Merge pull request #74 from dblock/ping-worker

Added server ping worker.

Author: dblock

.rubocop_todo.yml

@@ -1,6 +1,6 @@
 # This configuration was generated by
 # `rubocop --auto-gen-config`
-# on 2018-05-06 23:29:13 -0400 using RuboCop version 0.55.0.
+# on 2018-08-21 19:45:36 -0400 using RuboCop version 0.55.0.
 # The point is for the user to remove these configuration records
 # one by one as the offenses are removed from the code base.
 # Note that changes in the inspected code, or installation of new

CHANGELOG.md

@@ -1,10 +1,11 @@
 ### Changelog
 
-#### 0.6.2 (Next)
+#### 0.7.0 (Next)
 
 * [#60](https://github.com/slack-ruby/slack-ruby-bot-server/pull/60): Log caught Standard::Error backtrace at debug-level - [@alexagranov](https://github.com/alexagranov).
 * [#65](https://github.com/slack-ruby/slack-ruby-bot-server/pull/65): Updated Capybara and selenium-webdriver - [@dblock](https://github.com/dblock).
 * [#67](https://github.com/slack-ruby/slack-ruby-bot-server/pull/67): Only load the OTR::ActiveRecord::ConnectionManagement middleware when the OTR module is included. This module isn't needed when using Rails - [@darbyfrey](https://github.com/darbyfrey).
+* [#74](https://github.com/slack-ruby/slack-ruby-bot-server/pull/74): Added ping worker, will restart offline bots - [@dblock](https://github.com/dblock).
 * Your contribution here.
 
 #### 0.6.1 (3/29/2017)

README.md

@@ -152,6 +152,22 @@ SlackRubyBotServer.configure do |config|
 end
 ```
 
+#### Ping Worker
+
+Each `SlackRubyBotServer::Server` instance will start a ping worker that will periodically check the online status of the bot via the Slack web [auth.test](https://api.slack.com/methods/auth.test) and [users.getPresence](https://api.slack.com/methods/users.getPresence) APIs, and will forcefully close the connection if the bot goes offline, causing an automatic reconnect.
+
+You can configure the ping interval, number of retries before restart, and disable the ping worker entirely.
+
+```ruby
+SlackRubyBotServer.configure do |config|
+  config.ping = {
+    enabled: true, # set to false to disable the ping worker
+    ping_interval: 30, # interval in seconds
+    retry_count: 3 # number of unsuccessful retries until a restart
+  }
+end
+```
+
 ### Access Tokens
 
 By default the implementation of [Team](lib/slack-ruby-bot-server/models/team) stores a `bot_access_token` that grants a certain amount of privileges to the bot user as described in [Slack OAuth Docs](https://api.slack.com/docs/oauth). You may not want a bot user at all, or may require different auth scopes, such as `users.profile:read` to access user profile information via `Slack::Web::Client#users_profile_get`. To obtain the non-bot access token make the following changes.

UPGRADING.md

@@ -1,6 +1,24 @@
 Upgrading Slack-Ruby-Bot-Server
 ===============================
 
+### Upgrading to >= 0.7.0
+
+#### New Ping Worker
+
+Version 0.7.0 will automatically start a ping worker that checks for the bot's online status and forcefully terminate and restart disconnected bots. Set the ping `enabled` option to `false` to disable this behavior.
+
+```ruby
+SlackRubyBotServer.configure do |config|
+  config.ping = {
+    enabled: false
+  }
+end
+```
+
+If you are currently using a custom ping worker as suggested in [slack-ruby-client#208](https://github.com/slack-ruby/slack-ruby-client/issues/208), delete it.
+
+See [#74](https://github.com/slack-ruby/slack-ruby-bot-server/pull/74) for more information.
+
 ### Upgrading to >= 0.6.0
 
 #### Mongoid and ActiveRecord support

lib/slack-ruby-bot-server.rb

@@ -3,6 +3,7 @@
 require 'grape-swagger'
 require 'slack-ruby-bot'
 require 'slack-ruby-bot-server/server'
+require 'slack-ruby-bot-server/ping'
 require 'slack-ruby-bot-server/config'
 
 require 'slack-ruby-bot-server/ext'

lib/slack-ruby-bot-server/api/middleware.rb

@@ -8,7 +8,7 @@ module Api
     class Middleware
       def self.logger
         @logger ||= begin
-          $stdout.sync = true
+          STDOUT.sync = true
           Logger.new(STDOUT)
         end
       end

lib/slack-ruby-bot-server/app.rb

@@ -18,7 +18,7 @@ def self.instance
 
     def logger
       @logger ||= begin
-        $stdout.sync = true
+        STDOUT.sync = true
         Logger.new(STDOUT)
       end
     end

lib/slack-ruby-bot-server/config.rb

@@ -3,9 +3,11 @@ module Config
     extend self
 
     attr_accessor :server_class
+    attr_accessor :ping
     attr_accessor :database_adapter
 
     def reset!
+      self.ping = nil
       self.server_class = SlackRubyBotServer::Server
       self.database_adapter = if defined?(::Mongoid)
                                 :mongoid

lib/slack-ruby-bot-server/ping.rb

@@ -0,0 +1,99 @@
+module SlackRubyBotServer
+  class Ping
+    include Celluloid
+
+    attr_reader :client
+    attr_reader :options
+    attr_reader :error_count
+
+    def initialize(client, options = {})
+      @options = options
+      @client = client
+      @error_count = 0
+    end
+
+    def start!
+      every ping_interval do
+        check!
+      end
+    end
+
+    private
+
+    def check!
+      if online?
+        @error_count = 0
+      else
+        down!
+      end
+    rescue StandardError => e
+      case e.message
+      when 'account_inactive', 'invalid_auth' then
+        logger.warn "Error pinging team #{owner.id}: #{e.message}, terminating."
+        terminate
+      else
+        logger.warn "Error pinging team #{owner.id}: #{e.message}."
+      end
+    end
+
+    def offline?
+      !online?
+    end
+
+    def online?
+      presence.online
+    end
+
+    def presence
+      owner.ping![:presence]
+    end
+
+    def down!
+      logger.warn "DOWN: #{owner}, #{retries_left} #{retries_left == 1 ? 'retry' : 'retries'} left"
+      @error_count += 1
+      return if retries_left?
+      restart!
+    end
+
+    def restart!
+      logger.warn "RESTART: #{owner}"
+      driver.emit(:close, WebSocket::Driver::CloseEvent.new(1001, 'bot offline')) if driver
+      terminate
+    end
+
+    def ping_interval
+      options[:ping_interval] || 60
+    end
+
+    def retries_left?
+      retries_left >= 0
+    end
+
+    def retries_left
+      retry_count - error_count
+    end
+
+    def retry_count
+      options[:retry_count] || 2
+    end
+
+    def socket
+      client.instance_variable_get(:@socket) if client
+    end
+
+    def driver
+      socket.instance_variable_get(:@driver) if socket
+    end
+
+    def logger
+      @logger ||= begin
+        STDOUT.sync = true
+        Logger.new(STDOUT)
+      end
+    end
+
+    def owner
+      client.owner
+    end
+  end
+end

lib/slack-ruby-bot-server/server.rb

@@ -7,18 +7,36 @@ class Server < SlackRubyBot::Server
     def initialize(attrs = {})
       attrs = attrs.dup
       @team = attrs.delete(:team)
+      @ping_options = attrs.delete(:ping) || {}
       raise 'Missing team' unless @team
       attrs[:token] = @team.token
       super(attrs)
-      client.owner = @team
+      open!
     end
 
     def restart!(wait = 1)
       # when an integration is disabled, a live socket is closed, which causes the default behavior of the client to restart
       # it would keep retrying without checking for account_inactive or such, we want to restart via service which will disable an inactive team
       logger.info "#{team.name}: socket closed, restarting ..."
       SlackRubyBotServer::Service.instance.restart! team, self, wait
+      open!
+    end
+
+    private
+
+    attr_reader :ping_options
+
+    def create_ping
+      return unless !ping_options.key?(:enabled) || ping_options[:enabled]
+      SlackRubyBotServer::Ping.new(client, ping_options)
+    end
+
+    def open!
       client.owner = team
+      client.on :open do |_event|
+        worker = create_ping
+        worker.start! if worker
+      end
     end
   end
 end