CodePipeline concepts - Amazon CodePipeline
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

CodePipeline concepts

Modeling and configuring your automated release process is easier if you understand the concepts and terms used in Amazon CodePipeline. Here are some concepts to know about as you use CodePipeline.

For an example of a DevOps pipeline, see DevOps pipeline example.

The following terms are used in CodePipeline:


A pipeline is a workflow construct that describes how software changes go through a release process. Each pipeline is made up of a series of stages.


A stage is a logical unit you can use to isolate an environment and to limit the number of concurrent changes in that environment. Each stage contains actions that are performed on the application artifacts. Your source code is an example of an artifact. A stage might be a build stage, where the source code is built and tests are run. It can also be a deployment stage, where code is deployed to runtime environments. Each stage is made up of a series of serial or parallel actions.


An action is a set of operations performed on application code and configured so that the actions run in the pipeline at a specified point. This can include things like a source action from a code change, an action for deploying the application to instances, and so on. For example, a deployment stage might contain a deployment action that deploys code to a compute service like Amazon EC2 or Amazon Lambda.

Valid CodePipeline action types are source, build, test, deploy, approval, and invoke. For a list of action providers, see Valid action types and providers in CodePipeline .

Actions can run in series or in parallel. For information about serial and parallel actions in a stage, see the runOrder information in action structure requirements.

Pipeline executions

An execution is a set of changes released by a pipeline. Each pipeline execution is unique and has its own ID. An execution corresponds to a set of changes, such as a merged commit or a manual release of the latest commit. Two executions can release the same set of changes at different times.

While a pipeline can process multiple executions at the same time, a pipeline stage processes only one execution at a time. To do this, a stage is locked while it processes an execution. Two pipeline executions can't occupy the same stage at the same time. The execution waiting to enter the occupied stage is referred to an inbound execution. An inbound execution can still fail, be superseded, or be manually stopped. For more information about how inbound executions work, see How Inbound Executions Work.

Pipeline executions traverse pipeline stages in order. Valid statuses for pipelines are InProgress, Stopping, Stopped, Succeeded, Superseded, and Failed.

For more information, see PipelineExecution.

Stopped executions

The pipeline execution can be stopped manually so that the in-progress pipeline execution does not continue through the pipeline. If stopped manually, a pipeline execution shows a Stopping status until it is completely stopped. Then it shows a Stopped status. A Stopped pipeline execution can be retried.

There are two ways to stop a pipeline execution:

  • Stop and wait

  • Stop and abandon

For information about use cases for stopping an execution and sequence details for these options, see How pipeline executions are stopped.

Failed executions

If an execution fails, it stops and does not completely traverse the pipeline. Its status is FAILED status and the stage is unlocked. A more recent execution can catch up and enter the unlocked stage and lock it. You can retry a failed execution unless the failed execution has been superseded or is not retryable. You can roll back a failed stage to a previous successful execution.

Execution modes

To deliver the latest set of changes through a pipeline, newer executions pass and replace less recent executions already running through the pipeline. When this occurs, the older execution is superseded by the newer execution. An execution can be superseded by a more recent execution at a certain point, which is the point between stages. SUPERSEDED is the default execution mode.

In SUPERSEDED mode, if an execution is waiting to enter a locked stage, a more recent execution might catch up and supersede it. The newer execution now waits for the stage to unlock, and the superseded execution stops with a SUPERSEDED status. When a pipeline execution is superseded, the execution is stopped and does not completely traverse the pipeline. You can no longer retry the superseded execution after it has been replaced at this stage. Other available execution modes are PARALLEL or QUEUED mode.

For more information about execution modes and locked stages, see How executions are processed in SUPERSEDED mode.

Stage operations

When a pipeline execution runs through a stage, the stage is in the process of completing all of the actions within it. For information about how stage operations work and information about locked stages, see How executions are processed in SUPERSEDED mode.

Valid statuses for stages are InProgress, Stopping, Stopped, Succeeded, and Failed. You can retry a failed stage unless the failed stage is not retryable. For more information, see StageExecution. You can roll back a stage to a specified previous successful execution. A stage can be configured to roll back automatically on failure as detailed in Configuring stage rollback. For more information, see RollbackStage.

