Merge pull request #19 from tumblr/rc90

Rc90

Author: tbchrist

README.rdoc

@@ -75,6 +75,7 @@ If you have a question that isn't covered here, please feel free to email the au
 * <b>Dallas Marlow</b>: Developer
 * <b>Bob Patterson Jr</b>: Developer
 * <b>Tom Christ</b>: Developer
+* <b>Tyler Neely</b>: Developer
 
 Special thanks to <b>Tim Ellis</b> for testing and bug reports.
 
@@ -92,4 +93,4 @@ Unless required by applicable law or agreed to in writing, software
 distributed under the License is distributed on an "AS IS" BASIS,
 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
-limitations under the License.
+limitations under the License.

bin/jetpants

@@ -4,7 +4,7 @@
 # module in a Thor command processor, providing a command-line interface to
 # common Jetpants functionality.
 
-%w[thor pry highline/import colored].each {|g| require g}
+%w[thor pry pry-rescue highline/import colored].each {|g| require g}
 
 module Jetpants
 
@@ -222,15 +222,15 @@ module Jetpants
       end
 
       source.start_mysql if ! source.running?
-      error "source (#{source}) is not a standby slave" unless source.is_standby?
+      error "source (#{source}) is not a standby or backup slave" unless (source.is_standby? || source.for_backups?) 
 
       targets.each do |t|
         error "target #{t} already has a master; please clear out node (including in asset tracker) before proceeding" if t.master
         error "target #{t} is running a different version of MySQL than source #{source}! Cannot proceed with clone operation." if t.version_cmp(source) != 0
       end
 
       source.enslave_siblings!(targets)
-      targets.concurrent_each {|t| t.resume_replication; t.catch_up_to_master}
+      targets.concurrent_each {|t| t.resume_replication; t.catch_up_to_master(21600)}
       source.pool.sync_configuration
       puts "Cloning complete."
       Jetpants.topology.write_config
@@ -264,7 +264,12 @@ module Jetpants
       puts "This task turns an active slave into a standby slave."
       node = ask_node('Please enter node IP: ', options[:node])
       describe node
-      raise "Node is not an active slave" unless node.role == :active_slave
+      if node.running?
+        raise "Node is not an active slave" unless node.role == :active_slave
+      else
+        inform "Unable to connect to node #{node} to pull slave"
+        error  "Unable to pull slave" unless agree "Please confirm that #{node} is offline [yes/no]: "
+      end
       node.pool.mark_slave_standby(node)
       Jetpants.topology.write_config
     end
@@ -391,6 +396,8 @@ module Jetpants
     # (ie, if app configuration is a static file that needs to be deployed to webs.)
     desc 'shard_split_child_reads', 'shard split step 2 of 4: move reads to child shards'
     def shard_split_child_reads
+      s = ask_shard_being_split
+      s.move_reads_to_children
       Jetpants.topology.write_config
     end
     def self.after_shard_split_child_reads
@@ -453,8 +460,9 @@ module Jetpants
       # because the new last shard can't be created yet (chicken-and-egg problem -- master
       # must exist before we create the pool). The assumption is the hardware spec
       # of the new last shard and previous last shard will be the same.
-      raise "Not enough total spare machines!" unless Jetpants.topology.count_spares(like: last_shard_master) >= Jetpants.standby_slaves_per_pool + 1
-      raise "Not enough standby_slave role spare machines!" unless Jetpants.topology.count_spares(role: :standby_slave, like: last_shard_master) >= Jetpants.standby_slaves_per_pool
+      raise "Not enough total spare machines!" unless Jetpants.topology.count_spares(like: last_shard_master) >= last_shard.slaves_layout[:standby_slave] + last_shard.slaves_layout[:backup_slave] + 1
+      raise "Not enough standby_slave role spare machines!" unless Jetpants.topology.count_spares(role: :standby_slave, like: last_shard_master) >= last_shard.slaves_layout[:standby_slave]
+      raise "Not enough backup_slave role spare machines!" unless Jetpants.topology.count_spares(role: :backup_slave) >= last_shard.slaves_layout[:backup_slave]
       raise "Cannot find a spare master-role machine!" unless Jetpants.topology.count_spares(role: :master, like: last_shard_master) >= 1
 
