
 Changelog Master Feed Kaizen! Mop-up job (Changelog & Friends #114)
 Oct 24, 2025 
 The discussion kicks off with a recap of the exhilarating Pipely launch in Denver, followed by personal updates from the hosts. They dive into performance metrics, revealing impressive cache hit ratios and response time improvements. The exploration of open video standards leads to insights on transcoding challenges. A lively debate about future gatherings arises, while Gerhard showcases his astonishing 100Gbps home lab. They also touch on innovative home automation projects and conclude with light-hearted banter, leaving listeners eager for more! 
 AI Snips 
 Chapters 
 Transcript 
 Episode notes 
Stage Launch, Lunch, Then Metrics Surprise
- Gerhard describes launching Pipely on stage in Denver and then checking metrics after lunch when things started failing.
 - He under‑provisioned instances, traffic spiked, CPU hit 100% and the new CDN began to fail under load.
 
Always Use Blue‑Green Deploys
- Always introduce changes alongside the existing system so you can revert quickly if something breaks.
 - Use blue-green deployments to run new and old infrastructure in parallel and rollback by redirecting traffic.
 
Hike While Rollback Saved The Day
- While Pipely was overwhelmed, the hosts went hiking and avoided a production firefight because they had blue-green fallback.
 - The issue was fixed by restoring previous DNS records, routing traffic back to the old CDN within minutes.
 
