Machine-Readable UI
Note: This format is available in Terraform 0.15.3 and later.
By default, many Terraform commands display UI output as unstructured text, intended to be read by a user via a terminal emulator. This text stream is not a stable interface for integrations. Some commands support a -json flag, which enables a structured JSON output mode with a defined interface.
For long-running commands such as plan, apply, refresh, and test, the -json flag outputs a stream of JSON UI messages, one per line. These can be processed one message at a time, with integrating software filtering, combining, or modifying the output as desired.
The first message output has type version, and includes a ui key, which as of Terraform 1.1.0 has
value "1.0". The semantics of this version are:
- We will increment the minor version, e.g.
"1.1", for backward-compatible changes or additions. Ignore any object properties with unrecognized names to remain forward-compatible with future minor versions. - We will increment the major version, e.g.
"2.0", for changes that are not backward-compatible. Reject any input which reports an unsupported major version.
We will introduce new major versions only within the bounds of the Terraform 1.0 Compatibility Promises.
Sample JSON Output
Below is sample output from running terraform apply -json:
Each line consists of a JSON object with several keys common to all messages. These are:
@level: this is normally "info", but can be "error" or "warn" when showing diagnostics@message: a human-readable summary of the contents of this message@module: always "terraform.ui" when rendering UI output@timestamp: an RFC3339 timestamp of when the message was outputtype: defines which kind of message this is and determines how to interpret other keys which may be present
Clients presenting the logs as a user interface should handle unexpected message types by presenting at least the @message field to the user.
Messages will be emitted as events occur to trigger them. This means that messages related to several resources may be interleaved (if Terraform is running with concurrency above 1). The resource object value can be used to link multiple messages about a single resource.
Message Types
The following message types are supported:
Generic Messages
version: information about the Terraform version and the version of the schema used for the following messageslog: unstructured human-readable log linesdiagnostic: diagnostic warning or error messages; see theterraform validatedocs for more details on the format
Operation Results
resource_drift: describes a detected change to a single resource made outside of Terraformplanned_change: describes a planned change to a single resourcechange_summary: summary of all planned or applied changesoutputs: list of all root module outputs
Resource Progress
apply_start,apply_progress,apply_complete,apply_errored: sequence of messages indicating progress of a single resource through applyprovision_start,provision_progress,provision_complete,provision_errored: sequence of messages indicating progress of a single provisioner steprefresh_start,refresh_complete: sequence of messages indicating progress of a single resource through refresh
Test Results
test_abstract: describes the test operation that Terraform executestest_file: describes the status of a completed test filetest_run: describes the status of a completedrunblock within a test filetest_cleanup: describes the result of the test cleanup after a completed test filetest_summary: describes the overall status of the completed test operationtest_plan: describes the output of a givenrunblock whencommand = plantest_state: describes the output of a givenrunblock whencommand = applytest_interrupt: describes the result of an interrupted test operation
Version Message
A machine-readable UI command output will always begin with a version message. The following message-specific keys are defined:
terraform: the Terraform version which emitted this messageui: the machine-readable UI schema version defining the meaning of the following messages
Example
Resource Drift
If drift is detected during planning, Terraform will emit a resource_drift message for each resource which has changed outside of Terraform. This message has an embedded change object with the following keys:
resource: object describing the address of the resource to be changed; see resource object below for detailsaction: the action planned to be taken for the resource. Values:update,delete.
This message does not include details about the exact changes which caused the change to be planned. That information is available in the JSON plan output.
Example
Planned Change
At the end of a plan or before an apply, Terraform will emit a planned_change message for each resource which has changes to apply. This message has an embedded change object with the following keys:
resource: object describing the address of the resource to be changed; see resource object below for detailsprevious_resource: object describing the previous address of the resource, if this change includes a configuration-driven moveaction: the action planned to be taken for the resource. Values:noop,create,read,update,replace,delete,move.reason: an optional reason for the change, only used when the action isreplaceordelete. Values:tainted: resource was marked as taintedrequested: user requested that the resource be replaced, for example via the-replaceplan flagcannot_update: changes to configuration force the resource to be deleted and created rather than updateddelete_because_no_resource_config: no matching resource in configurationdelete_because_wrong_repetition: resource instance key has no correspondingcountorfor_eachin configurationdelete_because_count_index: resource instance key is outside the range of thecountargumentdelete_because_each_key: resource instance key is not included in thefor_eachargumentdelete_because_no_module: enclosing module instance is not in configuration
This message does not include details about the exact changes which caused the change to be planned. That information is available in the JSON plan output.
