19 KiB

Raw Blame History

layout	page_title	sidebar_current	description
docs	Configuring Resources	docs-config-resources	Resources are the most important element in a Terraform configuration. Each resource corresponds to an infrastructure object, such as a virtual network or compute instance.

Resources

Resources are the most important element in the Terraform language. Each resource block describes one ore more infrastructure objects, such as virtual networks, compute instances, or higher-level components such as DNS records.

Resource Syntax

Resource declarations can include a number of advanced features, but only a small subset are required for initial use. More advanced syntax features, such as single resource declarations that produce multiple similar remote objects, are described later in this page.

resource "aws_instance" "web" {
  ami           = "ami-a1b2c3d4"
  instance_type = "t2.micro"
}

A resource block declares a resource of a given type ("aws_instance") with a given local name ("web"). The name is used to refer to this resource from elsewhere in the same Terraform module, but has no significance outside of the scope of a module.

The resource type and name together serve as an identifier for a given resource and so must be unique within a module.

Within the block body (between { and }) are the configuration arguments for the resource itself. Most arguments in this section depend on the resource type, and indeed in this example both ami and instance_type are arguments defined specifically for the aws_instance resource type.

Resource Types and Arguments

Each resource is associated with a single resource type, which determines the kind of infrastructure object it manages and what arguments and other attributes are supported for each resource.

Each resource type in turn belongs to a provider, which is a plugin for Terraform that offers a collection of resource types that most often belong to a single cloud or on-premises infrastructure platform.

Most of the items within the body of a resource block are defined by and specific to the selected resource type, and these arguments can make full use of expressions and other dynamic Terraform language features.

However, there are some "meta-arguments" that are defined by Terraform itself and apply across all resource types. These arguments often have additional restrictions on what language features can be used with them, and are described in more detail in the following sections.

Resource Behavior

A resource block describes your intent for a particular infrastructure object to exist with the given settings. If you are writing a new configuration for the first time, the resources it defines will exist only in the configuration, and will not yet represent real infrastructure objects in the target platform.

Applying a Terraform configuration is the process of creating, updating, and destroying real infrastructure objects in order to make their settings match the configuration.

When Terraform creates a new infrastructure object represented by a resource block, the identifier for that real object is saved in Terraform's state, allowing it to be updated and destroyed in response to future changes. For resource blocks that already have an associated infrastructure object in the state, Terraform compares the actual configuration of the object with the arguments given in the configuration and, if necessary, updates the object to match the configuration.

This general behavior applies for all resources, regardless of type. The details of what it means to create, update, or destroy a resource are different for each resource type, but this standard set of verbs is common across them all.

The "meta-arguments" within resource blocks, defined in the following sections, allow some details of this standard resource behavior to be customized on a per-resource basis.

Resource Dependencies

As with other elements in the Terraform language, Terraform analyses any expressions within a resource block to find references to other objects, and infers from this a correct dependency ordering for creating, updating, or destroying each resource. Because of this, in most cases it is not necessary to mention explicitly any dependencies between resources.

However, in some less-common situations there are dependencies between resources that cannot be recognized implicitly in configuration. For example, if Terraform is being used to both manage access control policies and take actions that require those policies to be present, there may be a hidden dependency between the access policy and a resource whose creation depends on it.

In these rare cases, the depends_on meta-argument can be used to explicitly specify a dependency. This argument is available in all resource blocks, regardless of resource type. For example:

resource "aws_iam_role" "example" {
  name = "example"

  # assume_role_policy is omitted for brevity in this example. See the
  # documentation for aws_iam_role for a complete example.
  assume_role_policy = "..."
}

resource "aws_iam_instance_profile" "example" {
  # Because this expression refers to the role, Terraform can infer
  # automatically that the role must be created first.
  role = aws_iam_role.example.name
}

resource "aws_iam_role_policy" "example" {
  name   = "example"
  role   = aws_iam_role.example.name
  policy = jsonencode({
    "Statement" = [{
      # This policy allows software running on the EC2 instance to
      # access the S3 API.
      "Action" = "s3:*",
      "Effect" = "Allow",
    }],
  })
}

resource "aws_instance" "example" {
  ami           = "ami-a1b2c3d4"
  instance_type = "t2.micro"

  # Terraform can infer from this that the instance profile must
  # be created before the EC2 instance.
  iam_instance_profile = aws_iam_instance_profile.example

  # However, if software running in this EC2 instance needs access
  # to the S3 API in order to boot properly, there is also a "hidden"
  # dependency on the aws_iam_role_policy that Terraform cannot
  # automatically infer, so it must be declared explicitly:
  depends_on = [
    aws_iam_role_policy.example,
  ]
}

The depends_on meta-argument, if present, must be a list of references to other resources in the same module. Arbitrary expressions are not allowed in the depends_on argument value, because its value must be known before Terraform knows resource relationships and thus before it can safely evaluate expressions.

