Files

T

Antigravity Agent 504fd5fb42 [G1-Sync] Manual knowledge update

2026-05-10 22:08:15 +09:00

5.1 KiB

Raw Blame History

id, title, category, status, source_trust_level, verification_status, created_at, updated_at, tags, tech_stack, applied_in, aliases

title

Spinnaker / Tekton

Modern CI/CD beyond Jenkins. Spinnaker (multi-cloud), Tekton (K8s-native).

📖 핵심 개념

Spinnaker: Netflix 의 multi-cloud CD.
Tekton: K8s-native pipeline.
매 stage = container.
GitOps 친화.

💻 코드 패턴

Spinnaker pipeline (UI / JSON)

{
  "name": "Deploy",
  "stages": [
    { "type": "bake", "package": "my-app", "baseOs": "ubuntu" },
    { "type": "deploy", "clusters": [{ "account": "aws", "region": "us-east-1" }] },
    { "type": "manualJudgment", "message": "Approve prod?" },
    { "type": "deploy", "clusters": [{ "account": "aws-prod" }] }
  ]
}

→ Multi-cloud (AWS, GCP, Azure, K8s).

Spinnaker 의 강점

- Multi-cloud (이주 cluster).
- Canary (Kayenta automated analysis).
- Manual judgment (approval).
- Trigger (Jenkins, Docker, cron).
- Audit / RBAC.

Tekton (K8s-native)

apiVersion: tekton.dev/v1
kind: Pipeline
metadata:
  name: build-deploy
spec:
  tasks:
    - name: build
      taskRef: { name: build-image }
    - name: deploy
      runAfter: [build]
      taskRef: { name: deploy-app }

apiVersion: tekton.dev/v1
kind: Task
metadata:
  name: build-image
spec:
  steps:
    - name: build
      image: gcr.io/kaniko-project/executor
      args: [--dockerfile=./Dockerfile, --destination=my-image]

→ 매 task = pod.

PipelineRun

apiVersion: tekton.dev/v1
kind: PipelineRun
metadata:
  name: pr-1
spec:
  pipelineRef: { name: build-deploy }

→ Trigger.

Tekton Triggers

apiVersion: triggers.tekton.dev/v1beta1
kind: EventListener
metadata:
  name: github-listener
spec:
  triggers:
    - name: github-push
      interceptors:
        - ref: { name: 'github' }
      bindings:
        - ref: { name: github-binding }
      template:
        ref: build-deploy-template

→ GitHub webhook 가 PipelineRun.

vs Jenkins

Jenkins:
- Mature.
- Plugin 큰 ecosystem.
- Java VM.
- 큰 server.

Spinnaker:
- Multi-cloud first.
- Canary / blue-green.
- Netflix scale.

Tekton:
- K8s-native.
- 작은 (pod 별).
- GitOps.

→ K8s = Tekton. Multi-cloud / 큰 enterprise = Spinnaker. Legacy = Jenkins.

vs GitHub Actions

GitHub Actions:
- GitHub 친화.
- Action marketplace.
- 작은-중간 project.

Tekton:
- K8s-native (자체 cluster).
- 큰 organization.

→ Modern startup = GHA.
큰 enterprise + K8s = Tekton.

vs Argo Workflows

Argo Workflows:
- K8s-native (Tekton 와 비슷).
- DAG-based.
- 더 popular (Argo ecosystem).

Tekton:
- Linear pipeline.
- More 작은 + simpler.

→ Argo 가 더 powerful.
Tekton 가 더 simple.

Spinnaker canary (Kayenta)

{
  "stages": [
    {
      "type": "kayentaCanary",
      "canaryConfig": {
        "metricsAccountName": "datadog",
        "scoreThresholds": { "marginal": 75, "pass": 95 }
      }
    }
  ]
}

→ Datadog metrics 가 canary 의 OK / abort.

Tekton + ArgoCD

1. PR merge → Tekton build image.
2. Tekton update Helm chart (git).
3. ArgoCD detect change → deploy.

→ Build (Tekton) + Deploy (ArgoCD).

Spinnaker + ArgoCD

Spinnaker = orchestrator.
ArgoCD = K8s sync.

→ Spinnaker 가 multi-cluster orchestrate.

Production

Netflix: Spinnaker 의 creator.
Google: 자체 (similar to Tekton).
Capital One: Spinnaker.
IBM: Tekton 의 contributor.

When NOT?

Spinnaker:
- 작은 team (overkill).
- Single cloud (simpler tool).

Tekton:
- 작은 / non-K8s.
- GitHub-only (GHA 충분).

Cost

Spinnaker: free (open) + 큰 운영 cost.
Tekton: free + K8s cost.
GitHub Actions: per-use.
Jenkins: free + 운영.

→ 운영 cost 가 큰 factor.

Monitoring

- Pipeline duration.
- Failure rate.
- Queue depth.
- Stage-level metric.

→ Datadog / Prometheus.

Best practice

1. Pipeline as code (git).
2. 매 stage 의 timeout.
3. Auto-rollback.
4. Manual approval (production).
5. Canary / blue-green.
6. Audit log.

🤔 의사결정 기준

상황	추천
K8s-native	Tekton / Argo Workflows
Multi-cloud	Spinnaker
GitHub project	GitHub Actions
Legacy enterprise	Jenkins
작은 startup	GHA / GitLab CI
Canary auto-analysis	Spinnaker (Kayenta)

❌ 안티패턴

Spinnaker on single cloud: overkill.
Tekton + non-K8s: 잘못 선택.
모든 거 manual approval: slow.
No timeout: hang.
No audit: incident response 어려움.

🤖 LLM 활용 힌트

Spinnaker = multi-cloud + Netflix-scale.
Tekton = K8s-native pipeline.
Argo Workflows = Tekton 의 popular alternative.
GitOps + Spinnaker / Tekton = production.

5.1 KiB Raw Blame History