Example
Change Summary
Terraform outputs a change summary when a plan or apply operation completes. Both message types include a changes object, which has the following keys:
add: count of resources to be created (including as part of replacement)change: count of resources to be changed in-placeremove: count of resources to be destroyed (including as part of replacement)operation: one ofplan,apply, ordestroy
Example
Outputs
After a successful plan or apply, a message with type outputs contains the values of all root module output values. This message contains an outputs object, the keys of which are the output names. The outputs values are objects with the following keys:
action: for planned outputs, the action which will be taken for the output. Values:noop,create,update,deletevalue: for applied outputs, the value of the output, encoded in JSONtype: for applied outputs, the detected HCL type of the output valuesensitive: boolean value,trueif the output is sensitive and should be hidden from UI by default
Note that sensitive outputs still include the value field, and integrating software should respect the sensitivity value as appropriate for the given use case.
Example
Operation Messages
Performing Terraform operations to a resource will often result in several messages being emitted. The message types include:
apply_start: when starting to apply changes for a resourceapply_progress: periodically, showing elapsed time outputapply_complete: on successful operation completionapply_errored: when an error is encountered during the operationprovision_start: when starting a provisioner stepprovision_progress: on provisioner outputprovision_complete: on successful provisioningprovision_errored: when an error is enountered during provisioningrefresh_start: when reading a resource during refreshrefresh_complete: on successful refresh
Each of these messages has a hook object, which has different fields for each type. All hooks have a resource object which identifies which resource is the subject of the operation.
Apply Start
The apply_start message hook object has the following keys:
resource: aresourceobject identifying the resourceaction: the action to be taken for the resource. Values:noop,create,read,update,replace,deleteid_keyandid_value: a key/value pair used to identify this instance of the resource, omitted when unknown
Example
Apply Progress
The apply_progress message hook object has the following keys:
resource: aresourceobject identifying the resourceaction: the action being taken for the resource. Values:noop,create,read,update,replace,deleteelapsed_seconds: time elapsed since the apply operation started, expressed as an integer number of seconds
Example
Apply Complete
The apply_complete message hook object has the following keys:
resource: aresourceobject identifying the resourceaction: the action taken for the resource. Values:noop,create,read,update,replace,deleteid_keyandid_value: a key/value pair used to identify this instance of the resource, omitted when unknownelapsed_seconds: time elapsed since the apply operation started, expressed as an integer number of seconds
Example
Apply Errored
The apply_complete message hook object has the following keys:
resource: aresourceobject identifying the resourceaction: the action taken for the resource. Values:noop,create,read,update,replace,deleteelapsed_seconds: time elapsed since the apply operation started, expressed as an integer number of seconds
The exact detail of the error will be rendered as a separate diagnostic message.
Example
Provision Start
The provision_start message hook object has the following keys:
resource: aresourceobject identifying the resourceprovisioner: the type of provisioner
Example
Provision Progress
The provision_progress message hook object has the following keys:
resource: aresourceobject identifying the resourceprovisioner: the type of provisioneroutput: the output log from the provisioner
One provision_progress message is output for each log line received from the provisioner.
