TLDR Tech / DevOps
A one-line Kubernetes fix that saved 600 hours a year
Cloudflare engineers slashed Atlantis restart times from 30 minutes to 30 seconds by changing a single Kubernetes setting, reclaiming roughly 600 hours of blocked engineering time per year. The culprit was Kubernetes' default `fsGroupChangePolicy: Always`, which recursively changed permissions on millions of files in Atlantis' persistent volume with every restart—switching it to `OnRootMismatch` eliminated the bottleneck entirely.
Read the full story