       # In asset tracker, remove the last shard pool and replace it with a new pool. The new pool
@@ -470,15 +478,26 @@ module Jetpants
       # NOT to actually set the pool of the returned object.)
       new_last_shard_master = Jetpants.topology.claim_spare(role: :master, like: last_shard_master)
       new_last_shard_master.disable_read_only! if new_last_shard_master.running?
-      if Jetpants.standby_slaves_per_pool > 0
+      if last_shard.slaves_layout[:standby_slave] > 0
         # Verify spare count again, now that we can actually supply the new master as the :like context
-        raise "Not enough standby_slave role spare machines!" unless Jetpants.topology.count_spares(role: :standby_slave, like: new_last_shard_master) >= Jetpants.standby_slaves_per_pool
-        new_last_shard_slaves = Jetpants.topology.claim_spares(Jetpants.standby_slaves_per_pool, role: :standby_slave, like: new_last_shard_master)
+        raise "Not enough standby_slave role spare machines!" unless Jetpants.topology.count_spares(role: :standby_slave, like: new_last_shard_master) >= last_shard.slaves_layout[:standby_slave]
+        new_last_shard_slaves = Jetpants.topology.claim_spares(last_shard.slaves_layout[:standby_slave], role: :standby_slave, like: new_last_shard_master)
         new_last_shard_slaves.each do |x| 
           x.change_master_to new_last_shard_master
           x.resume_replication
         end
       end
+
+      # Set up backup slaves
+      if last_shard.slaves_layout[:backup_slave] > 0
+        raise "Not enough backup_slave role spare machines!" unless Jetpants.topology.count_spares(role: :backup_slave) >= last_shard.slaves_layout[:backup_slave]
+        new_last_shard_backup_slaves = Jetpants.topology.claim_spares(last_shard.slaves_layout[:backup_slave], role: :backup_slave)
+        new_last_shard_backup_slaves.each do |x| 
+          x.change_master_to new_last_shard_master
+          x.resume_replication
+        end
+      end
+
       new_last_shard = Shard.new(cutover_id, 'INFINITY', new_last_shard_master)
       new_last_shard.sync_configuration
       Jetpants.topology.pools << new_last_shard
@@ -498,7 +517,53 @@ module Jetpants
         'Deploy the configuration to all machines.',
       )
     end
-    
+
+    desc 'rebalance_backup_slaves', 'Add backup slaves to pools which contain too few (does not destroy existing slaves)'
+    def rebalance_backup_slaves
+      spares_available = Jetpants.topology.count_spares(role: :backup_slave)
+
+      raise "No backup slaves available" if spares_available == 0
+
+      # find shards that need a backup slave
+      need_backup_shards = Jetpants.topology.shards.reject { |shard| shard.backup_slaves.count >= shard.slaves_layout[:backup_slave] }
+
+      possible_iterations = spares_available < need_backup_shards.count ? spares_available : need_backup_shards.count
+
+      spares = Jetpants.topology.claim_spares(possible_iterations, role: :backup_slave)
+
+      need_backup_shards.first(possible_iterations)
+
+      # loop through and place a backup slave per pool that is without
+      need_backup_shards.limited_concurrent_map(5) do |shard|
+
+        if spares.count > 0
+
+          output "Considering shard #{shard}"
+
+          source = shard.standby_slaves.last
+          source.master.probe if source.master # fail early if there are any replication issues in this pool
+
+
+          targets = [spares.pop]
+
+          source.start_mysql if ! source.running?
+          error "Source (#{source}) is not a standby slave" unless source.is_standby?
+
+          targets.each do |t|
+            error "Target #{t} already has a master; please clear out node (including in asset tracker) before proceeding" if t.master
+            error "Target #{t} is running a different version of MySQL than source #{source}! Cannot proceed with clone operation." if t.version_cmp(source) != 0 
+            error "Target #{t} already has a pool!" if t.pool
+          end
+
+          source.enslave_siblings!(targets)
+          targets.concurrent_each {|t| t.resume_replication; t.catch_up_to_master(21600)}
+          source.pool.sync_configuration
+          output "Rebalance complete for #{shard}"
+
+        end
+
+      end
+    end
 
     no_tasks do
       def is_ip? address
@@ -516,7 +581,7 @@ module Jetpants
       def describe node
         puts "Node #{node} (#{node.hostname}:#{node.port}) has role #{node.role} in pool #{node.pool(true)}.".green
       end
-      
+
       def ask_node(prompt, supplied_node=false)
         node = supplied_node || ask(prompt)
         error "Node (#{node}) does not appear to be an IP address." unless is_ip? node
@@ -540,7 +605,7 @@ module Jetpants
           s = Jetpants.topology.shard shard_min, shard_max
           raise "Shard not found" unless s
         end
-        raise "Shard isn't in expected state" unless s.state == :deprecated
+        raise "Shard does not have children" if (s.children.nil? or s.children.count ==0)
         s
       end
     end
@@ -558,4 +623,20 @@ end
 # We load jetpants last so that plugins can monkeypatch Jetpants::CommandSuite if desired.
 require 'jetpants'
 
