Starting today, Dynoscale users can now connect their Heroku Node applications to the Dynoscale Heroku Addon. This is now possible with dynoscale-node. It ships request queuing data to Dynoscale so your dynos scale from real user traffic.
Dynoscale is smarter than Heroku's default scaling options and was built to help you save money. The core principle of Dynoscale is that Request Queuing is a better metric than Request Latency. Request queuing is the time spent doing everything for a request except executing your application's business logic. This means that intentionally long running code flows doesn't trigger a false positive auto scale event. This means that your Application will only auto scale when your current dyno count cannot handle user traffic.
The package requires Node.js 18 or newer. Mount dynoscale() as middleware on Express or Connect before your routes, or pass each request through it from node:http. The agent reports from the dyno whose name ends in .1, normally web.1. Install it with npm install dynoscale-node, deploy, and make an http request to start sampling.
Worker dynos are covered too. With BullMQ 5 or newer installed, name the queues in DYNOSCALE_BULLMQ_QUEUES or in the middleware config. Dynoscale then scales those worker dynos from queue latency, the same way Ruby apps scale Sidekiq and Resque workers.
To get started, follow the instructions outlined here in our knowledge base. If the agent is installed and samples are not arriving, use the Node troubleshooting guide.
heroku addons:create dscale