The CPU usage and bottleneck is normal and expected, eg doing SSR in react. I don’t believe the CPU changes would affect anything here though, per their clarification.
The point I’m trying to make is suspend is not ready for production.
I encounter a bug, suspended machine disappear from the dashboard in the organisation’s machines overview (they are correctly listed in app’s machine page).
Also, and it’s may be related, we can’t filter on the suspended machine state:
Hello there, I have a question and I can’t find answer on it in docs.
Is there a way stop/suspend machine which has not public http port?
For example I have frontend on nodejs which call backend http/grpc(no matter) by private net. Frontend can stop whithout activity it is ok but backend is aways run it is problem. I do not want expose backend port to public.
Take a look at flycast. Also fly machine suspend and machine API.
Flycast is easiest. Define your backend service with only a private ipv6, and instead of referencing it via a .internal address, use .flycast, and fly proxy will take care of suspend/resume.
1 Like
I’ve enabled this feature on my app, which in general has been an amazing change.
The only issue I’ve found that sometimes when an updated build is deployed upon a commit to my master branch in github, it often doesn’t pick up the changes in the build yet the machine states it is running on the most recent version of the app.
Hello everyone,
I just tried “suspense” and noticed machines with mounted volumes don’t suspend - they restart (4-5s vs 250ms).
EDIT: I first thought auto_stop_machines = "suspend" doesn’t work at all with mounted volumes, but just figured it’s because we had swap_size_mb enabled. “suspend” only works with no swap.