Hacker News new | comments | show | ask | jobs | submit login

He did, you do a rolling update, take a few systems out of active duty, upgrade them, and then begin directing some percent of traffic to them. That, of course, only works if you are of a scale to have that many production systems, and loose enough interdependencies to allow for new and old code to run at the same time. I think some of his point was that if you aren't able to do that sort of update, it was worth your time and effort to get there. I think he's probably right, but the work to get from where you are (it requires site outage to upgrade) to where you want to be (rolling upgrades, ability to run multiple versions in parallel, etc) is going to be site specific in most cases that it's hard to discuss that move in any depth.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact