Lessons learned in production
summary
A Node-based file uploader is already live handling streaming to disk, thumbnailing and shipping to S3, running on a three-server cluster and communicating between nodes with a Redis cluster. Deployment initially copies dependencies into the repo but moves to local npm installs. Errors use an in-flight registry with timestamps and a watcher to keep processes alive, extensive logging and metrics (node port of Coda Hale metrics) give visibility, and ops involvement from day one reduces midnight pager shocks.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 0:55JavaScript in full stack at Yammer
- 3:00Organizing JS code and dependencies
- 5:08Async Node.js error handling
- 7:56Visibility and metrics in Node.js
- 9:00Ops collaboration for Node.js reliability
- 10:02Closing remarks
- 10:45Q&A on Node.js deployment
related talks
42mFrom Localhost to Production
40mDeploying node.js at scale
33mTaking Node.js to The Ex-Stream: Stream to Scale
31mUsing NodeJS to Build Micro Services Architecture
34mYou Don't Know Node.js
31mKeynote Panel: Node.js at Massive Scale
43mVersioning, Syncing & Streaming Large Datasets Using DAT + Node
31mNode.js that's hugely reliable, fast, and scalable