Class: Switchman::Shard
- Inherits:
-
ActiveRecord::Base
- Object
- ActiveRecord::Base
- Switchman::Shard
- Defined in:
- app/models/switchman/shard_internal.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
- ._load(str) ⇒ Object
- .activate(shards) ⇒ Object
- .activate!(shards) ⇒ Object
- .categories ⇒ Object
- .clear_cache ⇒ Object
- .current(category = :default) ⇒ Object
- .default(reload_deprecated = false, reload: false, with_fallback: false) ⇒ Object
-
.determine_max_procs(max_procs_input, parallel_input = 2) ⇒ Object
given the provided option, determines whether we need to (and whether it’s possible) to determine a reasonable default.
-
.global_id_for(any_id, source_shard = nil) ⇒ Object
takes an id-ish, and returns an integral global id.
-
.integral_id_for(any_id) ⇒ Object
converts an AR object, integral id, string id, or string short-global-id to a integral id.
- .local_id_for(any_id) ⇒ Object
- .lookup(id) ⇒ Object
- .partition_by_shard(array, partition_proc = nil) ⇒ Object
-
.relative_id_for(any_id, source_shard, target_shard) ⇒ Object
takes an id-ish, and returns an integral id relative to target_shard.
- .shard_for(any_id, source_shard = nil) ⇒ Object
-
.short_id_for(any_id) ⇒ Object
takes an id-ish, and returns a shortened global string id if global, and itself if local.
-
.with_each_shard(*args) ⇒ Object
Parameters.
Instance Method Summary collapse
-
#_dump(depth) ⇒ Object
custom serialization, since shard is self-referential.
- #activate(*categories) ⇒ Object
-
#activate!(*categories) ⇒ Object
for use from console ONLY.
- #database_server ⇒ Object
- #database_server=(database_server) ⇒ Object
- #description ⇒ Object
- #destroy ⇒ Object
- #drop_database ⇒ Object
-
#global_id_for(local_id) ⇒ Object
takes an id local to this shard, and returns a global id.
-
#hash ⇒ Object
skip global_id.hash.
- #name ⇒ Object
- #name=(name) ⇒ Object
- #primary? ⇒ Boolean
-
#shard ⇒ Object
Shards are always on the default shard.
Class Method Details
._load(str) ⇒ Object
603 604 605 |
# File 'app/models/switchman/shard_internal.rb', line 603 def self._load(str) lookup(str.to_i) end |
.activate(shards) ⇒ Object
78 79 80 81 82 83 |
# File 'app/models/switchman/shard_internal.rb', line 78 def activate(shards) old_shards = activate!(shards) yield ensure active_shards.merge!(old_shards) if old_shards end |
.activate!(shards) ⇒ Object
85 86 87 88 89 90 91 92 93 94 95 96 97 |
# File 'app/models/switchman/shard_internal.rb', line 85 def activate!(shards) old_shards = nil currently_active_shards = active_shards shards.each do |category, shard| next if category == :unsharded unless currently_active_shards[category] == shard old_shards ||= {} old_shards[category] = currently_active_shards[category] currently_active_shards[category] = shard end end old_shards end |
.categories ⇒ Object
33 34 35 |
# File 'app/models/switchman/shard_internal.rb', line 33 def categories CATEGORIES.keys end |
.clear_cache ⇒ Object
142 143 144 |
# File 'app/models/switchman/shard_internal.rb', line 142 def clear_cache cached_shards.clear end |
.current(category = :default) ⇒ Object
74 75 76 |
# File 'app/models/switchman/shard_internal.rb', line 74 def current(category = :default) active_shards[category] || Shard.default end |
.default(reload_deprecated = false, reload: false, with_fallback: false) ⇒ Object
37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 |
# File 'app/models/switchman/shard_internal.rb', line 37 def default(reload_deprecated = false, reload: false, with_fallback: false) reload = reload_deprecated if reload_deprecated 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 if with_fallback && @default default = @default end # 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; rescue the fake default if the table doesn't exist @default = begin Shard.where(default: true).first || default rescue default end # rebuild current shard activations - it might have "another" default shard serialized there active_shards.replace(active_shards.map do |category, shard| shard = Shard.lookup((!shard || shard.default?) ? 'default' : shard.id) [category, shard] end.to_h) activate!(default: @default) if active_shards.empty? end @default end |
.determine_max_procs(max_procs_input, parallel_input = 2) ⇒ Object
given the provided option, determines whether we need to (and whether it’s possible) to determine a reasonable default.
