Class: Switchman::Shard

Inherits:
UnshardedRecord show all
Defined in:
lib/switchman/shard.rb

Constant Summary collapse

IDS_PER_SHARD =

ten trillion possible ids per shard. yup.

10_000_000_000_000
NIL_NIL_ID =

takes an id-ish, and returns a local id and the shard it’s local to. [nil, nil] if it can’t be interpreted. [id, nil] if it’s already a local ID. [nil, nil] if it’s a well formed id, but the shard it refers to does not exist

[nil, nil].freeze

Class Method Summary collapse

Instance Method Summary collapse

Class Method Details

._load(str) ⇒ Object 



600
601
602
 
# File 'lib/switchman/shard.rb', line 600

def self._load(str)
  lookup(str.to_i)
end

.activate(shards) ⇒ Object 



103
104
105
106
107
108
109
110
 
# File 'lib/switchman/shard.rb', line 103

def activate(shards)
  activated_classes = activate!(shards)
  yield
ensure
  activated_classes&.each do |klass|
    klass.connected_to_stack.pop
  end
end

.activate!(shards) ⇒ Object 



112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
 
# File 'lib/switchman/shard.rb', line 112

def activate!(shards)
  activated_classes = nil
  shards.each do |klass, shard|
    next if klass == UnshardedRecord

    next unless klass.current_shard != shard.database_server.id.to_sym ||
                klass.current_switchman_shard != shard

    (activated_classes ||= []) << klass
    klass.connected_to_stack << { shard: shard.database_server.id.to_sym,
                                  klasses: [klass],
                                  switchman_shard: shard }
  end
  activated_classes
end

.active_shards ⇒ Object 



128
129
130
131
132
 
# File 'lib/switchman/shard.rb', line 128

def active_shards
  sharded_models.filter_map do |klass|
    [klass, current(klass)]
  end.to_h
end

.clear_cache ⇒ Object 



155
156
157
 
# File 'lib/switchman/shard.rb', line 155

def clear_cache
  cached_shards.clear
end

.current(klass = ::ActiveRecord::Base) ⇒ Object 



98
99
100
101
 
# File 'lib/switchman/shard.rb', line 98

def current(klass = ::ActiveRecord::Base)
  klass ||= ::ActiveRecord::Base
  klass.current_switchman_shard
end

.default(reload: false, with_fallback: false) ⇒ Object 



62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
 
# File 'lib/switchman/shard.rb', line 62

def default(reload: false, with_fallback: false)
  if !@default || reload
    # Have to create a dummy object so that several key methods still work
    # (it's easier to do this in one place here, and just assume that sharding
    # is up and running everywhere else).  This includes for looking up the
    # default shard itself. This also needs to be a local so that this method
    # can be re-entrant
    default = DefaultShard.instance

    # if we already have a default shard in place, and the caller wants
    # to use it as a fallback, use that instead of the dummy instance
    default = @default if with_fallback && @default

    # the first time we need a dummy dummy for re-entrancy to avoid looping on ourselves
    @default ||= default

    # Now find the actual record, if it exists
    @default = begin
      find_cached("default_shard") { Shard.where(default: true).take } || default
    rescue
      default
    end

    # make sure this is not erroneously cached
    if @default.database_server.instance_variable_defined?(:@primary_shard)
      @default.database_server.remove_instance_variable(:@primary_shard)
    end

    # and finally, check for cached references to the default shard on the existing connection
    sharded_models.each do |klass|
      klass.connection.shard = @default if klass.connected? && klass.connection.shard.default?
    end
  end
  @default
end

.drop_database(name) ⇒ Object

Drops a specific database/schema from the currently active connection



620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
 
# File 'lib/switchman/shard.rb', line 620

def self.drop_database(name)
  sharding_config = Switchman.config || {}
  drop_statement = sharding_config["postgresql"]&.[](:drop_statement)
  drop_statement ||= sharding_config[:drop_statement]
  drop_statement = Array(drop_statement).map { |statement| statement.gsub("%{name}", name) } if drop_statement

  ::GuardRail.activate(:deploy) do
    # Shut up, Postgres!
    conn = ::ActiveRecord::Base.connection
    old_proc = conn.raw_connection.set_notice_processor {}
    begin
      drop_statement ||= "DROP SCHEMA #{name} CASCADE"
      Array(drop_statement).each do |stmt|
        ::ActiveRecord::Base.connection.execute(stmt)
      end
    ensure
      conn.raw_connection.set_notice_processor(&old_proc) if old_proc
    end
  end
end

.global_id_for(any_id, source_shard = nil) ⇒ Object

takes an id-ish, and returns an integral global id. returns nil if it can’t be interpreted



410
411
412
413
414
415
416
417
418
419
420
421
422
 
# File 'lib/switchman/shard.rb', line 410

def global_id_for(any_id, source_shard = nil)
  id = integral_id_for(any_id)
  return any_id unless id

  signed_id_operation(id) do |abs_id|
    if abs_id >= IDS_PER_SHARD
      abs_id
    else
      source_shard ||= Shard.current
      source_shard.global_id_for(abs_id)
    end
  end
end

.integral_id_for(any_id) ⇒ Object

converts an AR object, integral id, string id, or string short-global-id to a integral id. nil if it can’t be interpreted



337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
 
# File 'lib/switchman/shard.rb', line 337

def integral_id_for(any_id)
  case any_id
  when ::Arel::Nodes::Casted
    any_id = any_id.value
  when ::Arel::Nodes::BindParam
    any_id = any_id.value.value_before_type_cast
  end

  case any_id
  when ::ActiveRecord::Base
    any_id.id
  when /\A(\d+)~(-?\d+)\Z/
    local_id = $2.to_i
    signed_id_operation(local_id) do |id|
      return nil if id > IDS_PER_SHARD

      ($1.to_i * IDS_PER_SHARD) + id
    end
  when Integer, /\A-?\d+\Z/
    any_id.to_i
  end
end

.local_id_for(any_id) ⇒ Object 



365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
 
# File 'lib/switchman/shard.rb', line 365

def local_id_for(any_id)
  id = integral_id_for(any_id)
  return NIL_NIL_ID unless id

  return_shard = nil
  local_id = signed_id_operation(id) do |abs_id|
    if abs_id < IDS_PER_SHARD
      abs_id
    elsif (return_shard = lookup(abs_id / IDS_PER_SHARD))
      abs_id % IDS_PER_SHARD
    else
      return NIL_NIL_ID
    end
  end
  [local_id, return_shard]
end

.lookup(id) ⇒ Object

Raises:

  • (ArgumentError) 


134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
 
# File 'lib/switchman/shard.rb', line 134

def lookup(id)
  id_i = id.to_i
  return current if id_i == current.id || id == "self"
  return default if id_i == default.id || id.nil? || id == "default"

  id = id_i
  raise ArgumentError if id.zero?

  unless cached_shards.key?(id)
    cached_shards[id] = Shard.default.activate do
      find_cached(["shard", id]) { find_by(id: id) }
    end
  end
  cached_shards[id]
end

.partition_by_shard(array, partition_proc = nil) ⇒ Object 



289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
 
# File 'lib/switchman/shard.rb', line 289

def partition_by_shard(array, partition_proc = nil)
  shard_arrays = {}
  array.each do |object|
    partition_object = partition_proc ? partition_proc.call(object) : object
    case partition_object
    when Shard
      shard = partition_object
    when ::ActiveRecord::Base
      if partition_object.respond_to?(:associated_shards)
        partition_object.associated_shards.each do |a_shard|
          shard_arrays[a_shard] ||= []
          shard_arrays[a_shard] << object
        end
        next
      else
        shard = partition_object.shard
      end
    when Integer, /\A\d+\Z/, /\A(\d+)~(\d+)\Z/
      local_id, shard = Shard.local_id_for(partition_object)
      local_id ||= partition_object
      object = local_id unless partition_proc
    end
    shard ||= Shard.current
    shard_arrays[shard] ||= []
    shard_arrays[shard] << object
  end
  # TODO: use with_each_shard (or vice versa) to get
  # connection management and parallelism benefits
  shard_arrays.inject([]) do |results, (shard, objects)|
    results.concat(shard.activate { Array.wrap(yield objects) })
  end
end

.preload_cache ⇒ Object 



150
151
152
153
 
# File 'lib/switchman/shard.rb', line 150

def preload_cache
  cached_shards.reverse_merge!(active_shards.values.index_by(&:id))
  cached_shards.reverse_merge!(all.index_by(&:id))
end

.relative_id_for(any_id, source_shard, target_shard) ⇒ Object

takes an id-ish, and returns an integral id relative to target_shard. returns nil if it can’t be interpreted, or the integral version of the id if it refers to a shard that does not exist



386
387
388
389
390
391
392
393
394
395
 
# File 'lib/switchman/shard.rb', line 386

def relative_id_for(any_id, source_shard, target_shard)
  integral_id = integral_id_for(any_id)
  local_id, shard = local_id_for(integral_id)
  return integral_id unless local_id

  shard ||= source_shard
  return local_id if shard == target_shard

  shard.global_id_for(local_id)
end

.shard_for(any_id, source_shard = nil) ⇒ Object 



424
425
426
427
428
429
 
# File 'lib/switchman/shard.rb', line 424

def shard_for(any_id, source_shard = nil)
  return any_id.shard if any_id.is_a?(::ActiveRecord::Base)

  _, shard = local_id_for(any_id)
  shard || source_shard || Shard.current
end

.sharded_models ⇒ Object 



58
59
60
 
# File 'lib/switchman/shard.rb', line 58

def sharded_models
  @sharded_models ||= [::ActiveRecord::Base, UnshardedRecord].freeze
end

.short_id_for(any_id) ⇒ Object

takes an id-ish, and returns a shortened global string id if global, and itself if local. returns any_id itself if it can’t be interpreted



400
401
402
403
404
405
406
 
# File 'lib/switchman/shard.rb', line 400

def short_id_for(any_id)
  local_id, shard = local_id_for(any_id)
  return any_id unless local_id
  return local_id unless shard

  "#{shard.id}~#{local_id}"
end

.signed_id_operation(input_id) ⇒ Object

it’s tedious to hold onto this same kind of sign state and transform the result in multiple places, so here we can operate on the absolute value in a provided block and trust the sign will stay as provided. This assumes no consumer will return a nil value from the block.



329
330
331
332
333
 
# File 'lib/switchman/shard.rb', line 329

def signed_id_operation(input_id)
  sign = input_id.negative? ? -1 : 1
  output = yield input_id.abs
  output * sign
end

.with_each_shard(*args, parallel: false, exception: :raise, output: :simple) ⇒ Object

Parameters

  • shards - an array or relation of Shards to iterate over

  • classes - an array of classes to activate

    parallel: - true/false to execute in parallel, or an integer of how many
            sub-processes. Note that parallel invocation currently uses
            forking.
exception: - :ignore, :raise, :defer (wait until the end and raise the first
            error), or a proc
output: - :simple, :decorated (with database_server_id:shard_name)

Raises:

  • (ArgumentError) 


169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
 
# File 'lib/switchman/shard.rb', line 169

def with_each_shard(*args, parallel: false, exception: :raise, output: :simple)
  raise ArgumentError, "wrong number of arguments (#{args.length} for 0...2)" if args.length > 2

  return Array.wrap(yield) unless default.is_a?(Shard)

  if args.length == 1
    if Array === args.first && args.first.first.is_a?(Class)
      classes = args.first
    else
      scope = args.first
    end
  else
    scope, classes = args
  end

  parallel = [Environment.cpu_count || 2, 2].min if parallel == true
  parallel = 0 if parallel == false || parallel.nil?

  scope ||= Shard.all
  if ::ActiveRecord::Relation === scope && scope.order_values.empty?
    scope = scope.order(::Arel.sql("database_server_id IS NOT NULL, database_server_id, id"))
  end

  if parallel > 1
    if ::ActiveRecord::Relation === scope
      # still need a post-uniq, cause the default database server could be NULL or Rails.env in the db
      database_servers = scope.reorder("database_server_id").select(:database_server_id).distinct
                              .filter_map(&:database_server).uniq
      # nothing to do
      return if database_servers.count.zero?

      scopes = database_servers.to_h do |server|
        [server, scope.merge(server.shards)]
      end
    else
      scopes = scope.group_by(&:database_server)
    end

    # clear connections prior to forking (no more queries will be executed in the parent,
    # and we want them gone so that we don't accidentally use them post-fork doing something
    # silly like dealloc'ing prepared statements)
    if ::Rails.version < "7.1"
      ::ActiveRecord::Base.clear_all_connections!(nil)
    else
      ::ActiveRecord::Base.connection_handler.clear_all_connections!(:all)
    end

    parent_process_name = sanitized_process_title
    ret = ::Parallel.map(scopes, in_processes: (scopes.length > 1) ? parallel : 0) do |server, subscope|
      name = server.id
      last_description = name

      begin
        max_length = 128 - name.length - 3
        short_parent_name = parent_process_name[0..max_length] if max_length >= 0
        new_title = [short_parent_name, name].join(" ")
        Process.setproctitle(new_title)
        Switchman.config[:on_fork_proc]&.call
        with_each_shard(subscope, classes, exception: exception, output: :decorated) do
          last_description = Shard.current.description
          Parallel::ResultWrapper.new(yield)
        end
      rescue => e
        logger.error e.full_message
        Parallel::QuietExceptionWrapper.new(last_description, ::Parallel::ExceptionWrapper.new(e))
      end
    end.flatten

    errors = ret.select { |val| val.is_a?(Parallel::QuietExceptionWrapper) }
    unless errors.empty?
      raise errors.first.exception if errors.length == 1

      errors_desc = errors.map(&:name).sort.join(", ")
      raise Errors::ParallelShardExecError,
            "The following database server(s) did not finish processing cleanly: #{errors_desc}",
            cause: errors.first.exception
    end

    return ret.map(&:result)
  end

  classes ||= []

  result = []
  ex = nil
  old_stdout = $stdout
  old_stderr = $stderr
  scope.each do |shard|
    # shard references a database server that isn't configured in this environment
    next unless shard.database_server

    shard.activate(*classes) do
      if output == :decorated
        $stdout = Parallel::PrefixingIO.new(shard.description, $stdout)
        $stderr = Parallel::PrefixingIO.new(shard.description, $stderr)
      end

      result.concat Array.wrap(yield)
    rescue
      case exception
      when :ignore
        # ignore
      when :defer
        ex ||= $!
      when Proc
        exception.call
      # when :raise
      else
        raise
      end
    ensure
      $stdout = old_stdout
      $stderr = old_stderr
    end
  end
  raise ex if ex

  result
end

Instance Method Details

#_dump(_depth) ⇒ Object

custom serialization, since shard is self-referential



596
597
598
 
# File 'lib/switchman/shard.rb', line 596

def _dump(_depth)
  id.to_s
end

#activate(*classes, &block) ⇒ Object 



583
584
585
586
 
# File 'lib/switchman/shard.rb', line 583

def activate(*classes, &block)
  shards = hashify_classes(classes)
  Shard.activate(shards, &block)
end

#activate!(*classes) ⇒ Object

for use from console ONLY



589
590
591
592
593
 
# File 'lib/switchman/shard.rb', line 589

def activate!(*classes)
  shards = hashify_classes(classes)
  Shard.activate!(shards)
  nil
end

#database_server ⇒ Object 



553
554
555
 
# File 'lib/switchman/shard.rb', line 553

def database_server
  @database_server ||= DatabaseServer.find(database_server_id)
end

#database_server=(database_server) ⇒ Object 



557
558
559
560
 
# File 'lib/switchman/shard.rb', line 557

def database_server=(database_server)
  self.database_server_id = database_server.id
  @database_server = database_server
end

#description ⇒ Object 



566
567
568
 
# File 'lib/switchman/shard.rb', line 566

def description
  [database_server.id, name].compact.join(":")
end

#destroy ⇒ Object 



655
656
657
658
659
 
# File 'lib/switchman/shard.rb', line 655

def destroy
  raise("Cannot destroy the default shard") if default?

  super
end

#drop_database ⇒ Object 



604
605
606
607
608
609
610
611
612
613
614
615
 
# File 'lib/switchman/shard.rb', line 604

def drop_database
  raise "Cannot drop the database of the default shard" if default?
  return unless read_attribute(:name)

  begin
    activate do
      self.class.drop_database(name)
    end
  rescue ::ActiveRecord::StatementInvalid => e
    logger.error "Drop failed: #{e}"
  end
end

#global_id_for(local_id) ⇒ Object

takes an id local to this shard, and returns a global id



642
643
644
645
646
647
648
 
# File 'lib/switchman/shard.rb', line 642

def global_id_for(local_id)
  return nil unless local_id

  self.class.signed_id_operation(local_id) do |abs_id|
    abs_id + (id * IDS_PER_SHARD)
  end
end

#hash ⇒ Object

skip global_id.hash



651
652
653
 
# File 'lib/switchman/shard.rb', line 651

def hash
  id.hash
end

#name ⇒ Object 



539
540
541
542
543
544
545
546
 
# File 'lib/switchman/shard.rb', line 539

def name
  unless instance_variable_defined?(:@name)
    # protect against re-entrancy
    @name = nil
    @name = read_attribute(:name) || default_name
  end
  @name
end

#name=(name) ⇒ Object 



548
549
550
551
 
# File 'lib/switchman/shard.rb', line 548

def name=(name)
  write_attribute(:name, @name = name)
  remove_instance_variable(:@name) if name.nil?
end

#original_id ⇒ Object 



575
576
577
 
# File 'lib/switchman/shard.rb', line 575

def original_id
  id
end

#original_id_value ⇒ Object 



579
580
581
 
# File 'lib/switchman/shard.rb', line 579

def original_id_value
  id
end

#primary? ⇒ Boolean

Returns:

  • (Boolean) 


562
563
564
 
# File 'lib/switchman/shard.rb', line 562

def primary?
  self == database_server.primary_shard
end

#shard ⇒ Object

Shards are always on the default shard



571
572
573
 
# File 'lib/switchman/shard.rb', line 571

def shard
  Shard.default
end