On Friday 6:30 PM PST, the Sidekiq Pro gem server became unavailable due to a Linode data center outage in Fremont, CA. Connectivity was restored after 15 hours at Saturday 9:30 AM PST. I’m terribly sorry not only for the outage but the extended length of time.
I knew that the gem server was a single point of failure and had plans to build a redundant server later this summer. I had spent some time in Oct 2014 building a new, simpler gem server but hadn’t automated the process - the gem server was still a “unique snowflake” and had a number of files which couldn’t easily be recovered to bring up a replacement server.
While I was waiting for the server to become available, I started the process of automating the gem server build Saturday morning. Once connectivity was restored, I immediately copied the irreplacable files off-site. As of today (Monday) I was able to build a new DigitalOcean droplet with a simple shell script and a repository of those critical files necessary (TLS cert, .gem binary files, etc). Within 10 minutes of creation it was successfully serving Sidekiq Pro’s gems to my test app.
By the end of the month I will have two servers available, one in SF and one in NY. Each server will be rebuilt annually to ensure the build process is tested every 6 months and remains usable. I hope this will get us closer to the goal of 100% uptime.
If you are interested in app deploys that don’t depend on any gem servers, take a look at the
bundle package command.