. If this is not your default version, you’ll need to manage your own virtual environment.

Check your default version:

$ python -V
Python 3.12.6

If you need a newer Python, download and install it. Then you’ll have to set up the virtual environment yourself with something like this:

virtualenv -p /path/to/python3.12 venv
./venv/bin/pip install .

After this, pulumi commands should work. If 3.12 is your default version of Python, Pulumi should set up its own virtual environment, and you should not have to do this.

Shells other than Bash¶

Setup instructions in these docs are designed for use with the Bourne Again SHell (Bash). The Pulumi installer places the pulumi executable in a hidden folder in your home directory: ~/.pulumi/bin. The installer will add this to your default $PATH for you, but only on certain supported shells. If you use an alternative shell, you may need to do this step manually to avoid having to make an explicit path reference for every pulumi command.