Compatible Terraform Modules for Network Infrastructure Automation

Consul-Terraform-Sync (CTS) automates execution of Terraform modules through tasks. A task is a construct in CTS that defines the automation of Terraform and the module.

Module Specifications

Compatible modules for CTS follow the standard module structure. Modules can use syntax supported by Terraform version 0.13 and newer.

Compatibility Requirements

Below are the two required elements for module compatibility with CTS

  1. Root module - Terraform has one requirement for files in the root directory of the repository to function as the primary entrypoint for the module. It should encapsulate the core logic to be used by CTS for task automation. main.tf is the recommended filename for the main file where resources are created.
  2. services input variable - CTS requires all modules to have the following input variable declared within the root module. The declaration of the services variable can be included at the top of the suggested variables.tf file where other input variables are commonly declared. This variable functions as the response object from the Consul catalog API and surfaces network information to be consumed by the module. It is structured as a map of objects.

Optional Input Variables

In addition to the required services input variable, CTS provides additional, optional input variables to be used within your module. Support for an optional input variable requires two changes:

  1. Updating the Terraform Module to declare the input variable in the suggested variables.tf
  2. Adding configuration to the CTS task block to define the module input values that should be provided to the input variables

See below sections for more information on defining module input and declaring optional input variables in your Terraform module.

Module Input

A task monitors Consul objects that are defined by the task’s configuration. The Consul objects can be used for the module input that satisfies the requirements defined by the task’s Terraform module’s input variables.

A task’s module input is slightly different from the task’s condition, even though both monitor defined objects. The task’s condition monitors defined objects with a configured criteria. When this criteria is satisfied, the task will trigger.

The module input, however, monitors defined objects with the intent of providing values or metadata about these objects to the Terraform module. The monitored module input and condition objects can be the same object, such as a task configured with a condition "services" block and use_as_module_input set to true. The module input and condition can also be different objects and configured separately, such as a task configured with a condition "catalog-services and module_input "consul-kv" block. As a result, the monitored module input is decoupled from the provided condition in order to satisfy the Terraform module.

Each type of object that CTS monitors can only be defined through one configuration within a task definition. For example, if a task monitors services, the task cannot have both condition "services" and module_input "services" configured. See Task Module Input configuration for more details.

There are a few ways that a module input can be defined:

  • services list (deprecated) - The list of services to use as module input.
  • condition block’s use_as_module_input field - When set to true, the condition’s objects are used as module input.
    • Field was previously named source_includes_var (deprecated)
  • module_input blocks - This block can be configured multiple times to define objects to use as module input.
    • Block was previously named source_input (deprecated)

Multiple ways of defining a module input adds configuration flexibility, and allows for optional additional input variables to be supported by CTS alongside the services input variable.

Additional optional input variable types:

Services Module Input

Tasks configured with a services module input monitor for changes to services. Monitoring is either performed on a configured list of services or on any services matching a provided regex.

Sample rendered services input:

Terraform Modules - 图1

terraform.tfvars

  1. services = {
  2. "web.test-server.dc1" = {
  3. id = "web"
  4. name = "web"
  5. kind = ""
  6. address = "127.0.0.1"
  7. port = 80
  8. meta = {}
  9. tags = ["example"]
  10. namespace = ""
  11. status = "passing"
  12. node = "pm8902"
  13. node_id = "307625d3-a1cf-9e85-ff81-12017ca4d848"
  14. node_address = "127.0.0.1"
  15. node_datacenter = "dc1"
  16. node_tagged_addresses = {
  17. lan = "127.0.0.1"
  18. lan_ipv4 = "127.0.0.1"
  19. wan = "127.0.0.1"
  20. wan_ipv4 = "127.0.0.1"
  21. }
  22. node_meta = {
  23. consul-network-segment = ""
  24. }
  25. },
  26. }

In order to configure a task with the services module input, the list of services that will be used for the input must be configured in one of the following ways:

The services module input operates by monitoring the Health List Nodes For Service API and provides the latest service information to the input variables. A complete list of service information that would be provided to the module is expanded below:

AttributeDescription
idA unique Consul ID for this service. The service id is unique per Consul agent.
nameThe logical name of the service. Many service instances may share the same logical service name.
addressIP address of the service host — if empty, node address should be used.
portPort number of the service
metaList of user-defined metadata key/value pairs for the service
tagsList of tags for the service
namespaceConsul Enterprise namespace of the service instance
statusRepresentative status for the service instance based on an aggregate of the list of health checks
nodeName of the Consul node on which the service is registered
node_idID of the node on which the service is registered.
node_addressThe IP address of the Consul node on which the service is registered.
node_datacenterData center of the Consul node on which the service is registered.
node_tagged_addressesList of explicit LAN and WAN IP addresses for the agent
node_metaList of user-defined metadata key/value pairs for the node

Below is an example configuration for a task that will execute on a schedule and provide information about the services matching the regexp parameter to the task’s module.

