Server (failover) redundancy solution for Heroku app

Question

I have currently a node.js app deployed on a free web Dyno running on Heroku. As planning to make it production, I need to think about a redundancy and failover solution at a reasonable cost.

As I ran "Prodiction Check" on the Heroku Dashboard, it gave me a list of things to do to make it production. One of the things is "Dyno redundancy" that I should have at least 2 web dynos running for failover. Does it mean I should upgrade my Free Dyno to Hobby or Standart 1X, and should I also need to have two dyno of the same type, e.g. two Hobby dynos or two Standard 1X dynos?

How does Heroku handle failover from one Dyno to another one?

Thanks!

Answer 1

Heroku shares traffic between all available dynos, distributing requests using a random assignment algorithm. So all your dynos will always be serving incoming traffic.

This provides redundancy, not failover. If one dyno is choking on a very slow request, the app will still be available via the other dynos.

Failover is different. In the case of an application failure (say, the database is inaccessible) Heroku's router offers little help. To deal with more industrial workloads, you could use Amazon Route 53's DNS-level failover, which runs a health check against the backend and will reroute the domain name in the case of a Heroku crash.

However for many use-cases it is probably enough to simply offer a friendly, customised HTTP 503 error page, which you can configure in Heroku, to keep users happy during an outage.