Zero-downtime payment platforms

on YouTube ↗40m

" summary

Rails 3.2 app on Heroku with Postgres 9.x at peak doing ~$1,000/min uses timeouts (15s) around card charges to save Gateway-Dash-Down- placeholders, a cron every ~10 minutes to retry, and a separate Sinatra chocolate replayer behind Akamai that records raw requests for manual replay. Duplicate-safe replay uses Akamai request IDs, race conditions and Heroku random routing still cause daily spurious failovers, but the system materially keeps customers buying when gateways or Heroku briefly fail.

AI-written summary · about

" chapters (8)

1. 0:00Introduction to site uptime
2. 4:09Manual failover strategies
3. 8:16Automating payment recovery
4. 13:52Handling infrastructure failure
5. 17:10Implementing Chocolate replayer
6. 21:51Request replaying and dedupe
7. 26:59Addressing random routing
8. 30:20Q&A and final conclusions

" related talks

• 37mHigh Performance Political RevolutionsBraulio Carreno · RailsConf 2017
• 47mUptime == Money: High AvailabilityPaul Gross · RubyConf AU 2013
• 33mResilientSmit Shah · RailsConf 2015
• 37m`bundle install` Y U SO SLOW: Server EditionTerence Lee · Rubyfuza 2013
• 32mZero Downtime Deploys Made EasyMatt Duncan · RubyConf 2012
• 37mDeathmatch Bundler vs Rubygems.orgAndre Arko · GoRuCo 2013
• 33mHerding ElephantsClint Shryock · Big Ruby 2014
• 30mBanking with Django - how to not lose your customer's moneyAnssi Kääriäinen · DjangoCon Europe 2018