Infrastructure Plugins

AWS CDK synthesis/deployment and pipeline utility plugins.

CDK Synthesis

Plugin	Purpose	Compute	Secrets	Key Env Vars
cdk-synth	Synthesize CDK app to CloudFormation	MEDIUM	None (AWS IAM)	CDK_DEFAULT_REGION, CDK_DEFAULT_ACCOUNT, RESOLVED_SYNTH_PLUGIN

For CDK deployment plugins (cdk-deploy, cdk-deploy-multi-region), see Deploy Plugins.

Pipeline Utilities

Plugin	Purpose	Compute	Secrets	Key Env Vars
manual-approval	Native CDK approval gate that pauses the pipeline and waits for confirmation in the AWS CodePipeline console	SMALL	None	APPROVAL_COMMENT
manual-approval-custom	Approval gate that publishes an SNS notification, then polls for approval with a configurable timeout	SMALL	None (AWS IAM)	APPROVAL_TOPIC_ARN, APPROVAL_MESSAGE, APPROVAL_TIMEOUT, APPROVAL_URL
s3-cache	S3 build cache with zstd compression	SMALL	None (AWS IAM)	CACHE_BUCKET, CACHE_KEY, CACHE_PATHS, CACHE_ACTION
shell	Run arbitrary OS install/build commands on a clean Ubuntu 24.04 step (git/curl/jq/wget/apt) — the escape hatch when no language/tool plugin fits	SMALL	None	shellInstall, shellCommand, shellOutputDir

CDK Workflow

flowchart LR
    Synth[cdk-synth] --> CloudAssembly[cdk.out/]
    CloudAssembly --> SelfMutate[Self-Mutation]
    CloudAssembly --> Stages[Pipeline Stages]

Multi-Region Strategies

These strategies apply to the cdk-deploy-multi-region plugin. The CDK_DEPLOY_STRATEGY env var controls how stacks are deployed across regions:

Sequential

flowchart LR
    Start([Start]) --> Primary[Deploy Primary Region]
    Primary -->|Success| R2[Deploy Region 2]
    R2 -->|Success| R3[Deploy Region 3]
    R3 -->|Success| Done([Complete])
    Primary -->|Failure| Rollback1[Rollback Primary]
    R2 -->|Failure| Rollback2[Rollback Region 2]
    R3 -->|Failure| Rollback3[Rollback Region 3]
    Rollback1 --> Failed([Failed])
    Rollback2 --> Failed
    Rollback3 --> Failed

Deploys to each region one at a time in the order specified by CDK_REGIONS. If a deployment fails in any region, subsequent regions are skipped. This is the safest strategy and is recommended for production workloads.

Parallel

flowchart LR
    Start([Start]) --> Primary[Deploy Primary Region]
    Primary -->|Success| Fork{Fork}
    Fork --> R2[Deploy Region 2]
    Fork --> R3[Deploy Region 3]
    Fork --> R4[Deploy Region N]
    R2 --> Join{Join}
    R3 --> Join
    R4 --> Join
    Join -->|All Succeed| Done([Complete])
    Join -->|Any Failure| RollbackFailed[Rollback Failed Regions]
    RollbackFailed --> Partial([Partial - Successful Regions Kept])
    Primary -->|Failure| Failed([Failed])

Deploys to all regions simultaneously. Faster than sequential but provides less isolation between regions. Best suited for non-production environments or stateless workloads where region-level failures do not cascade.

Primary Region Canary Pattern

When CDK_PRIMARY_REGION is set, the deployment always starts with the primary region first regardless of the chosen strategy. Once the primary region deployment succeeds, the remaining regions proceed according to the selected strategy (sequential or parallel). This allows the primary region to serve as a canary, catching issues before they propagate to all regions.

Rollback on Failure

When CDK_ROLLBACK_ON_FAILURE=true, a failed deployment in any region triggers an automatic rollback of that region to the previous known-good state. In sequential mode, this also prevents deployment to subsequent regions. In parallel mode, regions that have already completed successfully are not rolled back – only the failed region is reverted.