#Fix Win 2019 issue

To run program on windows all files are copied in Windows\Temp folder that may occasionally be cleaned up. When certs are deleted, poller cannot verify call and raises exception.

# Why is this change needed?
Customers are suffering from "Agent occasionally stops working" issue

# How does it address the issue?
Propagate certificate exception, so win service can  handle it properly. Win service tries to copy certificates from PROGRAMDATA folder, so agent can continue to work properly

# How was this tested ?
bb release

cr https://code.amazon.com/reviews/CR-19039551

Author: Alena Kastsiukavets

lib/instance_agent/agent/base.rb

@@ -27,6 +27,10 @@ def run
         begin
           perform
           @error_count = 0
+        rescue Seahorse::Client::NetworkingError => e
+          log(:error, "Failed to execute the command. Your certificates might have been deleted" )
+          # TODO: verify error message is "certificate verify failed"
+          raise e
         rescue Aws::Errors::MissingCredentialsError
           log(:error, "Missing credentials - please check if this instance was started with an IAM instance profile")
           @error_count = @error_count.to_i + 1

lib/winagent.rb

@@ -17,8 +17,10 @@
 
 class InstanceAgentService < Daemon
 
-  def initialize 
+  def initialize
+    @app_root_folder = File.join(ENV['PROGRAMDATA'], "Amazon/CodeDeploy")
     InstanceAgent::Platform.util = InstanceAgent::WindowsUtil
+
     cert_dir = File.expand_path(File.join(File.dirname(__FILE__), '..\certs'))
     Aws.config[:ssl_ca_bundle] = File.join(cert_dir, 'ca-bundle.crt')
     ENV['AWS_SSL_CA_DIRECTORY'] = File.join(cert_dir, 'ca-bundle.crt')
@@ -32,6 +34,7 @@ def description
 
   def service_main
     read_config
+    @attempt_count = 0
     log(:info, 'started')
      shutdown_flag = false
      while running? && !shutdown_flag
@@ -73,8 +76,7 @@ def expand_conf_path(key)
   end 
 
   def read_config
-    default_root = File.join(ENV['PROGRAMDATA'], "Amazon/CodeDeploy")
-    default_config = File.join(default_root, "conf.yml")
+    default_config = File.join(@app_root_folder, "conf.yml")
     InstanceAgent::Config.config({:config_file => default_config,
             :on_premises_config_file => File.join(default_root, "conf.onpremises.yml")})
     InstanceAgent::Config.load_config
@@ -87,6 +89,16 @@ def read_config
 
   def with_error_handling
     yield
+  rescue Seahorse::Client::NetworkingError => e
+    @attempt_count = @attempt_count + 1
+    if @attempt_count > 3
+      log(:error, "Failed to recover after certificate issue:" + e.inspect)
+      exit
+    end
+    log(:error, "Custom:" + e.inspect)
+    # try to copy certs from application root folder
+    @certs_backup_folder = File.join(@app_root_folder, "certs/.")
+    FileUtils.cp_r(@certs_backup_folder, @cert_dir)
   rescue SocketError => e
     log(:info, "#{description}: failed to run as the connection failed! #{e.class} - #{e.message} - #{e.backtrace.join("\n")}")
     sleep InstanceAgent::Config.config[:wait_after_connection_problem]