Machine restart policy

The Machine restart policy defines whether and how flyd restarts a Machine after its main process exits. The restart policy applies per Machine, and is not an app-wide setting.

The restart policies are:

• no: Never try to restart a Machine automatically when its main process exits, whether that’s on purpose or on a crash. no is the default when you use the --rm option to create a Machine with fly m run that auto-destroys on exit.

• always: Always restart a Machine automatically and never let it enter a stopped state, even when the main process exits cleanly. always is the default when you create a Machine with fly m run and for Fly Postgres app Machines. Recommended for “always-on” apps with no services configured, since the Machine restarts regardless of the exit code.

• on-fail (or on-failure for the Machines API and when viewed in the Machine config): Try up to 10 times to automatically restart the Machine if it exits with a non-zero exit code, before letting it stop. Recommended for most Machines with services configured, since Fly Proxy can wake them on request. on-fail lets Machines be restarted if they crash, and allows your app Machines to effectively scale down by exiting cleanly. on-fail is the default when there’s no explicit restart policy in a Machine’s config, such as Machines created by fly launch and fly deploy. Machines with a schedule also default to the on-fail restart policy.

Check a Machine’s restart policy

Display a Machine’s status and its config in json format:

fly m status -d <machine id> 

Example output with a restart policy of always:

...
Config:
{
  "init": {},
  "image": "registry-1.docker.io/flyio/hellofly:latest",
  "restart": {
    "policy": "always"
  },
  "guest": {
    "cpu_kind": "shared",
    "cpus": 1,
    "memory_mb": 256
  },
  "dns": {}
}

Change a Machine’s restart policy with flyctl

Update the Machine config:

fly m update <machine id> --restart <no | always | on-fail>

The following example updates a Machine’s restart policy to on-fail:

fly m update 3908032c794088 --restart on-fail

Configuration changes to be applied to machine: 3908032c794088 (my-app-name)

      ... // 2 identical lines
        "image": "registry-1.docker.io/flyio/hellofly:latest",
        "restart": {
-      "policy": "always"
+      "policy": "on-failure"
        },
        "guest": {
      ... // 6 identical lines

? Apply changes? (y/N)

Enter y to apply the changes.

Change a Machine’s restart policy with the Machines API

With the Machines API, you can set the restart policy, and the maximum number of retries when the policy is on-failure.

Endpoint: POST /apps/{app_name}/machines/{machine_id}

For example:

...
    "restart": {
      "max_retries": 5,
      "policy": "on-failure"
    },
...

Refer to the Machines API docs for more information about updating a Machine.

Set a restart policy in your Fly.toml

You can also set a default app-level restart policy in your Fly.toml file:

[[restart]]
  policy = "<never | always | on-failure>"
  retries = 10
  processes = ["app"]

A restart policy can be targeted to a specific process group. If a group is not specified, all machines in an app will have the same default restart policy. If needed, you can still apply different policies on individual machines using the Flyctl or Machines API methods above.