Our graphs on 2 separate apps started looking very chaotic scaling up and down at the same time while not deploying a new version. Did fly.io release some new autoscaling algo?
there seems to have an issue with waking up serverless machines when they suspended.
if its only 1 app on one region then it wakes up fast, less than a second i belive.
if its scaled app on multi regions then the wake up time is couple of seconds for some reason? maybe load balancer maybe proxy issue?
i just tested and its so, i scaled down my machines to 1 in single region only and it wakes up fast again.
maybe you have same issue as mine
Hey @ViliamKopecky
I took a look at the logs and it seems the behavior you observe is because it takes quite a lot of time for the app to boot and pass the health checks:
2025-11-15 17:19:27.727251000 Starting machine
2025-11-15 17:21:37.521513000 Health check on port 3000 is now passing.
Proxy starts a machine, repeatedly tries to connect to it and eventually bails out due to timeout. It then retries the request on another machine, starts it and so on until one of them eventually passes the health checks.
Since now there are more machines running than needed some of them are stopped.
Was the startup time always like this or has something changed recently?
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.