bosh_aws_cpi 1.2905.0 → 1.2915.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 632acbb4fb95e22626dd45a7298d4cc60c500f7f
-  data.tar.gz: fa7198502d1716dbf7cb8f24b718a45a7c8e919b
+  metadata.gz: c22ae9747a085884024355d95211b787433c092b
+  data.tar.gz: 0d2512c06b0de933e4a9eb2ad2986c63db30998b
 SHA512:
-  metadata.gz: 1bcffafb2072a71739a19cd3d052201e60df33db20847b1f1c889e822aba280467ddd4582a9349336b1e81e55261f5ef3306e3426c5ad802d354df98d7e3d3a3
-  data.tar.gz: e2715f4b35c9cd59ce6a7ac57f4c68bebd6e51ec74b25c57ad4a771f5da2c21bd379f14024e5765c661f21a27d979adb15a0a7026baea5653e35fdae9393bd46
+  metadata.gz: 3d5476682f1ab3435ce1c31184b43614cd581ec48235c661502ce0beecc43264fddd7b27a19c4c9810c4d2adf71c6fb004fcb8909c8b32654f914445c34aa8ed
+  data.tar.gz: 8d75ae99b097651d136dd3c24fa92eaabba49a12731e8c3263f2c7ca7ca205b08fb78d404319c9dacec1f2da5601e6e51e88f0178f16f70a8ed0f8fd61380f38

data/lib/cloud/aws/cloud.rb

@@ -205,14 +205,20 @@ module Bosh::AwsCloud
 
         logger.info("Deleting volume `#{volume.id}'")
 
-        tries = 10
-        sleep_cb = ResourceWait.sleep_callback("Waiting for volume `#{volume.id}' to be deleted", tries)
+        # Retry 1, 6, 11, 15, 15, 15.. seconds. The total time is ~10 min.
+        # VolumeInUse can be returned by AWS if disk was attached to VM
+        # that was recently removed.
+        tries = ResourceWait::DEFAULT_WAIT_ATTEMPTS
+        sleep_cb = ResourceWait.sleep_callback(
+          "Waiting for volume `#{volume.id}' to be deleted",
+          { interval: 5, total: tries }
+        )
         ensure_cb = Proc.new do |retries|
           cloud_error("Timed out waiting to delete volume `#{volume.id}'") if retries == tries
         end
-        error = AWS::EC2::Errors::Client::VolumeInUse
+        errors = [AWS::EC2::Errors::VolumeInUse, AWS::EC2::Errors::RequestLimitExceeded]
 
-        Bosh::Common.retryable(tries: tries, sleep: sleep_cb, on: error, ensure: ensure_cb) do
+        Bosh::Common.retryable(tries: tries, sleep: sleep_cb, on: errors, ensure: ensure_cb) do
           volume.delete
           true # return true to only retry on Exceptions
         end
@@ -484,7 +490,6 @@ module Bosh::AwsCloud
       cloud_error("Cannot find EBS volume on current instance")
     end
 
-
     private
 
     attr_reader :az_selector
@@ -573,11 +578,35 @@ module Bosh::AwsCloud
       # AWS might still lie and say that the disk isn't ready yet, so
       # we try again just to be really sure it is telling the truth
       attachment = nil
-      Bosh::Common.retryable(tries: 15, on: AWS::EC2::Errors::IncorrectState) do
+
+      logger.debug("Attaching '#{volume.id}' to '#{instance.id}' as '#{device_name}'")
+
+      # Retry every 1 sec for 15 sec, then every 15 sec for ~10 min
+      # VolumeInUse can be returned by AWS if disk was attached to VM
+      # that was recently removed.
+      tries = ResourceWait::DEFAULT_WAIT_ATTEMPTS
+      sleep_cb = ResourceWait.sleep_callback(
+        "Attaching volume `#{volume.id}' to #{instance.id}",
+        { interval: 0, tries_before_max: 15, total: tries }
+      )
+
+      Bosh::Common.retryable(
+        on: [
+          AWS::EC2::Errors::IncorrectState,
+          AWS::EC2::Errors::VolumeInUse,
+          AWS::EC2::Errors::RequestLimitExceeded
+        ],
+        sleep: sleep_cb,
+        tries: tries
+      ) do |retries, error|
+        # Continue to retry after 15 attempts only for VolumeInUse
+        if retries > 15 && error.instance_of?(AWS::EC2::Errors::IncorrectState)
+          cloud_error("Failed to attach disk: #{error.message}")
+        end
+
         attachment = volume.attach_to(instance, device_name)
       end
 