-Jetpants::CommandSuite.start
+def with_debug
+  Pry::rescue do
+    begin
+      yield if block_given?
+    rescue => e
+      Pry.config.prompt = proc {|_, _, _| "# debug > "}
+      puts "Entering debugging session (debug_exceptions is enabled).  Resume with ctrl+d."
+      Pry::rescued e
+    end
+  end
+end
+
+if Jetpants.debug_exceptions
+  with_debug { Jetpants::CommandSuite.start }
+else
+  Jetpants::CommandSuite.start
+end

etc/jetpants.yaml.sample

@@ -21,6 +21,13 @@ mysql_grant_privs:
     - LOCK TABLES
     - EXECUTE
 
+# Specific schemata may be optionally ignored while cloning.  If this is not
+# overridden it will default to information_schema and performance_schema.
+mysql_clone_ignore:
+    - some_temporary_schema
+    - information_schema
+    - performance_schema
+
 # Define what directory to put exported files in. If you have /var/lib/mysql
 # on an SSD RAID mount and the rest of your OS on a rotational disk mount,
 # it's fine to put the export_location on the rotational disk, which probably

jetpants.gemspec

@@ -2,13 +2,13 @@ require 'rake'
 
 Gem::Specification.new do |s|
   s.name = "jetpants"
-  s.version = "0.8.3"
+  s.version = "0.9.0"
 
   s.homepage = 'https://github.com/tumblr/jetpants/'
   s.required_rubygems_version = Gem::Requirement.new(">= 0") if s.respond_to? :required_rubygems_version=
   s.required_ruby_version = '>= 1.9.2'
   s.authors = ["Evan Elias", "Dallas Marlow", "Bob Patterson Jr.", "Tom Christ"]
-  s.date  = "2013-05-17"
+  s.date  = "2014-06-09"
   s.email = ["[email protected]", "[email protected]", "[email protected]", "[email protected]"]
   s.files = FileList['Gemfile', 'README.rdoc', 'doc/*.rdoc', 'lib/**/*.rb', 'bin/**', 'scripts/*.rb', 'plugins/**/*.rb', 'etc/jetpants.yaml.sample'].to_a
   s.require_paths = ["lib"]
@@ -24,8 +24,10 @@ Gem::Specification.new do |s|
   s.add_runtime_dependency 'sequel', '~> 3.36'
   s.add_runtime_dependency 'net-ssh', '~> 2.3'
   s.add_runtime_dependency 'pry', '~> 0.9.8'
+  s.add_runtime_dependency 'pry-rescue', '~> 1.4.0'
   s.add_runtime_dependency 'thor', '~> 0.15'
   s.add_runtime_dependency 'highline', '~> 1.6.12'
   s.add_runtime_dependency 'colored', '~> 1.2'
   s.add_runtime_dependency 'collins_client', '~> 0.2.7'
+  s.add_runtime_dependency 'bloom-filter', '~> 0.2.0'
 end

lib/jetpants.rb

@@ -23,6 +23,7 @@ module Jetpants
   @config = {
     'max_concurrency'         =>  20,         # max threads/conns per database
     'standby_slaves_per_pool' =>  2,          # number of standby slaves in every pool
+    'backup_slaves_per_pool'  =>  1,          # number of backup slaves in every pool
     'mysql_schema'            =>  'test',     # database name
     'mysql_app_user'          =>  'appuser',  # mysql user for application
     'mysql_app_password'      =>  '',         # mysql password for application
@@ -31,6 +32,7 @@ module Jetpants
     'mysql_root_password'     =>  false,      # mysql root password. omit if specified in /root/.my.cnf instead.
     'mysql_grant_ips'         =>  ['192.168.%'],  # mysql user manipulations are applied to these IPs
     'mysql_grant_privs'       =>  ['ALL'],    # mysql user manipulations grant this set of privileges by default
+    'mysql_clone_ignore'      =>  ['information_schema', 'performance_schema'], # these schemata will be ignored during cloning
     'export_location'         =>  '/tmp',     # directory to use for data dumping
     'verify_replication'      =>  true,       # raise exception if the 2 repl threads are in different states, or if actual repl topology differs from Jetpants' understanding of it
     'plugins'                 =>  {},         # hash of plugin name => arbitrary plugin data (usually a nested hash of settings)
@@ -39,6 +41,8 @@ module Jetpants
     'compress_with'           =>  false,      # command line to use for compression in large file transfers
     'decompress_with'         =>  false,      # command line to use for decompression in large file transfers
     'private_interface'       =>  'bond0',    # network interface corresponding to private IP
+    'output_caller_info'      =>  false,      # includes calling file, line and method in output calls
+    'debug_exceptions'        =>  false,      # open a pry session when an uncaught exception is thrown
   }
 
   config_paths = ["/etc/jetpants.yaml", "~/.jetpants.yml", "~/.jetpants.yaml"]

lib/jetpants/db.rb

@@ -12,6 +12,7 @@ module Jetpants
   # functional lines.
   class DB
     include CallbackHandler
+    include Output
 
