tgwrap 0.9.2

A (terragrunt) wrapper around a (terraform) wrapper around ....

tg-wrap

This app simply wraps terragrunt (which is a wrapper around terraform, which is a wrapper around cloud APIs, which is...).

Wait, why on earth do we need a wrapper for a wrapper (for a wrapper)?

Well, first of all it is pretty opinionated so what works for us, doesn't necessarily work for you.

But our reasoning for creating this is as follows:

1. Less typing

terraform is great, and in combination with terragrunt even greater! But let's face it, terragrunt does not excel in conciseness! The options are pretty long, which leads to lots of typing. We don't like typing!

2. Testing modules locally

However, more importantly, we are heavily utilising TERRAGRUNT_SOURCE when developing.

The thing is that as long as you use run-all you can use one setting for that variable (and conveniently set it as an environment variable), while if you run a regular command, you need to specify the full path. Which is obviously different for each project.

Which leads to (even) more typing, and worse: a higher chance for errors.

Luckily you can use run-all and add the appriopriate flags to ensure it behaves like a regular plan|apply|destroy etc. But again, more typing.

Nothing a bunch a aliases can't solve though!

3. But the original reason was: Errors when using run-all are challenging

One of the main boons of terragrunt is the ability to break up large projects in smaller steps while still retaining the inter-dependencies. However, when working on such a large project and something goes wrong somewhere in the middle is pretty challenging.

terragrunt's error messages are pretty massive, and this is extrapolated with every individual project in your dependency chain.

And if it fails somewhere at the front, it keeps on trying until the last one, blowing up your terminal in the process.

So we wanted a possibility to run the projects step by step, using the dependency graph of terragrunt and have a bit more control over it.

And when you can run it step by step, you can make the process also re-startable, which is also pretty handy!

And this was not something a bunch of aliases could solve, hence this wrapper was born. And while we we're at it, replacing the aliases with this was then pretty straightforward next step as well.

4. Analyzing plan files

When using the run-all, analyzing what is about to be changed is not going to be easier. Hence we created the tgwrap analyze function that lists all the planned changes and (if a config file is availabe) calculates a drift score and runs a terrasafe style validation check.

It needs a config file as follows:

---
#
# Critically of resources as interpreted by 'tgwrap analyze'.
# It uses it for a 'terrasafe' like validation if resources can safely be deleted.
# On top of that it tries to analyze and quantify the drift impact of the changes,
# so that this can be monitored.
#
low:
  # defaults:
  #   terrasafe_level: ignore_deletions
  #   drift_impact:
  #     default: minor
  #     delete: medium
  azuread_application.: {} # if we you want to use the defaults
  azuread_app_role_assignment: # or if you want to override these
    drift_impact:
      delete: minor
  # and so on, and so forth
medium:
  # defaults:
  #   terrasafe_level: ignore_deletion_if_recreation
  #   drift_impact:
  #     default: medium
  #     delete: major
  azurerm_data_factory_linked_service_key_vault.: {}
  # and so on, and so forth
high:
  # defaults:
  #   terrasafe_level: unauthorized_deletion
  #   drift_impact:
  #     default: major
  #     update: medium
  azuread_group.:
    drift_impact:
      create: minor
      update: minor
  azurerm_application_insights.: {}
  # and so on, and so forth

Speeding up the performance of analyze

This analyze function turned out to be pretty slow, where most of the time went into the terragrunt show function that is executed for each individual module.

This was a bit surprising as the plan file is already available on the file system, but it turns out that terragrunt is taking quite a bit of time for managing the depdencies. Even when you're excluding the external dependencies and are located in a particular module.

So, if you add the following to your root terragrunt.hcl:

terraform {
  after_hook "link_to_current_module" {
    commands = ["init", "plan", "apply", "validate", "destroy"]
    execute  = ["bash", "-c", "ln -sf $(pwd) ${get_terragrunt_dir()}/.terragrunt-cache/current"]
  }
}

The directory where the plan file is stored (including the other resources that terraform needs) becomes predictable and it becomes possible to run a native terraform show (instead terragrunt show) which dramatically speed up things.

Just set the proper value as an environment variable:

export TGWRAP_PLANFILE_DIR=".terragrunt-cache/current"

Or pass it along with the --planfile-dir|-P option and it will use that.

Logging the results

tgwrap supports logging the analyze results to an Azure Log Analytics custom table.

For that, the custom table need to be present, including a data collection endpoint and associated data collection rule.

When you want to activate this, just pass --data-collection-endpoint (or, more conveniently, set the TGWRAP_ANALYZE_DATA_COLLECTION_ENDPOINT environment variable) with the url to which the data can be posted.

Note that for this to work, tgwrap assumes that there is a functioning azure cli available on the system.

