Improvements to PaaS UI settings

Fresh Produce
1

image
image1492×930 127 KB

We just shipped many more settings forms that are tweakable via dashboard UI per-app!

These take effect immediately upon saving which means some changes require deploying the app.

As for the updated fly.toml

  • If you have a connected GitHub repo: we will create a PR you can merge
  • If you don’t have a connected GitHub repo: we will advise you to run fly config save -a APP_NAME on your terminal to save the current state.

This is far from all what a fly.toml can do but hopefully this will let one scale their apps better

image
image1740×950 186 KB

Also we updated out machines list page to let you tweak more than scale for process groups! The “Edit app scale” button became a cog in the group name to make things easier to manage per group.

Extensive list of changes

Yes I asked a LLM to scan my commits, sorry if this feels robot-y:

Manage more of your fly.toml directly from the app dashboard

  • Edit the [deploy] section from the Configuration page: pick a deploy strategy (rolling, immediate, canary, blue/green), set a release command, tweak wait_timeout, release_command_timeout, and max_unavailable.
  • Edit top-level app settings from the same page: primary region, kill signal, kill timeout, console command, and swap size.
  • Configure a per-process-group restart policy (always, never, on-failure) and retry count from the process group scaling form.

Better guardrails and hints

  • min_machines_running warnings now count only machines in your primary region, so you see accurate under-scaled warnings.
  • If your app has mounted volumes, the Configuration page will tell you upfront that the “canary” deploy strategy isn’t compatible.

Better consistency with flyctl

  • Duration fields (kill_timeout, wait_timeout, release_command_timeout) accept the same units as flyctl: ms, s, m, h.
  • Kill signal is now a free-form field with common suggestions rather than a rigid dropdown — pick anything valid (SIGABRT, SIGHUP, etc.), not just the six defaults.
  • The on-failure restart policy — the platform default — is now a first-class option you can select explicitly and see round-trip through the form.
  • max_unavailable is preserved for canary deploys (previously only kept for rolling).

Under the hood

  • Config PRs now cover the new [deploy], [[restart]], and top-level scalar changes.
  • All the new fields are validated at the form level with clear, actionable error messages before we ever hit the API.