The depends_on argument should be used only as a last resort. When using it, always include a comment explaining why it is being used, to help future maintainers understand the purpose of the additional dependency.

Multiple Resource Instances

By default, a single resource block corresponds to only one real infrastructure object. Sometimes it is desirable to instead manage a set of similar objects of the same type, such as a fixed pool of compute instances. You can achieve this by using the count meta-argument, which is allowed in all resource blocks:

resource "aws_instance" "server" {
  count = 4 # create four similar EC2 instances

  ami           = "ami-a1b2c3d4"
  instance_type = "t2.micro"

  tags {
    Name = "Server ${count.index}"
  }
}

When the count meta-argument is present, a distinction exists between the resource block itself -- identified as aws_instance.server -- and the multiple resource instances associated with it, identified as aws_instance.server[0], aws_instance.server[1], etc. When count is not present, a resource block has only a single resource instance, which has no associated index.

For resource blocks where count is set, an additional count object is available for use in expressions, which has an attribute count.index that provides the distinct index for each instance.

The Resource Behavior section above described how each resource corresponds to a real infrastructure object. It is in fact resource instances that correspond to infrastructure objects, and so when count is used a particular resource block has a distinct infrastructure object associated with each of its instances, and each is separtely created, updated, or destroyed when the configuration is applied.

The count meta argument accepts expressions in its value, similar to the resource-type-specific arguments for a resource. However, Terraform must interpret the count argument before any actions are taken from remote resources, and so (unlike the resource-type-specifc arguments) the count expressions may not refer to any resource attributes that are not known until after a configuration is applied, such as a unique id generated by the remote API when an object is created.

For example, count can be used with an input variable that carries a list value, to create one instance for each element of the list:

variable "subnet_ids" {
  type = list(string)
}

resource "aws_instance" "server" {
  # Create one instance for each subnet
  count = length(var.subnet_ids)

  ami           = "ami-a1b2c3d4"
  instance_type = "t2.micro"
  subnet_id     = var.subnet_ids[count.index]

  tags {
    Name = "Server ${count.index}"
  }
}

Note that the separate resource instances created by count are still identified by their index, and not by the string values in the given list. This means that if an element is removed from the middle of the list, all of the indexed instances after it will see their subnet_id values change, which will cause more remote object changes than were probably intended. The practice of generating multiple instances from lists should be used sparingly, and with due care given to what will happen if the list is changed later.

Selecting a Non-default Provider Configuration

As described in the providers guide, Terraform optionally allows the definition of multiple alternative ("aliased") configurations for a single provider, to allow management of resources in different regions in multi-region services, etc. The provider meta-argument overrides Terraform's default behavior of selecting a provider configuration based on the resource type name.

By default, Terraform takes the initial word in the resource type name (separated by underscores) and selects the default configuration for that named provider. For example, the resource type google_compute_instance is associated automatically with the default configuration for the provider named google.

By using the provider meta-argument, an aliased provider configuration can be selected:

# default configuration
provider "google" {
  region = "us-central1"
}

# alternative, aliased configuration
provider "google" {
  alias  = "europe"
  region = "europe-west1"
}

resource "google_compute_instance" "example" {
  # This "provider" meta-argument selects the google provider
  # configuration whose alias is "europe", rather than the
  # default configuration.
  provider = google.europe

  # ...
}

A resource always has an implicit dependency on its associated provider, to ensure that the provider is fully configured before any resource actions are taken.

The provider meta-argument value must always be a literal provider name followed by an alias name separated by a dot. Arbitrary expressions are not permitted for provider because it must be resolved while Terraform is constructing the dependency graph, before it is safe to evaluate expressions.

Lifecycle Customizations

The general lifecycle for resources is described above in the section Resource Behavior. Some details of that behavior can be customized using the special nested block lifecycle within a resource block body:

resource "azurerm_resource_group" "example" {
  # ...

  lifecycle {
    create_before_destroy = true
  }
}

The lifecycle block and its contents are meta-arguments, available for all resource blocks regardless of type. The following lifecycle meta-arguments are supported:

create_before_destroy (bool) - By default, when Terraform must make a change to a resource argument that cannot be updated in-place due to remote API limitations Terraform will instead destroy the existing object and then create a new replacement object with the new configured arguments.

