Skip to content

Latest commit

 

History

History
232 lines (187 loc) · 10.5 KB

PIPELINES-VARIABLE-REFERENCE.md

File metadata and controls

232 lines (187 loc) · 10.5 KB
Raw

Pipelines Environment Variable Usage

Environment variables (also known as environment parameters) are a first grade citizen of the pipelines utility.

The pipelines utility has support for Bitbucket Cloud Pipelines Plugin environment variables. This document describes along the list of default variables [BBPL-ENV] the level of support and provides useful information for setting and using these variables.

This document is a lengthier review of all variables at the time of writing and provides additional information next to the short introduction in the read me, both for limitations of the pipelines utility to this regard (and therefore feature planning) and usage guidelines for "dot env" environment files in project scope in a suggested (and best-intend compatible) fashion (façon).

Default Variables

Variable Name Remarks
CI all options; always set to "true"
BITBUCKET_BOOKMARK --trigger <ref> where <ref> is bookmark:<name>; for Mercurial projects
BITBUCKET_BRANCH --trigger <ref> where <ref> is branch:<name>, pr:<name>, pr:<source>:<destination>; (source) branch
BITBUCKET_BUILD_NUMBER all options; always set to "0"
BITBUCKET_CLONE_DIR all options; set by pipelines, it is the deploy directory inside the container (not clone directory) as pipelines has more options than cloning (it currently actually never clones)
BITBUCKET_COMMIT all options; always set to "0000000000000000000000000000000000000000"
BITBUCKET_DEPLOYMENT_ENVIRONMENT -/-; currently unsupported
BITBUCKET_DEPLOYMENT_ENVIRONMENT_UUID -/-; currently unsupported
BITBUCKET_EXIT_CODE all options; set to the exit status of the script for use in the after-script
BITBUCKET_GIT_HTTP_ORIGIN -/-; currently unsupported
BITBUCKET_GIT_SSH_ORIGIN -/-; currently unsupported
BITBUCKET_PARALLEL_STEP all options; in a parallel step set to zero-based index of the current step in the group, e.g. 0, 1, 2, ...
BITBUCKET_PARALLEL_STEP_COUNT all options; in a parallel step set to the total number of steps in the group, e.g. 5.
BITBUCKET_PR_DESTINATION_BRANCH --trigger <ref> where <ref> is pr:<source>:<destination> for the <destination> branch (see as well BITBUCKET_BRANCH); destination branch
BITBUCKET_PR_ID -/-; currently unsupported
BITBUCKET_PROJECT_KEY -/-; currently unsupported
BITBUCKET_PROJECT_UUID -/-; currently unsupported
BITBUCKET_REPO_FULL_NAME -/-; currently unsupported
BITBUCKET_REPO_OWNER all options; always set to current username from environment or if not available "nobody" (which might align w/ the Apache httpd project)
BITBUCKET_REPO_OWNER_UUID -/-; currently unsupported
BITBUCKET_REPO_SLUG all options; always set to base name of project directory
BITBUCKET_REPO_UUID -/-; currently unsupported
BITBUCKET_STEP_RUN_NUMBER all options; defaults to "1" and is set to "1" after the first run step
BITBUCKET_STEP_TRIGGERER_UUID -/-; currently unsupported
BITBUCKET_TAG --trigger <ref> where <ref> is tag:<name>; Git projects

Comments on Support and Usage

All environment variables are supported in the meaning they can be explicitly passed with any value of the users wish (or command) into the container. That is even true for those variables "always" set to a specific value - variables passed by the user will always override any automatically "always" set values.

Unsupported Features and Environment Variables

Some variables yet do not make sense as the pipelines utility does not support the underlying feature.

For example, the variable BITBUCKET_DEPLOYMENT_ENVIRONMENT and the depployments are not (yet) supported, therefore the variable BITBUCKET_DEPLOYMENT_ENVIRONMENT is not set. Setting it before call makes not much sense as it would be set for every script, and not the deployment script only.

However: Some variables which are valid per project and which are marked "currently unsupported" above do make sense to be put into the auto-loading environment files (.env and .env.dist). As long as their values are not considered a secret, they can be added including the values to .env.dist which intention is to be committed in the project repository (see your VCS/SCMs' documentation, e.g. for git see git-ignore [GIT-IGNORE]).

Otherwise, instead of poisoning your local environment, project secrets can be put into the private .env file (or any other *.env file via the --env-file option).

