us-east-1

Jul 20, 2016

I have fond memories of operating Heroku in us-east-1. Be it the electrical storm of 2012, or the ELB failures of Christmas Eve, the region is plagued by more than its fair share of outages and problems. With Heroku being entirely single region, these major cross-AZ issues would take us down every time. Here are a few favorites and links to their postmorterms:

Today, us-east-1 experienced yet another major API outage; with the API being degraded or completely down completely for two hours between 1115a and 115p PT. True to their reputation, Amazon was slow to update their status page, leaving users struggling to understand what was going on for themselves.

About a year ago, Stripe undertook the major project of moving all its servers over to us-west-2 (Oregon), and completed it within the last year. The project had a few major objectives, getting our fleet onto VPC being one for example, but among them was the simple goal of anchoring into one of AWS’ most modern and problem-free data centers.

The project involved an incredible amount of effort on the part of many people, but seems to be paying off already. While my friends at places like Heroku and Citus labored under duress waiting for AWS to come back online, the silence over on the west coast was deafening.

Did I make a mistake? Please consider sending a pull request.