// References cold start optimization history and techniques. Covers Docker image optimization, Fargate CPU tuning, SOCI lazy loading, and Lambda Bedrock discovery fix. Use when optimizing startup performance.
References Serverless OpenClaw architecture. Covers network design, CDK stack structure, DynamoDB data model, API protocols, and container design.
Loads Serverless OpenClaw project context. Provides background knowledge needed for development including project overview, tech stack, architecture decisions, and data models.
References Serverless OpenClaw cost optimization guidelines. Validates that no resources exceeding the cost target ($1/month) are created during implementation. Use when writing CDK stacks or making infrastructure changes.
Development workflow guide. Covers build, test, lint, format commands, package structure, Git hooks, TDD methodology, and coding conventions. Use when setting up dev environment, running tests, or following project conventions.
Guides Phase 1 MVP implementation steps. Pass a specific step number as an argument to get the goals, deliverables, validation criteria, and detailed design for that step.
Guides Phase 2 Lambda Container Migration steps. Pass a specific step number (2-1 through 2-5) to get the goals, deliverables, validation criteria, and detailed design for that step.
| name | cold-start |
| description | References cold start optimization history and techniques. Covers Docker image optimization, Fargate CPU tuning, SOCI lazy loading, and Lambda Bedrock discovery fix. Use when optimizing startup performance. |
| allowed-tools | Read, Glob, Grep |
Full analysis: docs/cold-start-optimization.md
| Target | Before | After | Savings |
|---|---|---|---|
| Docker image size | 2.22 GB | 1.27 GB | 43% reduction |
| Fargate cold start | ~120s (0.25 vCPU) | ~40-57s (1 vCPU) | ~80s |
| Lambda cold start | 56s | 1.35s | ~54s |
make cold-start # Full measurement (stops tasks + cleans TaskState first)
make cold-start-warm # Warm start measurement (skip idle wait)
make task-status # Verify task state before measuring
make task-stop # Stop all tasks
make task-stop-recent # Stop only most recent task
Important: Always use make cold-start ā never run the measurement script directly without stopping existing tasks first.
| Optimization | Savings | Technique |
|---|---|---|
| Remove AWS CLI | -358 MB | Replaced aws s3 sync with @aws-sdk/client-s3 Node.js code (s3-sync.ts) |
| chown layer optimization | -134 MB | COPY --chown=openclaw:openclaw instead of separate RUN chown |
| Pre-generated openclaw config | ā | Handled in Dockerfile (skip runtime onboard) |
| start-openclaw.sh simplification | ā | Removed S3 restore/onboard/sleep 2 |
| zstd compression (P6) | -16% (258ā217 MB) | --compression zstd in docker buildx |
./scripts/deploy-image.sh --soci # Linux only, ~50% additional pull-time reduction
SOCI CI: .github/workflows/deploy-image.yml
| CPU | Memory | Cold Start |
|---|---|---|
| 0.25 vCPU | 512 MB | OOM Kill |
| 0.25 vCPU | 2048 MB | ~120s |
| 0.5 vCPU | 2048 MB | ~80s |
| 1 vCPU | 2048 MB | ~40-57s (default) |
| 2 vCPU | 4096 MB | Configurable via env |
Configure via env vars: FARGATE_CPU, FARGATE_MEMORY
Promise.all (~3-5s saved)Root cause: Bedrock auto-discovery scanned all AWS regions on startup ā 56s delay
Fix: bedrockDiscovery.enabled: false in OpenClaw config ā 1.35s cold start
{
"bedrockDiscovery": { "enabled": false }
}
runEmbeddedPiAgent() ā bypasses Gateway WS server (no :18789 socket bind)extensionAPI.js via file:// URL (not in exports map)| Version | Cold Start | Status |
|---|---|---|
| v2026.2.9 | 65.3s | Working |
| v2026.2.13 | 57.9s | Working ā pinned |
| v2026.2.14 | N/A | BROKEN (scope system) |
Pin location: packages/container/Dockerfile
ARG OPENCLAW_VERSION=2026.2.13
Configurable via build-arg or env var.
now < prewarmUntil