Coming next week: How we moved off Heroku onto AWS after the Heroku outage.

Cloud services go down. If that's a problem for your business, plan for it.

Heroku is on AWS, so I'm not sure that would help. I agree cloud services have disadvantages but you pay a significant productivity tax to manage your own hardware, otherwise you're always putting your faith in someone else to keep your site online.

A Heroku outage is not necessarily caused by an AWS outage.

Not necessarily, but I think history would show, usually.

In our experience, Heroku has orders of magnitude more issues than AWS. Look at their status page for a list of issues since only Jan 25th: https://status.heroku.com/

They pretty much have an issue every other day. Sure it usually only effects a small subset of users, but you wind up getting bit by them more often than necessary.

We were on Heroku for more than a year before we couldn't tolerate the sporadic outages and performance degradations anymore.

But you also don't automatically get AWS's level of reliability by skipping Heroku and maintaining your own deployments on AWS. Once you go that route your uptime becomes AWS minus your competence level.

Could easily be worse than Heroku, especially for a very small shop.

OK, can't argue with that.

Did you guys look into exporting your data from mysql to postgres with taps? (http://devcenter.heroku.com/articles/taps). I've done this many times and it's been pretty simple.

We tried taps but it wasn't working for us and Craig Kerstiens, Heroku's Django guy, told us to use the pg:backups approach instead.

Great details in the post. Thanks for sharing!