Action executions

An action execution is the process of completing a configured action that operates on designated artifacts. These can be input artifacts, output artifacts, or both. For example, a build action might run build commands on an input artifact, such as compiling application source code. Action execution details include an action execution ID, the related pipeline execution source trigger, and the input and output artifacts for the action.

Valid statuses for actions are InProgress, Abandoned, Succeeded, or Failed. For more information, see ActionExecution.

Execution types

A pipeline or stage execution can be either a standard or a rolled-back execution.

For standard types, the execution has a unique ID and is a full pipeline run. A pipeline rollback has a stage to be rolled back and a successful execution for the stage as the target execution to which to roll back. The target pipeline execution is used to retrieve source revisions and variables for the stage to rerun.

Action types

Action types are preconfigured actions that are available for selection in CodePipeline. The action type is defined by its owner, provider, version, and category. The action type provides customized parameters that are used to complete the action tasks in a pipeline.

For information about the Amazon Web Services and third-party products and services you can integrate into your pipeline based on action type, see Integrations with CodePipeline action types.

For information about the integration models supported for action types in CodePipeline, see Integration model reference.

For information about how third-party providers can set up and manage action types in CodePipeline, see Working with action types.


A transition is the point where a pipeline execution moves to the next stage in the pipeline. You can disable a stage's inbound transition to prevent executions from entering that stage, and then you can enable the transition to allow executions to continue. When more than one execution arrives at a disabled transition, only the latest execution continues to the next stage when the transition is enabled. This means that newer executions continue to supersede waiting executions while the transition is disabled, and then after the transition is enabled, the execution that continues is the superseding execution.

A pipeline contains stages, which contain actions, which are separated by transitions that can be disabled and enabled.


Artifacts refers to the collection of data, such as application source code, built applications, dependencies, definitions files, templates, and so on, that is worked on by pipeline actions. Artifacts are produced by some actions and consumed by others. In a pipeline, artifacts can be the set of files worked on by an action (input artifacts) or the updated output of a completed action (output artifacts).

Actions pass output to another action for further processing using the pipeline artifact bucket. CodePipeline copies artifacts to the artifact store, where the action picks them up. For more information about artifacts, see Input and output artifacts.

Source revisions

When you make a source code change, a new version is created. A source revision is the version of a source change that triggers a pipeline execution. An execution processes source revisions. For GitHub and CodeCommit repositories, this is the commit. For S3 buckets or actions, this is the object version.

You can start a pipeline execution with a source revision, such as a commit, that you specify. The execution will process the specified revision and override what would have been the revision used for the execution. For more information, see Start a pipeline with a source revision override.


Triggers are events that start your pipeline. Some triggers, such as starting a pipeline manually, are available for all source action providers in a pipeline. Certain triggers depend on the source provider for a pipeline. For example, CloudWatch events must be configured with event resources from Amazon CloudWatch that have the pipeline ARN added as a target in the event rule. Amazon CloudWatch Events is the recommended triggers for automatic change detection for pipelines with a CodeCommit or S3 source action. Webhooks are a type of trigger configured for third-party repository events. For example, WebhookV2 is a trigger type that allows Git tags to be used to start pipelines with third-party source providers such as, GitHub Enterprise Server,, GitLab self-managed, or Bitbucket Cloud. In the pipeline configuration, you can specify a filter for triggers, such as push or pull request. You can filter code push events on Git tags, branches, or file paths. You can filter pull request events on event (opened, updated, closed), branches, or file paths.

For more information about triggers, see Start a pipeline in CodePipeline. For a tutorial that walks you through using Git tags as triggers for your pipeline, see Tutorial: Use Git tags to start your pipeline.


A variable is a value that can be used to dynamically configure actions in your pipeline. Variables can be either declared on the pipeline level, or emitted by actions in the pipeline. Variable values are resolved at the time of pipeline execution and can be viewed in the execution history. For variables declared at the pipeline level, you can either define default values in the pipeline configuration, or override them for a given execution. For variables emitted by an action, the value is available after an action succesfully completes. For more information, see Variables.