Redis To Go

» see redistogo.com as it lived, on the wayback machine

~ the goodbye ~

I have written code that talks to Redis for years and have never once rented one. The work I do puts the process and its memory on the same board, often a board that has to get airborne, so a key-value store reached through a URL in an environment variable belongs to a different shape of engineering. Redis To Go was built for that shape and almost nothing else: the developer with a Rails app on Heroku who needed a job queue by Friday.

It started in July 2010. Heroku had written the first Redis add-on for its own platform, then spun it out and handed it to a separate team to run. A third party therefore operated the default way of getting Redis into a Heroku app, and kept operating it for twelve years. Exceptional Cloud Services ran it next to Airbrake and Exceptional, two error trackers, and the group claimed more than fifty thousand developers across the lot.

Rackspace bought the whole company on 28 March 2013, a month after it had bought ObjectRocket, a managed database outfit. The press release led on error tracking and treated the Redis hosting as the sweetener. Redis To Go kept serving under the new owner for another nine years, changing hands three times in total by the count of one of the Heroku engineers who had been there at the spin-out.

~ the chronicle ~

~ the free plan ~

The free tier is why this one gets remembered. You attached an instance with a single command, paid nothing, and shipped. A Heroku dyno restarts and has no persistent disk, so running Redis next to your own app was not an available option, and the free add-on filled that hole for a decade of tutorials, side projects and job queues that never grew up.

The shutdown notice knew it. Rackspace recommended ObjectRocket for Redis, listed the four data centers an instance could live in, and then put one line in capitals: "Please note, ObjectRocket Redis DOES NOT have a free tier plan available." The capitals are Rackspace's own.

~ stuck on three ~

The technical cause of death was a version number. Redis To Go never offered anything past Redis 3.x, and by 2022 that mattered, because Sidekiq, the Ruby background job library a great many of these instances existed to serve, had moved to requiring Redis 4. Upgrade one and you lost the other.

Mike Perham, who writes Sidekiq, posted the account that stuck: "Sidekiq didn't kill RTG, they chose to die. RTG decided not to offer future versions of Redis past 3.x. Once the owners stop investing time and effort into a company, they're choosing to let it die while pocketing any remaining profit." Rackspace never answered that, and never explained why paying customers were not simply moved onto ObjectRocket, the Redis service it had owned since 2013.

~ eight weeks in summer ~

The notice went up around 13 June 2022 for a shutdown on 8 August, so anybody running a production queue on it had eight weeks, most of them inside the northern hemisphere's vacation season. Heroku mailed its own customers and then left its add-on documentation untouched, so for a while the official page was still advertising the number one Redis provider with over fifty thousand instances while the service was being wound down behind it. On the day, the instances were switched off and the homepage became the notice. It is still the notice today.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »