WIP: AASM approach sketch

Gemfile

@@ -2,6 +2,7 @@ source "https://rubygems.org"
 
 ruby "3.2.0"
 
+gem "aasm"
 gem "bootsnap", require: false
 gem "importmap-rails"
 gem "net-ssh"

Gemfile.lock

@@ -1,6 +1,8 @@
 GEM
   remote: https://rubygems.org/
   specs:
+    aasm (5.5.0)
+      concurrent-ruby (~> 1.0)
     actioncable (7.1.3.2)
       actionpack (= 7.1.3.2)
       activesupport (= 7.1.3.2)
@@ -239,6 +241,7 @@ PLATFORMS
   x86_64-linux
 
 DEPENDENCIES
+  aasm
   bootsnap
   cloudflare-rails
   debug

app/controllers/configs_controller.rb

@@ -11,12 +11,11 @@ def show
     server = Server.find_by_ip!(ip)
 
     if server.machine_config
-      # Using 1.second.ago seems silly but is required to show the correct status in the UI at the moment
-      server.update!(last_request_for_configuration_at: 1.second.ago, last_configured_at: Time.now)
+      server.configure!
       headers["Content-Type"] = "text/yaml"
       render plain: server.machine_config.generate_config
     else
-      server.update!(last_configured_at: nil, last_request_for_configuration_at: Time.now)
+      server.request_configuration!
       render plain: "No configuration found for server with IP #{server.ip}", status: 420
     end
   end

app/controllers/servers_controller.rb

@@ -42,7 +42,7 @@ def bootstrap
   def rescue
     server = Server.find(params[:id])
 
-    server.rescue
+    server.rescue!
 
     redirect_to servers_path, notice: "Server #{server.name} is rebooting in rescue mode"
   end

app/models/server.rb

@@ -7,6 +7,61 @@ class Server < ApplicationRecord
   has_one :machine_config, dependent: :destroy
   has_one :config, through: :machine_config
 
+  include AASM
+
+  aasm timestamps: true do
+    state :pending, initial: true
+    state :rescuing
+    state :failed_rescue
+    state :rescued
+    state :installing_talos
+    state :awaiting_configuration
+    state :configuring
+    state :configured
+    state :ready
+    state :inaccessible
+
+    event :rescue, before: :before_rescue do
+      transitions to: :rescuing
+    end
+
+    event :await_rescued do
+      transitions from: %i[rescuing failed_rescue], to: :rescued, guard: :bootstrappable?
+      transitions from: :rescuing, to: :failed_rescue, guard: :rescue_timeout?
+    end
+
+    event :install_talos, before: :bootstrap! do
+      transitions from: :rescued, to: :installing_talos
+    end
+
+    event :request_configuration do
+      before do
+        assign_attributes(last_configured_at: nil, last_request_for_configuration_at: Time.now)
+      end
+
+      transitions to: :awaiting_configuration
+    end
+
+    event :configure do
+      before do
+        # 1.second.ago seems silly but is required to show the correct status in the UI at the moment
+        assign_attributes(last_configured_at: Time.now, last_request_for_configuration_at: 1.second.ago)
+      end
+
+      transitions to: :configuring
+    end
+
+    event :await_configured do
+      transitions from: %i[configuring failed_configure], to: :configured, guard: :talos_accessible?
+      transitions from: :configuring, to: :failed_configure, guard: :configure_timeout?
+    end
+
+    event :await_ready do
+      transitions from: %i[configured failed_ready], to: :ready, guard: :ready?
+      transitions from: :configured, to: :failed_ready, guard: :ready_timeout?
+    end
+  end
+
   attr_accessor :sync # set to true to sync changed attributes to hetzner
 
   validates_uniqueness_of :name, allow_nil: true
@@ -85,8 +140,8 @@ def bootstrap!
     update!(accessible: false)
   end
 
-  def rescue
-    raise "#rescue is not implemented for #{self.class.name}"
+  def before_rescue
+    raise "#before_rescue is not implemented for #{self.class.name}"
   end
 
   def reset

app/models/server/hetzner_cloud.rb

@@ -17,7 +17,7 @@ def bootstrappable?
     session&.shutdown!
   end
 
-  def rescue
+  def before_rescue
     ::HetznerCloud.active_rescue_system(id)
 
     # Hetzner appears to temporarily lock the server immediately after enabling rescue mode.

app/models/server/hetzner_dedicated.rb

@@ -17,7 +17,7 @@ def bootstrappable?
     session&.shutdown!
   end
 
-  def rescue
+  def before_rescue
     Hetzner.active_rescue_system(id)
 
     if Hetzner.reset_state(id).fetch("operating_status") == "shut off"