Autostop doesn't work

Questions / Help
,
1

Hello,

This is the first time I’ve encountered this issue with the auto-stop functionality. All my other apps handle it perfectly fine.

For context, it’s a database that I access using its flycast address (myapp.flycast:port).

It remains constantly running, even without usage. Therefore, the auto-stop feature doesn’t work. However, if I run fly machines stop my-machine-id, it restarts fine when in use, and the health checks run smoothly.

Do you have any suggestions?

For your information, here’s my fly.toml:

app = "myapp"
primary_region = "cdg"
kill_signal = "SIGINT"
kill_timeout = "5s"

[experimental]
  auto_rollback = true

[env]
  PORT = "27017"

[[mounts]]
....some data

[[services]]
  internal_port = 27017
  auto_start_machines = true
  auto_stop_machines = true
  min_machines_running = 0
  processes = ["app"]
  protocol = "tcp"

  [services.concurrency]
    hard_limit = 25
    soft_limit = 20
    type = "connections"

  [[services.ports]]
    port = 27017

  [[services.tcp_checks]]
    grace_period = "1s"
    interval = "15s"
    restart_limit = 0
    timeout = "2s"

[[vm]]
  cpu_kind = "shared"
  cpus = 2
  memory_mb = 512
2

Hi… Databases in particular are prone to being kept awake by open connections from client machines, even when there isn’t any query traffic.

You might try looking for those…

$ fly ssh console -a mongodb-app-name
# ss -tn state established '( sport = 27017 )'

(Note that the spaces before and after the parentheses are (weirdly) mandatory.)

On Debian, ss can be found in the iproute2 package.

3

Hi! Thanks for your answer!

The weird part is that I just recreated an app, identical to an old one, in a new region, and with more disk space. And on the old one, everything was working fine.

Indeed, I had zombie connections that I was able to cut. I will keep an eye on this issue and see if it comes back!

4

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.