High-Availability Django
summary
Theatlantic.com migrates from PHP to Django and cuts catastrophic deploy pauses by switching from mod_wsgi to uWSGI Zerg mode and preloading workers, using an A/B symlink deploy and warming workers with cache‑busted requests to achieve zero-downtime deploys. An nginx proxy cache holds responses when backends fail, and query changes (using .values() instead of full model hydration) shave ~80 ms per request, moving end-to-end times from ~80 seconds at worst to ~150 ms in steady state. Monitoring uses New Relic/opbeat and a uWSGI stats page.
AI-written summary · about
chapters (11)
- 0:00Intro
- 0:15Overview
- 2:44Profiling
- 3:59Caching
- 6:19Query Optimization
- 8:11Server Upgrade
- 9:25Directory Structure
- 11:00Stage and Live
- 13:04ModWhiskey Documentation
- 13:37YouWhisky Monitoring
- 16:29Questions