  1. task {
  2. name = "services_condition_task"
  3. description = "execute on changes to services whose name starts with web"
  4. providers = ["my-provider"]
  5. module = "path/to/services-condition-module"
  6. condition "schedule" {
  7. cron = "* * * * Mon"
  8. }
  9. module_input "services" {
  10. regexp = "^web.*"
  11. }
  12. }

Consul KV Module Input

Tasks configured with a Consul KV module input monitor Consul KV for changes to KV pairs that satisfy the provided configuration. The Consul KV module input operates by monitoring the Consul KV API and provides these key values to the task’s module.

Sample rendered consul KV input:

Terraform Modules - 图2

terraform.tfvars

  1. consul_kv = {
  2. "my-key" = "some value"
  3. }

To configure a task with the Consul KV module input, the KVs which will be used for the input must be configured in one of the following ways:

Below is a similar example to the one provided in the Consul KV Condition section. However, the difference in this example is that instead of triggering based on a change to Consul KV, this task will instead execute on a schedule. Once execution is triggered, Consul KV information is then provided to the task’s module.

  1. task {
  2. name = "consul_kv_schedule_task"
  3. description = "executes on Monday monitoring Consul KV"
  4. module = "path/to/consul-kv-module"
  5. condition "schedule" {
  6. cron = "* * * * Mon"
  7. }
  8. module_input "consul-kv" {
  9. path = "my-key"
  10. recurse = true
  11. datacenter = "dc1"
  12. namespace = "default"
  13. }
  14. }

Catalog Services Module Input

Tasks configured with a Catalog Services module input monitors for service and tag information provided by the Catalog List Services API. The module input is a map of service names to a list of tags.

Sample rendered catalog-services input:

Terraform Modules - 图3

terraform.tfvars

  1. catalog_services = {
  2. "api" = ["prod", "staging"]
  3. "consul" = []
  4. "web" = ["blue", "green"]
  5. }

To configure a task with the Catalog Services module input, the catalog services which will be used for the input must be configured in one of the following ways:

Note: Currently there is no support for a module_input "catalog-services" block.

Example of a catalog-services condition which supports module input through use_as_module_input:

  1. task {
  2. name = "catalog_services_condition_task"
  3. description = "execute on registration/deregistration of services"
  4. providers = ["my-provider"]
  5. module = "path/to/catalog-services-module"
  6. condition "catalog-services" {
  7. datacenter = "dc1"
  8. namespace = "default"
  9. regexp = "web.*"
  10. use_as_module_input = true
  11. node_meta {
  12. key = "value"
  13. }
  14. }
  15. }

How to Create a Compatible Terraform Module

You can read more on how to create a module or work through a tutorial to build a module. CTS is designed to integrate with any module that satisfies the specifications in the following section.

The repository hashicorp/consul-terraform-sync-template-module can be cloned and used as a starting point for structuring a compatible Terraform module. The template repository has the files described in the next steps prepared.

First, create a directory to organize Terraform configuration files that make up the module. You can start off with creating two files main.tf and variables.tf and expand from there based on your module and network infrastructure automation needs.

The main.tf is the entry point of the module and this is where you can begin authoring your module. It can contain multiple Terraform resources related to an automation task that uses Consul service discovery information, particularly the required services input variable. The code example below shows a resource using the services variable. When this example is used in automation with CTS, the content of the local file would dynamically update as Consul service discovery information changes.

Terraform Modules - 图4

main.tf

  1. # Create a file with service names and their node addresses
  2. resource "local_file" "consul_services" {
  3. content = join("\n", [
  4. for _, service in var.services : "${service.name} ${service.id} ${service.node_address}"
  5. ])
  6. filename = "consul_services.txt"
  7. }

Something important to consider before authoring your module is deciding the condition under which it will execute. This will allow you to potentially use other types of CTS provided input variables in your module. It will also help inform your documentation and how users should configure their task for your module.

Services Variable

To satisfy the specification requirements for a compatible module, copy the services variable declaration to the variables.tf file. Your module can optionally have other variable declarations and CTS provided input variables in addition to var.services.

Terraform Modules - 图5

variables.tf

  1. variable "services" {
  2. description = "Consul services monitored by Consul-Terraform-Sync"
  3. type = map(
  4. object({
  5. id = string
  6. name = string
  7. kind = string
  8. address = string
  9. port = number
  10. meta = map(string)
  11. tags = list(string)
  12. namespace = string
  13. status = string
  14. node = string
  15. node_id = string
  16. node_address = string
  17. node_datacenter = string
  18. node_tagged_addresses = map(string)
  19. node_meta = map(string)
  20. cts_user_defined_meta = map(string)
  21. })
  22. )
  23. }

Keys of the services map are unique identifiers of the service across Consul agents and data centers. Keys follow the format service-id.node.datacenter (or service-id.node.namespace.datacenter for Consul Enterprise). A complete list of attributes available for the services variable is included in the documentation for CTS tasks.

