蓝绿部署与渐进式发布
使用蓝绿部署和渐进式流量发布安全地部署边缘应用，降低风险并支持即时回滚。
蓝绿部署与渐进式发布 是 CoddyKit 上的免费 Edge Computing with Cloudflare Workers & Deno 课时。 这是第 4 节课，共 4 节。 你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。 这是 Edge Computing with Cloudflare Workers & Deno 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 Edge Computing with Cloudflare Workers & Deno 课程共包含 4 节课。
本课时的部分内容尚未翻译，以英文显示。
The Risk of Big-Bang Deploys
Shipping a new version to 100% of users at once is risky, a hidden bug hits everyone instantly.
Progressive deployment strategies reduce blast radius by exposing changes gradually and enabling fast rollback.
Blue-Green Deployments
In a blue-green setup you run two environments:
- Blue, the current live version
- Green, the new version, fully deployed but not yet receiving traffic
You switch traffic to green when ready, and switch back instantly if something breaks.
Gradual (Canary) Rollouts
A gradual rollout sends a small percentage of traffic to the new version first, then increases it as confidence grows.
- 5% then 25% then 50% then 100%
- Watch error rates at each step
Also called a canary release.
Versioned Uploads
Cloudflare lets you upload a new Worker version without deploying it to all traffic.
wrangler versions upload
# uploads a new version, 0% traffic by defaultSplitting Traffic by Percentage
Use gradual deployments to assign a percentage of requests to each version.
wrangler versions deploy \
<new-version-id>@10% <old-version-id>@90%Manual Routing for Control
For custom logic you can route in code, for example send a cohort to the new behavior based on a hash of the user ID.
function inCanary(userId) {
// simple stable bucketing
return (userId.charCodeAt(0) % 100) < 10; // ~10%
}Watch the Right Metrics
During a rollout, monitor:
- Error rate / 5xx responses
- Latency (p50, p95, p99)
- CPU time
- Business KPIs
Compare the new version against the baseline before increasing traffic.
Instant Rollback
The key benefit of progressive deploys is fast rollback, shift traffic back to the known-good version immediately.
wrangler versions deploy <old-version-id>@100%Automating Rollback
Tie rollout steps to CI checks. If error rates exceed a threshold, the pipeline automatically reverts to the previous version.
if (errorRate > 0.02) {
await rollbackToPrevious();
}Database Compatibility
During blue-green, both versions may run at once, so schema changes must be backward compatible.
- Add columns before removing old ones
- Use expand-then-contract migrations
Best Practices Summary
Deploy safely at the edge by:
- Uploading versions without full deploy
- Splitting traffic gradually and watching metrics
- Keeping rollback one command away
- Making migrations backward compatible
Quick Check
What is the defining characteristic of a canary (gradual) rollout?
Recap
You learned safe deployment strategies:
- Blue-green keeps a hot standby for instant switching
- Gradual rollouts shift traffic in steps while you watch metrics
- Versioned uploads and percentage splits make this easy on Cloudflare
- Backward-compatible migrations keep both versions healthy
Progressive delivery turns scary deploys into routine, reversible events.
常见问题解答
「蓝绿部署与渐进式发布」课时是免费的吗？
是的 — 「蓝绿部署与渐进式发布」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 Edge Computing with Cloudflare Workers & Deno 课程的其余内容，请升级到 CoddyKit PRO。 Edge Computing with Cloudflare Workers & Deno 课程共包含 4 节课。
「蓝绿部署与渐进式发布」这节课中我会学到什么？
使用蓝绿部署和渐进式流量发布安全地部署边缘应用，降低风险并支持即时回滚。 你通过在浏览器中直接运行的动手代码来练习 Edge Computing with Cloudflare Workers & Deno，全天候 AI 导师会在你学习这节课的过程中回答你的问题。
学习 Edge Computing with Cloudflare Workers & Deno 需要有经验吗？
无需任何先前经验。CoddyKit 上的 Edge Computing with Cloudflare Workers & Deno 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。 这是第 4 节课，共 4 节。
「蓝绿部署与渐进式发布」课时需要多长时间？
大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。
我能在这节 Edge Computing with Cloudflare Workers & Deno 课中编写并运行代码吗？
能。每节 Edge Computing with Cloudflare Workers & Deno 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。
此课程中的所有课时
- Deno Deploy 与 CLI
- 自定义运行时环境
- 边缘测试与 CI/CD
- 蓝绿部署与渐进式发布