Suspended machines are not started after deploy: new env vars are not set and health-checks are not run

[hilja]

Describe the bug

If I have all my machines in a suspended state and do a deploy. The machines are not started after the deploy, and:

• New env vars are not applied
• Health-checks are not run

I only know if my machine actually works when the first request hits it. Maybe I forgot to set a critical env var or import is missing, whatever.

I'm running long tasks on Chrome and it takes 18s to boot, but it comes out of suspension in a second or so. Would be really neat to have the initial boot happen on deploy so subsequent starts are fast.

I guess this is more of a feature request, but also it feels kinda broken :) The forums have a post on this suggesting opening a issue here.

Possible fix

Option to start suspended machines would be neat, something like:

flyctl deploy --force-start

Recommendations

For now, should I have an extra step after the deploy in my CI which starts the machines with fly machine start? That command needs a machine id so I would have to get those first, totally doable but not best DX :)

Info

fly version 0.3.231

[[services]]
  internal_port = 8080
  processes = ["app"]
  protocol = "tcp"
  auto_stop_machines = "suspend"
  auto_start_machines = true
  min_machines_running = 0

[hilja]

Starting machines is actually not hard at all:

MACHINE_IDS=$(flyctl machine list -a=foo --quiet)
flyctl machine start $MACHINE_IDS -a=foo