Lesson 7 — Beginner

Containers & Deployment

Tumhari service laptop par chal rahi hai — ab use public URL tak le jaana hai. Wo tabhi hoga jab environment bhi saath chale. Is lesson mein application vs environment vs machine, containers aur images, Dockerfile likhna, layers aur cache, container logs se startup debugging, FastAPI ko containerize karna, cloud deploy aur production secrets dekhoge.

Containers kya hain — aur kyun

Teen cheezein alag hain: application (tumhara code), environment (Python version, dependencies, env vars, OS libraries) aur machine (laptop, server, cloud VM). Local pe theek chalna sirf ye prove karta hai ki tumhara laptop sahi hai — production ki machine alag hoti hai. Container isi gap ko bridge karta hai.

Images, layers aur cache

$ docker build -t order-service:1.0 .
 => [1/5] FROM python:3.12-slim             0.8s
 => [2/5] WORKDIR /app                      0.0s
 => [3/5] COPY pyproject.toml uv.lock ./    0.1s
 => [4/5] RUN uv sync --no-dev             12.4s
 => [5/5] COPY . .                          0.2s
 => naming to docker.io/library/order-service:1.0

• Har Dockerfile instruction ek layer banati hai — aur layer cache hoti hai. Usme sirf ek line badli toh wahi layer dobara banti hai, baaki purane cache se aate hain.
• Order matter karta hai: pehle dependency files + install (kam badalta hai), phir app code (aksar badalta hai). Ulti order = har code change par poora install.
• Tags do: order-service:1.0 jaisa version. Latest par kabhi rely mat karo — server pe kya chal raha hai kisi ko nahi pata.
• .dockerignore = build context ka filter — .git, .env, .venv build mein jaane se pehle hi bahar.

Dockerfile — instruction ka instruction

FROM python:3.12-slim
WORKDIR /app

# Pehle dependency files — inke badle par hi install dobara
COPY pyproject.toml uv.lock ./
RUN pip install uv && uv sync --no-dev

# Phir baaki code
COPY . .

EXPOSE 8000
CMD ["uv", "run", "uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8000"]

• FROM = base image. python:3.12-slim = chhota official image — zaroorat ke bina bloat nahi.
• WORKDIR = andar ka current directory — COPY, CMD sab yahan se chalte hain. Path confusion isi se aata hai.
• EXPOSE sirf documentation hai (kaunsa port andar chal raha hai) — host se port tabhi khulta hai jab run mein -p do.
• CMD = container start ka command. Image banana alag hai, chalana alag — doosra command de kar override bhi ho jaata hai.

FastAPI service ko chalao

# Build -> chala -> smoke test (cloud se pehle, hamesha local)
docker build -t order-service:1.0 .
docker run --rm -p 8000:8000 --env-file .env order-service:1.0
# Ab: http://localhost:8000/docs

# Env vars safely — image ke andar bake nahi, run ke waqt do
docker run --rm -p 8000:8000 -e DATABASE_URL="postgres://app@db:5432/orders" -e LOG_LEVEL=info order-service:1.0

# Debug
docker ps
docker logs -f <container-id>
docker exec -it <container-id> sh

• -p 8000:8000 = host:container. Andar app 8000 pe chalti hai, bahar tum 8000 se pahunche — dono alag hain.
• --env-file .env ya -e: secrets image ke andar kabhi mat COPY karo — wo layers mein chhap jaate hain aur docker history se nikal jaate hain.
• Local smoke test: build -> run -> curl /health. Cloud mein jaake debug karna sabse mehnga tareeqa hai.
• docker exec sh se andar jaakar dekho: env vars set hain? files hain? Python import ho raha hai?

Logs, .dockerignore aur debugging

# .dockerignore — context chhota + secrets ke andar jaane se bachav
.git
.env
.env.*
__pycache__/
.pytest_cache/
.venv/
node_modules/

# Container ka crash dekho
docker logs <container-id>
=> Traceback (most recent call last):
=>   File "/app/app/main.py", line 3, in <module>
=>     from app.routes import orders
=> ModuleNotFoundError: No module named "app.routes"
# matlab: COPY/WORKDIR order galat — code sahi jagah nahi pahauncha

• Logs stdout/stderr se aate hain — print() aur uvicorn logging dono docker logs mein dikhte hain. Alag log file banane ki zarurat nahi, production rule bhi wahi hai.
• Common startup failures: missing env var (KeyError), wrong import path, port mismatch (app 8000 pe, platform dusra de raha hai), dev dependency --no-dev mein chali gayi.
• .env image mein gaya toh tar/history se recover ho sakta hai — .gitignore ki tarah .dockerignore bhi zaroori hai.
• docker history image se har layer dikhti hai — isse verify karo ki koi secret layer mein toh nahi pada.

Public URL pe deploy aur production secrets

# Generic flow — platform koi bhi ho, steps same hain
docker build -t registry.example.com/team/order-service:1.0 .
docker push registry.example.com/team/order-service:1.0
# Platform par: image reference + env vars -> service chalu -> public URL

# Public URL ka smoke test (docs UI nahi, curl se)
curl -sS https://order-service.example.com/health
curl -sS -X POST https://order-service.example.com/orders -H "Content-Type: application/json" -d @order.json

# Production config alag — local .env se kabhi mat uthao:
# DATABASE_URL, API_KEY, LOG_LEVEL platform ke secrets manager/dashboard se

• Flow: image banao -> registry push -> platform ko image do -> env vars/secrets set -> health check. Code ek baar build hota hai, har environment alag config leta hai.
• Public URL aate hi pehla kaam: health endpoint + ek real POST curl se — scriptable hai, CI mein bhi chalega. Sirf homepage kholna smoke test nahi.
• Local vs production config alag: .env sirf local (gitignored), production mein platform ke secrets manager se env vars inject hote hain.
• Secret leak ho jaye toh turant rotate karo. Repo mein pada hai toh history se hatana bhi padega — prevention hi sasta hai.

Exercise

Common mistakes

• COPY . . se .env aur .venv bhi image mein: Secrets layers mein chhap jaate hain aur image bhaari. Fix: .dockerignore pehle likho — .env, .venv, .git sab bahar.
• Dependency aur code ki layer order ulta: Har chhote code change par poora install chalega — build minutes se hours mein. Fix: dependency files + install neeche, code COPY upar.
• Port confusion: EXPOSE se port khulta nahi — sirf document hota hai. Fix: run mein -p host:container, aur app ko 0.0.0.0 par bind karo (localhost nahi).
• Cloud mein jaake debug karna: Deploy ke baad error dhoondhna mehnga aur dheema hai. Fix: local par wahi image build + run + curl smoke test, tabhi push.

docker build -t kitchen/paneer-box:1.0 .      # box pack hua
docker run -p 8000:8000 kitchen/paneer-box:1.0 # outlet par khola
docker push registry/kitchen/paneer-box:1.0   # dusre outlet bheja
# Wahi box, wahi khana — sirf outlet (machine) badla.