Aws Codepipeline

CodePipeline is AWS's native CI/CD orchestration — deeply integrated with CodeBuild (build), CodeDeploy (deploy), ECS (container deployment), Lambda (serverless), and GitHub/BitBucket (source). For agents: StartPipelineExecution triggers a pipeline run on-demand, GetPipelineState monitors current stage status, ListPipelineExecutions queries execution history, and custom action providers enable integrating external tools (agents can implement a custom action that CodePipeline polls). Confidence is docs-derived.

CodePipeline API via boto3 (Python). cp.start_pipeline_execution(name="my-pipeline") triggers a pipeline run — returns pipelineExecutionId. cp.get_pipeline_state(name="my-pipeline") retrieves current stage statuses (inProgress, succeeded, failed). cp.get_pipeline_execution(pipelineName, pipelineExecutionId) retrieves execution details. cp.list_pipeline_executions(pipelineName, maxResults=10) lists recent executions. cp.stop_pipeline_execution(pipelineName, pipelineExecutionId, abandon=True) stops execution. cp.put_job_success_result(jobId, currentRevision, executionDetails) signals custom action success (for custom action providers). cp.put_job_failure_result(jobId, failureDetails) signals failure. cp.list_pipelines() lists all pipelines.

IAM-based auth: SigV4 via boto3. IAM roles: codepipeline:StartPipelineExecution, codepipeline:GetPipelineState, codepipeline:ListPipelineExecutions, codepipeline:GetPipelineExecution, codepipeline:StopPipelineExecutionRun. HTTPS enforced. No OAuth2. Pipeline service role: IAM role assumed by CodePipeline to access CodeBuild, S3, CodeDeploy, ECS, etc. — requires cross-service permissions. Custom actions: worker polls cp.poll_for_jobs() — must have codepipeline:PollForJobs + codepipeline:PutJobSuccessResult/PutJobFailureResult.

Execution is async: start_pipeline_execution returns immediately — poll get_pipeline_state for progress. Concurrent executions: by default only one active execution per pipeline — additional starts are queued. Stage transitions: can be enabled/disabled via enable_stage_transition/disable_stage_transition. Artifact storage: S3 bucket managed by CodePipeline — access artifacts from S3 after successful build. CloudWatch Events: pipeline state changes emit EventBridge events — trigger automation without polling. CodePipeline uptime at status.aws.amazon.com. Manual approval actions: human approval gate in pipeline — programmatically approve via put_approval_result.

docs.aws.amazon.com/codepipeline covers API reference, pipeline creation guide, custom action integration, and CloudWatch Events documentation. Getting started: AWS free tier (1 active pipeline/month), create pipeline via Console, boto3 for programmatic control. boto3: pip install boto3. AWS CLI: aws codepipeline start-pipeline-execution. Terraform: aws_codepipeline resource. Community via AWS re:Post and Stack Overflow (aws-codepipeline tag). Mature service with good documentation; well-integrated with the broader AWS DevOps toolchain.