A payload as below will be posted, and the log analytics table should be able to accomodate for that:

[
  {
    "scope": "terragrunt/dlzs/data-platform/global/platform/rbac/",
    "principal": "myself",
    "repo": "https://gitlab.com/my-git-repo.git",
    "creations": 0,
    "updates": 0,
    "deletions": 0,
    "minor": 0,
    "medium": 0,
    "major": 0,
    "unknown": 0,
    "total": 0,
    "score": 0.0,
    "details": [
      {
        "drifts": {
          "minor": 0,
          "medium": 0,
          "major": 0,
          "unknown": 0,
          "total": 0,
          "score": 0.0
        },
        "all": [],
        "creations": [],
        "updates": [],
        "deletions": [],
        "unauthorized": [],
        "unknowns": [],
        "module": ""
      }
    ]
  }
]

More than a wrapper

Over time, tgwrap became more than a wrapper, blantly violating #1 of the unix philosophy: 'Make each program do one thing well'.

For instance, the 'analyze' functionality is already an example, but more features such as deploying a landing zone has crept into the application. It makes sense for how we're using it, but we're fully aware this makes it less generically applicable.

Usage

# general help
tgwrap --help

tgwrap run -h
tgwrap run-all -h

# run a plan
tgwrap plan # which is the same as tgwrap run plan

# run-all a plan
tgwrap run-all plan

# run-all with excluding a particular directory
tgwrap run-all plan -E 'excluded-dir/*'
# or a directory somewhere further down in the path
tgwrap run-all plan -E '**/excluded-dir/**'

# or do the same in step-by-step mode
tgwrap run-all plan -s

# or excluding (aka ignoring) external dependencies
tgwrap run-all plan -sx

# if you want to add additional arguments it is recommended to use -- as separator (although it *might* work without)
tgwrap output -- -json

Note: special precautions are needed when passing on parameters that contain quotes. For instance, if you want to move state like below, escape the double quote in the staate address:

tgwrap state mv 'azuread_group.this[\"viewers\"]' 'azuread_group.this[\"readers\"]'

A word about escaping inputs

Your shell is escaping special characters such as * and " before passing it to the program (tgwrap in this case). So some inputs need to be escaped in order to function properly.

For example:

# to exclude certain modules from an action (such as analyze):
tgwrap analyze -E 'integrations/\*/\*'

# to import a resource that has a " in its address:
tgwrap import -a 'azuread_group.this[\"my_group\"]' -i ${GROUP_ID}

Deploy manifests

In order to easily deploy a new version of the terraform (and associated terragrunt) modules, we include a manifest file in the root of the landing zone:

---
git_repository: ssh://[e-mail]/my-org/my-terraform-modules-repo.git
base_path: terragrunt/my-platform # path where the terragrunt modules are stored 
config_path: terragrunt/config/platform-dev # path (relative to base path) where the config files are stored

deploy: # which modules do you want to deploy
  dtap:
    applies_to_stages:
      - dev
      - tst
      - acc
      - prd
    source_stage: dev
    source_dir: platform # optional, if the source modules are not directly in the stage dir, but in <stage>/<source_dir> directory
    base_dir: platform # optional, if you want to deploy the base modules in its own dir, side by side with substacks
    config_dir: ../../../config # this is relative to where you run the deploy
    configs:
      - my-config.hcl
      - ../my-ss-config-dir
    # skip_all_modules: True # could be handy if you only want to deploy substacks and no regular modules
    exclude_modules: # these modules will always be excluded, can be omitted
      - my-specific-module
    include_modules: {} # omit or use an empty dict for all of them
      # or specify your modules as follows
      # base: {} # just a simple include
      # networking-connected: # or a bit more complicated
      #  - source: networking
      #  - target: networking-connected

substacks:
  is01:
    source: shared-integration/intsvc01
    target: integration/is01
    exclude_modules:  # a list of modules that will always be excluded, can be omitted
      - my-specific-module
    configs:
      - my-ss-config.hcl
      - ../my-ss-config-dir
  is02:
    applies_to_stages: # optional list of stages to include the substack in
      - dev
    source: shared-integration/intsvc01
    target: integration/is02

# global configuration files that are deployed as well
# note that these files are typically applicable to all landing zones and stages!
# this might lead to unexpected behaviour
# note that you can exclude syncing these files with a command line switch
global_config_files:
  root-terragrunt:
    source: ../../terragrunt.hcl # relative to base_path
    target: ../../terragrunt.hcl # can be omitted, then it is same as source path
  terrasafe-config:
    source: ../../terrasafe-config.json

Generating change logs

tgwrap can generate a change log by running:

tgwrap change-log [--changelog-file ./CHANGELOG.md]

The (optional) change log file will be, if passed along. tgwrap then checks the file for the existance of a start and end markers, in the following format:

  start_marker = '<!-- BEGINNING OF OF TGWRAP CHANGELOG SECTION -->'
  end_marker = '<!-- END OF TGWRAP CHANGELOG SECTION -->'

If they exist, everything between these lines will be replaced by the new change log.

Development

In order to develop, you need to apply it to your terragrunt projects. For that you can use the --terragrunt-working-dir option and just run it from the poetry directory. Alternatively you can use the tgwrap-dev script and invoke that from your terragrunt directories. Either put it in your PATH or create an alias for convenience.