-      logger.info("Attaching '#{volume.id}' to '#{instance.id}' as '#{device_name}'")
       ResourceWait.for_attachment(attachment: attachment, state: :attached)
 
       device_name = attachment.device

data/lib/cloud/aws/resource_wait.rb

@@ -7,7 +7,8 @@ module Bosh::AwsCloud
     # a sane amount of retries on AWS (~25 minutes),
     # as things can take anywhere between a minute and forever
     DEFAULT_TRIES = 54
-    MAX_SLEEP_EXPONENT = 5
+    MAX_SLEEP_TIME = 15
+    DEFAULT_WAIT_ATTEMPTS = 600 / MAX_SLEEP_TIME # 10 minutes
 
     def self.for_instance(args)
       raise ArgumentError, "args should be a Hash, but `#{args.class}' given" unless args.is_a?(Hash)
@@ -134,12 +135,21 @@ module Bosh::AwsCloud
       end
     end
 
-    def self.sleep_callback(description, tries)
+    def self.sleep_callback(description, options)
+      max_sleep_time = options.fetch(:max, MAX_SLEEP_TIME)
       lambda do |num_tries, error|
-        sleep_time = 2**[num_tries, MAX_SLEEP_EXPONENT].min # Exp backoff: 1, 2, 4, 8 ... up to max 32
+        if options[:tries_before_max] && num_tries >= options[:tries_before_max]
+          time = max_sleep_time
+        else
+          if options[:exponential]
+            time = [options[:interval] ** num_tries, max_sleep_time].min
+          else
+            time = [1 + options[:interval] * num_tries, max_sleep_time].min
+          end
+        end
         Bosh::AwsCloud::ResourceWait.logger.debug("#{error.class}: `#{error.message}'") if error
-        Bosh::AwsCloud::ResourceWait.logger.debug("#{description}, retrying in #{sleep_time} seconds (#{num_tries}/#{tries})")
-        sleep_time
+        Bosh::AwsCloud::ResourceWait.logger.debug("#{description}, retrying in #{time} seconds (#{num_tries}/#{options[:total]})")
+        time
       end
     end
 
@@ -163,7 +173,10 @@ module Bosh::AwsCloud
       tries = args.fetch(:tries, DEFAULT_TRIES).to_i
       target_state = args.fetch(:target_state)
 
-      sleep_cb = self.class.sleep_callback("Waiting for #{desc} to be #{target_state}", tries)
+      sleep_cb = self.class.sleep_callback(
+        "Waiting for #{desc} to be #{target_state}",
+        { interval: 2, total: tries, max: 32, exponential: true }
+      )
       errors << AWS::EC2::Errors::RequestLimitExceeded
       ensure_cb = Proc.new do |retries|
         cloud_error("Timed out waiting for #{desc} to be #{target_state}, took #{time_passed}s") if retries == tries

data/lib/cloud/aws/version.rb

@@ -1,5 +1,5 @@
 module Bosh
   module AwsCloud
-    VERSION = '1.2905.0'
+    VERSION = '1.2915.0'
   end
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: bosh_aws_cpi
 version: !ruby/object:Gem::Version
-  version: 1.2905.0
+  version: 1.2915.0
 platform: ruby
 authors:
 - VMware
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2015-03-27 00:00:00.000000000 Z
+date: 2015-04-07 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: aws-sdk
@@ -30,42 +30,42 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 1.2905.0
+        version: 1.2915.0
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 1.2905.0
+        version: 1.2915.0
 - !ruby/object:Gem::Dependency
   name: bosh_cpi
   requirement: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 1.2905.0
+        version: 1.2915.0
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 1.2905.0
+        version: 1.2915.0
 - !ruby/object:Gem::Dependency
   name: bosh-registry
   requirement: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 1.2905.0
+        version: 1.2915.0
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 1.2905.0
+        version: 1.2915.0
 - !ruby/object:Gem::Dependency
   name: httpclient
   requirement: !ruby/object:Gem::Requirement
@@ -138,7 +138,7 @@ dependencies:
         version: '0'
 description: |-
   BOSH AWS CPI
-  327a58
+  5653ce
 email: support@cloudfoundry.com
 executables:
 - aws_cpi