Example
Provision Complete
The provision_complete message hook object has the following keys:
resource: aresourceobject identifying the resourceprovisioner: the type of provisioner
Example
Provision Errored
The provision_errored message hook object has the following keys:
resource: aresourceobject identifying the resourceprovisioner: the type of provisioner
Example
Refresh Start
The refresh_start message hook object has the following keys:
resource: aresourceobject identifying the resourceid_keyandid_value: a key/value pair used to identify this instance of the resource
Example
Refresh Complete
The refresh_complete message hook object has the following keys:
resource: aresourceobject identifying the resourceid_keyandid_value: a key/value pair used to identify this instance of the resource
Example
Resource Object
The resource object is a decomposed structure representing a resource address in configuration, which is used to identify which resource a given message is associated with. The object has the following keys:
addr: the full unique address of the resource as a stringmodule: the address of the module containing the resource, in the formmodule.foo.module.bar, or an empty string for a root module resourceresource: the module-relative address, which is identical toaddrfor root module resourcesresource_type: the type of resource being addressedresource_name: the name label for the resourceresource_key: the address key (countorfor_eachvalue), ornullif the neither are usedimplied_provider: the provider type implied by the resource type; this may not reflect the resource's provider if provider aliases are used
Example
Test Abstract
The terraform test command emits a test_abstract message describing the test files and run blocks that Terraform executes during the upcoming operation.
Example
Test File
Throughout a terraform test execution, Terraform will produce several progress updates for each test file. The progress field can be starting, teardown, or complete. Each test file will emit each progress update exactly once. When a test file emits the complete progress update, it will also include a status field containing one of pass, error, fail, or skip denoting the overall status of the completed test file.
Example
Test Run
While executing run blocks within a test file, Terraform will also produce several status updates for each run block. The progress field for a run block progress update can be starting, running, teardown, and complete. The starting and complete progress updates will be emitted exactly once. While the running and teardown progress updates can be emitted multiple times.
The starting, running and teardown updates will also include an elapsed field indicating the number of milliseconds the current test operation (for starting and running) or the current destroy operation (for teardown) has been in progress.
The complete progress update will also include a status field containing one of pass, error, fail, or skip denoting the overall status of the completed run block`.
Not every run block will emit teardown progress updates, as only the most recently executed run blocks reference the latest in-memory state files that need to be torn down. In addition, the run block tear down process is only initiated once the overall test file has already emitted its teardown status update. This means you can expect the complete progress update to be issued for a run block before any teardown updates are provided. There will always be a complete progress update issued by the enclosing test file when the tear down process for all run blocks is complete.
Examples
Test Cleanup
After Terraform completes the execution of each test file, terraform test may emit a series of test_cleanup messages detailing any state it could not destroy. You must locate and destroy the resources listed in these resources manually.
As the test framework can manage multiple state files for each test file, you can see multiple versions of this message for each state file that holds resources that Terraform could not destroy. Using the @testrun field, you can pinpoint the run block that last updated the state file to locate the resources that Terraform could not destroy.
Example
Test Summary
After the test operation has completed all test files, terraform test emits a test_summary message with the status of the overall test operation.
Example
Test Plan
In -verbose mode, terraform test emits a test_plan message for all run blocks that executed a plan operation. The test_plan message contains a subset of the attributes from the terraform show -json <planfile> command output and the terraform providers schema -json command output.
The message contains the following fields representing the plan: plan_format_version, output_changes, resource_changes, resource_drift and relevant_attributes. These match the respective fields in the JSON Plan Representation.
The message contains the following fields representing the provider schemas: provider_format_version and provider_schemas. These match the respective fields in the JSON Providers Schema Representation.
Example
Test State
In -verbose mode, terraform test emits a test_state message for all run blocks that executed an apply operation. The test_state message contains a subset of the terraform show -json command output and the terraform providers schema -json command output.
The message contains the following fields representing the state: state_format_version, root_module, and outputs. These match the respective fields in the JSON Values Representation that are embedded in complete JSON state representation.
The message contains the following fields representing the provider schemas: provider_format_version and provider_schemas. These match the respective fields in the JSON Providers Schema Representation.
Example
Test Interrupt
When terraform test is impeded or canceled, it emits a test_interrupt message that describes the interrupted operation. This message includes any state that Terraform could not destroy before exiting and any leftover resources that Terraform did not finish creating.
As with the Test Cleanup message, Terraform might have been maintaining multiple state files. The state field contains the resources from the main state for the test configuration, while the states field contains a map of run block names to the resources that each run block created, which Terraform could not destroy.
Finally, the planned field contains any resources that were in the process of being created by the interrupted run block, which you can identify using the @testrun field.
Example