The create_before_destroy meta-argument changes this behavior so that the new, replacement object is created first, and then the prior object is destroyed only once the replacement is created.

This is an opt-in behavior because many remote object types have unique name requirements or other constraints that must be accommodated for both an new and an old object to exist concurrently. Some resource types offer special options to append a random suffix onto each object name to avoid collisions, for example. Terraform Core cannot automatically activate such features, so you must understand the constrants for each resource type before using create_before_destroy with it.
prevent_destroy (bool) - This meta-argument, when set to true, will cause Terraform to reject with an error any plan that would destroy the infrastructure object associated with the resource, as long as the argument remains present in the configuration.

This can be used as a measure of safety against the accidental replacement of objects that may be costly to reproduce, such as database instances. However, it will make certain configuration changes impossible to apply, and will prevent the use of the terraform destroy command once such objects are created, and so this option should be used sparingly.

Since this argument must be present in configuration for the protection to apply, note that this setting does not prevent the remote object from being destroyed if the resource block were removed from configuration entirely: in that case, the prevent_destroy setting is removed along with it, and so Terraform will allow the destroy operation to succeed.
ignore_changes (list of attribute names) - By default, Terraform detects any difference between the current settings of a real infrastructure object and plans to update the remote object to match configuration.

In some rare cases, settings of a remote object are modified by processes outside of Terraform, which Terraform would then attempt to "fix" on the next run. In order to make Terraform share management responsibilities of a single object with a separate process, the ignore_changes meta-argument specifies resource attributes that Terraform should ignore when planning updates to the associated remote object.

The arguments corresponding to the given attribute names are considered when planning a create operation, but are ignored when planning an update.
```
  resource "aws_instance" "example" {
    # ...

    lifecycle {
      ignore_changes = [
        # Ignore changes to tags, e.g. because a management agent
        # updates these based on some ruleset managed elsewhere.
        tags,
      ]
    }
  }
```
Instead of a list, the special keyword all may be used to instruct Terraform to ignore all attributes, which means that Terraform can create and destroy the remote object but will never propose updates to it.

Only attributes defined by the resource type can be ignored. ignore_changes cannot be applied to itself or to any other meta-arguments.

The lifecycle settings all effect how Terraform constructs and traverses the dependency graph. As a result, only literal values can be used because the processing happens to early for arbitrary expression evaluation.

Local-only Resources

While most resource types correspond to an infrastructure object type that is managed via a remote network API, there are certain specialized resource types that operate only within Terraform itself, calculating some results and saving those results in the state for future use.

For example, local-only resource types exist for generating private keys, issuing self-signed TLS certificates, and even generating random ids. While these resource types often have a more marginal purpose than those managing "real" infrastructure objects, they can be useful as glue to help connect together other resources.

The behavior of local-only resources is the same as all other resources, but their result data exists only within the Terraform state. "Destroying" such a resource means only to remove it from the state, discarding its data.

Operation Timeouts

Some resource types provide a special timeouts nested block argument that allows you to customize how long certain operations are allowed to take before being considered to have failed. For example, aws_db_instance allows configurable timeouts for create, update and delete operations.

Timeouts are handled entirely by the resource type implementation in the provider, but resource types offering these features follow the convention of defining a child block called timeouts that has a nested argument named after each operation that has a configurable timeout value. Each of these arguments takes a string representation of a duration, such as "60m" for 60 minutes, "10s" for ten seconds, or "2h" for two hours.

resource "aws_db_instance" "example" {
  # ...

  timeouts {
    create = "60m"
    delete = "2h"
  }
}

The set of configurable operations is chosen by each resource type. Most resource types do not support the timeouts block at all. Consult the documentation for each resource type to see which operations it offers for configuration, if any.

Resource Provisioners

Some infrastructure objects require some special actions to be taken after they are created before they can become fully functional. For example, compute instances may require configuration to be uploaded or a configuration management program to be run before they can begin their intended operation.

Create-time actions like these can be described using resource provisioners. A provisioner is another type of plugin supported by Terraform, and each provisioner takes a different kind of action in the context of a resource being created.

Provisioning steps should be used sparingly, since they represent non-declarative actions taken during the creation of a resource and so Terraform is not able to model changes to them as it can for the declarative portions of the Terraform language.

Provisioners can also be defined to run when a resource is destroyed, with certain limitations.

The provisioner and connection block types within resource blocks are meta-arguments available across all resource types. Provisioners and their usage are described in more detail in the Provisioners section.

19 KiB Raw Blame History