Terraform variables when passed as module arguments can be lossy for object types. This allows CTS to declare the full variable with every object attribute in the generated root module, and pass the variable to a child module that contains a subset of these attributes for its variable declaration. Modules compatible with CTS may simplify the var.services declaration within the module by omitting unused attributes. For example, the following services variable has 4 attributes with the rest omitted.

Terraform Modules - 图6

variables.tf

  1. variable "services" {
  2. description = "Consul services monitored by Consul-Terraform-Sync"
  3. type = map(
  4. object({
  5. id = string
  6. name = string
  7. node_address = string
  8. port = number
  9. status = string
  10. })
  11. )
  12. }

Catalog Services Variable

If you are creating a module for a catalog-services condition, then you have the option to add the catalog_services variable, which contains service registration and tag information. If your module would benefit from consuming this information, you can copy the catalog_services variable declaration to your variables.tf file in addition to the other variables.

Terraform Modules - 图7

variables.tf

  1. variable "catalog_services" {
  2. description = "Consul catalog service names and tags monitored by Consul-Terraform-Sync"
  3. type = map(list(string))
  4. }

The keys of the catalog_services map are the names of the services that are registered with Consul at the given datacenter. The value for each service name is a list of all known tags for that service.

We recommend that if you make a module with with a catalog-services condition, that you document this in the README. This way, users that want to configure a task with your module will know to configure a catalog-services condition block.

Similarly, if you use the catalog_services variable in your module, we recommend that you also document this usage in the README. Users of your module will then know to set the catalog-services condition use_as_module_input configuration to be true. When this field is set to true, CTS will declare the catalog_services variable in the generated root module, and pass the variable to a child module. Therefore, if this field is configured inconsistently, CTS will error and exit.

Consul KV Variable

If you are creating a module for a consul-kv condition, then you have the option to add the consul_kv variable, which contains a map of the keys and values for the Consul KV pairs. If your module would benefit from consuming this information, you can copy the consul_kv variable declaration to your variables.tf file in addition to the other variables.

Terraform Modules - 图8

variables.tf

  1. variable "consul_kv" {
  2. description = "Keys and values of the Consul KV pairs monitored by Consul-Terraform-Sync"
  3. type = map(string)
  4. }

If your module contains the consul_kv variable, we recommend documenting the usage in the README file so that users know to set the use_as_module_input configuration to true in the consul-kv condition. Setting the field to true instructs CTS to declare the consul_kv variable in the generated root module and pass the variable to a child module. Therefore, if this field is configured inconsistently, CTS will error and exit.

Module Input Variables

Network infrastructure differs vastly across teams and organizations, and the automation needs of practitioners are unique based on their existing setup. Input variables can be used to serve as customization parameters to the module for practitioners.

  1. Identify areas in the module where practitioners could tailor the automation to fit their infrastructure.
  2. Declare input variables and insert the use of variables throughout module resources to expose these options to practitioners.
  3. Include descriptions to capture what the variables are and how they are used, and specify custom validation rules for variables to provide context to users the expected format and conditions for the variables.
  4. Set reasonable default values for variables that are optional, or omit default values for variables that are required module arguments.
  5. Set the sensitive argument for variables that contain secret or sensitive values. When set, Terraform will redact the value from output when Terraform commands are run.

Terraform is an explicit configuration language and requires variables to be declared, typed, and passed explicitly through as module arguments. CTS abstracts this by creating intermediate variables at the root level from the module input. These values are configured by practitioners within the task block. Value assignments are parsed to interpolate the corresponding variable declaration and are written to the appropriate Terraform files. A few assumptions are made for the intermediate variables: the variables users provide CTS are declared and supported by the module, matching name and type.

Module Guidelines

This section covers guidelines for authoring compatible CTS modules.

Scope

We recommend scoping the module to a few related resources for a provider. Small modules are easier and more flexible for end users to adopt for CTS. It allows them to iteratively combine different modules and use them as building blocks to meet their unique network infrastructure needs.

Complexity

Consider authoring modules with low complexity to reduce the run time for Terraform execution. Complex modules that have a large number of dependencies may result in longer runs, which adds delay to the near real time network updates.

Providers

The Terraform module must declare which providers it requires within the terraform.required_providers block. We suggest to also include a version constraint for the provider to specify which versions the module is compatible with.

Aside from the required_providers block, provider configurations should not be included within the sharable module for network integrations. End users will configure the providers through CTS, and CTS will then translate provider configuration to the generated root module appropriately.

Documentation

Modules for CTS are Terraform modules and can effectively run independently from the consul-terraform-sync daemon and Consul environment. They should be written and designed with Terraform best practices and should be clear to a Terraform user what the module does and how to use it. Module documentation should be named README or README.md. The description should capture what the module should be used for and the implications of running it in automation with CTS.