# auto-deploy

**Repository Path**: boranli_admin/auto-deploy

## Basic Information

- **Project Name**: auto-deploy
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-07-13
- **Last Updated**: 2026-07-14

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# auto-deploy

`auto-deploy` is a simplified Huawei Cloud single-ECS web app deployment skill.
It is scoped to one-machine deployments. Clustered, high-availability, and
multi-resource production architectures are intentionally outside this skill.

It is intentionally optimized for the shortest successful path:

1. Analyze the repo.
2. Generate a small deployment plan.
3. Provision one ECS with bundled Terraform.
4. Upload source code over SSH.
5. Start the app and verify health.

## Main Command

```bash
python scripts/auto_deploy.py --repo /path/to/app --app-name demo-api --region cn-north-4 --yes
```

## Common Variants

Plan only:

```bash
python scripts/auto_deploy.py --repo /path/to/app --app-name demo-api --region cn-north-4 --plan-only
```

Validate Terraform only:

```bash
python scripts/auto_deploy.py --repo /path/to/app --app-name demo-api --region cn-north-4 --validate-only --skip-preflight
```

Reuse an existing VPC/subnet:

```bash
python scripts/auto_deploy.py --repo /path/to/app --app-name demo-api --network-mode reuse --existing-vpc-id <vpc-id> --existing-subnet-id <subnet-id> --yes
```

Restrict ports:

```bash
python scripts/auto_deploy.py --repo /path/to/app --app-name demo-api --app-port 8081 --no-http --no-https --public-cidr 203.0.113.10/32 --ssh-cidr 198.51.100.7/32 --yes
```

Resume a failed deploy after infrastructure already exists:

```bash
python scripts/auto_deploy.py --repo /path/to/app --app-name demo-api --workdir /path/to/app/.auto-deploy --resume-from deploy --skip-prepare --skip-upload
```

Docker cold builds can take longer on first deploy because large base images are
pulled through the ECS EIP. Docker and Docker Compose deployments default to a
`200 Mbps` EIP bandwidth recommendation, while native deployments keep the
lighter `5 Mbps` default. Docker build timeout defaults to `2400` seconds and can
be overridden:

```bash
python scripts/auto_deploy.py --repo /path/to/app --app-name demo-api --docker-build-timeout 3600 --yes
```

Inject application configuration without uploading `.env`:

```bash
python scripts/auto_deploy.py --repo /path/to/app --app-name demo-api --env DATABASE_URL=postgres://... --env JWT_SECRET=... --yes
python scripts/auto_deploy.py --repo /path/to/app --app-name demo-api --env-file ./deploy.env --yes
```

If analysis detects required variables and `--yes` is used, missing keys fail
fast. Use `--allow-missing-env` only when you accept the deployment risk.

## Supported Deploy Paths

- `native`: preferred for recognizable Node.js, Python, Java, Go, Rust, and .NET web apps.
- `docker-compose`: used when the repo explicitly ships a Compose app.
- `docker`: used only when the repo is container-first and no clear native runtime was detected.

Frontend/backend split projects should use Docker Compose. Native split
deployment is blocked by default; `--allow-native-fullstack` is an explicit
override for simple projects.

Removed from the main path:

- Artifact-native fallback
- Static-only special deployer
- Kubernetes deployer
- Clustered or multi-ECS deployment
- RDS/Redis/OBS/MQ/ELB/CCE/Kubernetes stacks

Prepare external databases, Redis, queues, or object storage separately, then
inject their connection values with `--env` or `--env-file`. Reports and summary
files show env key names only, never values.

SSH ingress defaults to the detected control-machine public IP `/32`. If public
IP detection fails, provide `--ssh-cidr x.x.x.x/32`; use
`--ssh-cidr 0.0.0.0/0` only when intentionally opening SSH to the internet.

## Why It Is Faster Now

- Preflight runs after analysis and skips Docker Hub checks for native apps.
- SSH keys are reused and generated as `ed25519`, not regenerated as timestamped `RSA 4096`.
- Region/flavor/image resolution no longer runs as a separate workflow stage.
- VPC conflict handling uses a direct Huawei Cloud API precheck before apply.
- Deploy no longer does duplicate Docker pre-install plus fallback orchestration.
- Failed deploy recovery can resume from `deploy` or `verify` without rerunning Terraform.
- Docker build timeouts retry once to reuse partially downloaded layers.
- Previous timing/report files are archived as `*-attempt-N.*` before a resume overwrites canonical outputs.

## Outputs

Each run writes under `<repo>/.auto-deploy`:

- `analysis.json`
- `resource-spec.json`
- `deployment-plan.md`
- `timings.json`
- `run-summary.json`
- `deployment-test-report.md`

## Cleanup

```bash
python scripts/destroy.py --workdir .auto-deploy/terraform
```

Destroy only through the recorded Terraform workdir.