Product updates

June 2021

Rails Autoscale free plan

The 7-day trial is out. The forever free plan is in! More details in the plans & pricing docs.

February 2021

Scheduled autoscaling

You can now define weekly or daily schedules with a custom autoscale range. Read more in the launch announcment and the docs.

January 2021

Agent gem 0.10.2

v0.10.2 adds support for Ruby 3.

Agent gem 0.10.0

v0.10.0 brings several fixes and a few new features:

Check out the gem’s change log for more details.

October 2020

Dashboard refresh

I’ve removed the old Heroku nav bar (they no longer require it for add-ons) and updated the Rails Autoscale nav bar to include app info and wrap more nicely with lots of worker processes. You can also now adjust the vertical scale of your queue time chart, making it much easier to see differences in lower queue time values.

September 2020

Pricing changes

The old plans (Bronze, Silver, Gold, Platinum) have been replaced by plans specific to dyno types—Bronze-Standard, Bronze-Performance, Silver-Standard, etc. The official listing of available plans is on the Heroku Elements page, and more details are on Rails Autoscale’s Plans & Pricing docs page.

August 2020

Settings can now be updated via the API

The Rails Autoscale API can now be used to update your app settings. Some customers are using this endpoint to have different autoscale settings on a schedule, such as more lenient settings on the weekend.

Immediately apply updated autoscale settings

Previously, updating your “autoscale limit” settings would have no immediate effect. The limit was only changed when an upscale or downscale threshold was breached.

Now, Rails Autoscale will immediately check to see if your app is running within your new autoscale limit. If your app is scaled outside that range, you’ll autoscale up or down as necessary to satisfy the limit.

July 2020

Agent gem 0.9.0

v0.9.0 brings major improvements for worker backends and logging:

  • Official Resque support
  • Corrected handling of failed jobs in Delayed Job
  • Better handling of jobs without a specified queue name
  • A “quiet mode” option if you find the gem too chatty in your logs
  • An improved “debug mode” when you need all the detail in your logs

June 2020

Product updates page

The page you’re reading right now. 😁 I haven’t been great about tracking and sharing product updates, and this is my attempt to do better.

Documentation site

The official docs site is live! Prior to this, the only documentation was in the Heroku Dev Center, and it was insufficient. The new docs site will be the official canonical docs, frequently updated, and as extensive as possible.

May 2020

Agent gem 0.8.1, 0.8.2, and 0.8.3

A few bugfix releases were needed to address edge cases with Delayed Job and Que reporting.

Beta testing for Resque support

Autoscaling Resque has been the top feature request ever since Sidekiq autoscaling was launched in 2019. Currently beta testing with several customers.

Beta testing for Puma request body wait

Request queue times can be skewed when Puma spends time waiting for large request bodies to be transferred by slow clients. That time does not indicate a capacity issue, and thus should not be factored into request queue time for autoscaling purposes.

Allow autoscaling free/hobby dynos

Heroku does not allow free and hobby dynos to be scaled to more than one dyno, so Rails Autoscale had previously shown an error message if you attempted to enable autoscaling on these dynos. Some customers are autoscaling worker dynos down to zero dynos, though, and there’s no reason you shouldn’t be able to autoscale free/hobby dynos from 0-to-1. You can do that now.

April 2020

Dyno usage page

Your Rails Autoscale dashboard now links to a “Dyno Usage” page. This page shows a daily breakdown of how many dyno-hours you’ve used, the cost of those dynos, and the money saved by autoscaling. Here’s one crazy example.

March 2020

Agent gem 0.8.0 with DJ and Que support

Delayed Job and Que have joined Sidekiq in the job backends supported by worker autoscaling. 🚀