This is as long as you consider your local system safe (if your computer is connected to the internet this might not be the case and might be an overall short-coming with the Atlassian Bitbucket Cloud Pipelines Plugin, take care [!] as Secured Variables are just masked and can be read by any user who has write access to a Bitbucket repository - similar to read access on a/your local system).

Dealing w/ Secrets as Variable Values

For secret environment parameters, put them into the .env file only (or other .env files - if at all) which by intention is/are not to be committed in the project repository.

However: Required variables (with secrets) should be added with their name only to .env.dist file which is the project distribution parameter definition so that it clearly is documented which variables are required (this can be very helpful to document which secrets are needed and for non-secret what [sane] defaults are).

Example Configuration of an Environment File

Example configuration in a git based project with "dot env" files as they work w/ the pipelines utility:

The pipelines utility adheres to environment files as by the Docker project. That is mainly b/c it is highly dependent on the Docker client utility and the environment file-format in the Docker project ships with all needed features:

  • environment variables,
  • environment parameter imports / definitions and
  • comments .

The following example consists of three files: First a .gitignore file that takes care that no environment files but .env.dist is getting committed to the git repository.

This is necessary as these files might contain secrets (even in plain text) which should never go into the projects history (applies to git-utility managed projects only for .gitignore see your VCS utility if different for similar options).

From pipelines perspective, the file intended to be committed is the .env.dist file. It should contain all major variables but no secrets. This is possible by writing down the variable (parameter) name only. One such variable in the example is AWS_ACCESS_KEY_ID. The key-id and the access key are considered a secret in the following example.

The .env file then (can) contain private information if ok persisted in the local file-system. Different dot-env files can be provided w/ the --env-file <path> option of the pipelines (or docker) utility.

  • .gitignore file:

    # ...

# Private environment variable files (dot env)
/.env*
!/.env.dist

# ...

    (for more information on the .gitignore file format best see the gitignore documentation [GIT-IGNORE])

  • .env.dist file:

    # ---
# Pipeline: Gitlab Cloud Project
#
BITBUCKET_REPO_OWNER=ACME.Inc.
BITBUCKET_REPO_OWNER_UUID=894510a8-be2a-4660-a276-10fd5280cc61
BITBUCKET_REPO_SLUG=sudo-make-sandwich
BITBUCKET_REPO_UUID=894510a8-be2a-4660-a276-10fd5280cc61

# ---
# Pipeline: AWS ECR
#
AWS_REGION=eu-west-1
AWS_ACCESS_KEY_ID
AWS_SECRET_ACCESS_KEY

    (secrets in this .env file are hidden, only their variable names are given, e.g. AWS_SECRET_ACCESS_KEY)

  • .env file:

    # ---
# Pipeline: AWS ECR
#
AWS_REGION=eu-west-1
AWS_ACCESS_KEY_ID=ETI1ZUOWJ2GO5Q8IDFDC
AWS_SECRET_ACCESS_KEY=IzuobhF55og9fel6hleyuo4UOA0lUL9+GE2RmH6J

This example provides the project-wide information which environment variables are required for it. Additionally the project-wide default values are provided and the configuration setup does allow (local) (secret) configuration.

The comment blocks for a section are multi-line and follow an exemplary format of the first line of the comment block having a separating three-dash-line and a single (empty) line at the end after the second line containing a section title so that it is easy to diff between .env and .env.dist (and other) files.

Why Multi-Line Comments?

Why multi-line comments? In practice I found that single line comments are not that practical with text differs for sections, actually this is how this format came to life; when using various diff-utilities to compare these various dot-env files.

It is however up to every users discreet own experience(s) and need(s) on how to introduce sections in dot env files. It is only a subjective suggestion.

Note that there is a single empty line before any other but the first section comment block if the first section comment block starts on the very first line of the file. This is crucial for all differs I've run across.

Environment Parameters Considered for Distribution

If a project fully relies on Atlassian Bitbucket Cloud Pipelines Environment (by my reading of their documentation) the following environment variables and values should be distributed within the project:

  • BITBUCKET_PROJECT_KEY
  • BITBUCKET_PROJECT_UUID
  • BITBUCKET_REPO_FULL_NAME
  • BITBUCKET_REPO_OWNER
  • BITBUCKET_REPO_OWNER_UUID
  • BITBUCKET_REPO_SLUG - if not matching project base name
  • BITBUCKET_REPO_UUID

Note: Perhaps UUID values are gained best via the Bitbucket Cloud REST API on the shell command line.

References