# Deploy Tidb AWS

> Deploy or manage TiDB clusters on AWS with Terraform and TiUP.

- Skill: `tenfyzhong/deploy-tidb-aws` (Agent Skill, multi-file: 16 files)
- Install (CLI): `npx skillmds@latest add tenfyzhong/deploy-tidb-aws`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tenfyzhong/deploy-tidb-aws/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: DevOps & Infra
- Author: tenfyzhong (https://skillmd.com/u/tenfyzhong)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/tenfyzhong/deploy-tidb-aws

---


# Deploy TiDB AWS

## Overview

Use this skill to create a self-contained Terraform project from the bundled template, provision or destroy AWS VMs with Terraform, and deploy, start, or validate TiDB with TiUP from the center VM.
Scaffolded projects default to `$HOME/test/tidb-aws-<namespace>` so later agents can find the deployment directory and operate Terraform after the original agent exits.
Each generated project includes a local `README.md` with Terraform operation commands.

Before running live infrastructure commands, read [references/deployment.md](references/deployment.md).

## Defaults

When `--namespace` is omitted, it defaults to `tidb-cluster-<YYMMDDHHMM>` (current local time).

When the prompt omits a count, use:

- `n_pd = 1`
- `n_tidb = 3`
- `n_tikv = 3`
- `n_tiflash = 1`
- `n_ticdc = 3`

Honor explicit counts from the prompt, including `0` for optional roles. Keep `n_pd = 1`; the template assumes one PD/monitoring/Grafana host.

Treat named extra service nodes as standalone EC2 instances, not TiUP cluster components. Only add extra service nodes when the prompt explicitly requests them.

When the prompt omits a TiDB version, resolve the latest stable release before `tiup cluster deploy`. Do not deploy `nightly` by default.

Use the new TiCDC architecture by default. If the prompt explicitly asks for old or legacy TiCDC architecture, scaffold with `--ticdc-architecture old` so the generated topology does not set `server_configs.cdc.newarch`.

## Quick Start

Resolve the skill directory, then scaffold a Terraform project:

```shell
python3 <skill-dir>/scripts/scaffold_project.py
```

Namespace defaults to `tidb-cluster-<YYMMDDHHMM>` (e.g. `tidb-cluster-2506261730`). Override with `--namespace` if needed:

```shell
python3 <skill-dir>/scripts/scaffold_project.py --namespace my-cluster
```

When `--target` is omitted, the scaffold script creates the project at `$HOME/test/tidb-aws-<namespace>`.
The script prints the absolute deployment path; report that path to the user.

Pass explicit counts only when the user requested them:

```shell
python3 <skill-dir>/scripts/scaffold_project.py \
  --target "$HOME/test/tidb-aws-cluster" \
  --namespace tidb-cluster \
  --n-pd 1 \
  --n-tidb 3 \
  --n-tikv 3 \
  --n-tiflash 1 \
  --n-ticdc 3 \
  --ticdc-architecture new \
  --extra-service service-name=1
```

Then follow [references/deployment.md](references/deployment.md) from the generated project directory.
Use the generated `README.md` in that directory as the local Terraform operation guide.

## Safety

- Before any Terraform command, verify AWS identity for the configured AWS profile. If the identity check fails, run AWS SSO login for that profile and retry the identity check.
- Do not run `terraform apply`, `terraform destroy`, or remote TiUP commands unless the user explicitly asked for deployment, startup, or cleanup.
- Do not print private key material, Terraform state contents, or AWS credentials.
- Warn the user that the template exposes SSH, Grafana, and TiDB Dashboard ports publicly, matching the source repo.
- Use a saved Terraform plan before apply unless the user explicitly asks for non-interactive deployment.
- Start TiUP clusters with `tiup cluster start <cluster-name> --yes`; do not add `--init`.
- After deployment/startup, display and report the absolute deployment path, the generated `README.md` path, and node IPs for all core and extra service nodes.