Hacker News Reader

Give yourself enough time that you're confident you'll be able to finish the maintenance within that timeframe. I was bitten more than once by scheduling a maintenance window too small, and then becoming frantic as the maintenance window passed and I still wasn't finished.

View on HN

Don't overestimate by orders of magnitude though. There was one change where I expected it would take 5 minutes, I said "Gimme an hour to be safe", and my friend (who was in charge of the server) posted a maintenance window for 5 hours. It actually took about a minute, and then people were left thinking "Has the downtime happened? Is it going to go down later?"

It seems the lesson is also to communicate the downtime effectively with your clients.

Do you have a staging/test site, a mirror of your live site? Normally you apply all the changes to the test site so that you know the upgrade works, and how long it will take. Then you can schedule downtime as required.

On other projects, yeah, but not on this one. Or actually, this one had a staging site (which I tested changes on first), but its data was about 8 months out of date and its configuration didn't exactly match the production site. Volunteer project, limited resources.

Reply on news.ycombinator.com