     # IP address (as a string) of the MySQL instance
     attr_reader :ip
@@ -70,6 +71,9 @@ def initialize(ip, port=3306)
       # This is ephemeral, only known to Jetpants if you previously called
       # DB#start_mysql or DB#restart_mysql in this process
       @options = []
+
+      # Mutex used to lock probing state
+      @probe_mutex = Mutex.new
     end
 
     ###### Host methods ########################################################
@@ -97,19 +101,6 @@ def same_host_as?(db)
 
     ###### Misc methods ########################################################
 
-    # Displays the provided output, along with information about the current time,
-    # self, and optionally a Jetpants::Table name.
-    def output(str, table=nil)
-      str = str.to_s.strip
-      str = nil if str && str.length == 0
-      str ||= "Completed (no output)"
-      output = Time.now.strftime("%H:%M:%S") + " [#{self}] "
-      output << table.name << ': ' if table
-      output << str
-      print output + "\n"
-      output
-    end
-    
     # DB objects are sorted as strings, ie, by calling to_s
     def <=> other
       to_s <=> other.to_s

lib/jetpants/db/import_export.rb

@@ -325,13 +325,22 @@ def clone_to!(*targets)
         destinations[t] = t.mysql_directory
         raise "Over 100 MB of existing MySQL data on target #{t}, aborting copy!" if t.data_set_size > 100000000
       end
-      [self, targets].flatten.concurrent_each {|t| t.stop_query_killer; t.stop_mysql}
-      targets.concurrent_each {|t| t.ssh_cmd "rm -rf #{t.mysql_directory}/ib_logfile*"}
 
       # Construct the list of files and dirs to copy. We include ib_lru_dump if present
       # (ie, if using Percona Server with innodb_buffer_pool_restore_at_startup enabled)
       # since this will greatly improve warm-up time of the cloned nodes
-      files = ['ibdata1', 'mysql', 'test', app_schema]
+      databases = mysql_root_cmd("SHOW DATABASES").split("\n").select { |row|
+        row.include?('Database:')
+      }.map{ |line|
+        line.split(":").last.strip
+      }.reject { |s|
+        Jetpants.mysql_clone_ignore.include? s
+      }
+
+      [self, targets].flatten.concurrent_each {|t| t.stop_query_killer; t.stop_mysql}
+      targets.concurrent_each {|t| t.ssh_cmd "rm -rf #{t.mysql_directory}/ib_logfile*"}
+
+      files = (databases + ['ibdata1', app_schema]).uniq
       files << 'ib_lru_dump' if ssh_cmd("test -f #{mysql_directory}/ib_lru_dump 2>/dev/null; echo $?").chomp.to_i == 0
 
       fast_copy_chain(mysql_directory, 
@@ -343,4 +352,4 @@ def clone_to!(*targets)
     end
 
   end
-end
+end

lib/jetpants/db/replication.rb

@@ -164,7 +164,7 @@ def enslave_siblings!(targets)
         t.enable_read_only!
       end
       resume_replication # should already have happened from the clone_to! restart anyway, but just to be explicit
-      catch_up_to_master
+      catch_up_to_master 21600
       enable_monitoring
     end
 
@@ -209,7 +209,7 @@ def seconds_behind_master
     # returns true if slave lag is zero 3 times in a row. Gives up if this does
     # not occur within a one-hour period. If a large amount of slave lag is
     # reported, this method will automatically reduce its polling frequency.
-    def catch_up_to_master(timeout=3600, threshold=3, poll_frequency=5)
+    def catch_up_to_master(timeout=21600, threshold=3, poll_frequency=5)
       raise "This instance is not a slave" unless master
       resume_replication if @repl_paused
 

lib/jetpants/db/server.rb

@@ -108,5 +108,16 @@ def enable_monitoring(*services)
     def disable_monitoring(*services)
     end
 
+    # No built-in effect. Use when performing actions which will cause the
+    # server to go offline or become unresponsive, as an escalated enable_monitoring
+    # Plugins can override and/or implement before/after hooks
+    def set_downtime(hours)
+    end
+
+    # No built-in effect. Use when performing actions which will cause the
+    # server to go offline or become unresponsive, as an escalated disable_monitoring
+    # Plugins can override and/or implement before/after hooks
+    def cancel_downtime
+    end
   end
 end