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 |
# File 'app/models/switchman/shard_internal.rb', line 513 def determine_max_procs(max_procs_input, parallel_input=2) max_procs = nil if max_procs_input max_procs = max_procs_input.to_i max_procs = nil if max_procs == 0 else return 1 if parallel_input.nil? || parallel_input < 1 cpus = Environment.cpu_count if cpus && cpus > 0 max_procs = cpus * parallel_input end end return max_procs 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
494 495 496 497 498 499 500 501 502 503 |
# File 'app/models/switchman/shard_internal.rb', line 494 def global_id_for(any_id, source_shard = nil) id = integral_id_for(any_id) return any_id unless id if id >= IDS_PER_SHARD id else source_shard ||= Shard.current source_shard.global_id_for(id) 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
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 |
# File 'app/models/switchman/shard_internal.rb', line 432 def integral_id_for(any_id) if ::Rails.version >= '4.2' && any_id.is_a?(::Arel::Nodes::Casted) any_id = any_id.val end case any_id when ::ActiveRecord::Base any_id.id when /^(\d+)~(\d+)$/ local_id = $2.to_i # doesn't make sense to have a double-global id return nil if local_id > IDS_PER_SHARD $1.to_i * IDS_PER_SHARD + local_id when Integer, /^\d+$/ any_id.to_i else nil end end |
.local_id_for(any_id) ⇒ Object
457 458 459 460 461 462 463 464 465 466 467 |
# File 'app/models/switchman/shard_internal.rb', line 457 def local_id_for(any_id) id = integral_id_for(any_id) return NIL_NIL_ID unless id if id < IDS_PER_SHARD [id, nil] elsif shard = lookup(id / IDS_PER_SHARD) [id % IDS_PER_SHARD, shard] else NIL_NIL_ID end end |
.lookup(id) ⇒ Object
99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 |
# File 'app/models/switchman/shard_internal.rb', line 99 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 == 0 unless cached_shards.has_key?(id) cached_shards[id] = Shard.default.activate do # can't simply cache the AR object since Shard has a custom serializer # that calls this method attributes = Switchman.cache.fetch(['shard', id].join('/')) do shard = find_by_id(id) if shard attributes = shard.attributes if ::Rails.version < '4.2' attributes.each_key do |key| attributes[key] = attributes[key].unserialize if attributes[key].is_a?(::ActiveRecord::AttributeMethods::Serialization::Attribute) end end attributes else :nil end end if attributes == :nil nil else shard = Shard.new attributes.each do |attr, value| shard.send(:"#{attr}=", value) end shard.instance_variable_set(:@new_record, false) # connection info doesn't exist in database.yml; # pretend the shard doesn't exist either shard = nil unless shard.database_server shard end end end cached_shards[id] end |
.partition_by_shard(array, partition_proc = nil) ⇒ Object
397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 |
# File 'app/models/switchman/shard_internal.rb', line 397 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, /^\d+$/, /^(\d+)~(\d+)$/ local_id, shard = Shard.local_id_for(partition_object) local_id ||= partition_object object = local_id if !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 |
.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
473 474 475 476 477 478 479 480 |
# File 'app/models/switchman/shard_internal.rb', line 473 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
505 506 507 508 509 |
# File 'app/models/switchman/shard_internal.rb', line 505 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 |
.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
485 486 487 488 489 490 |
# File 'app/models/switchman/shard_internal.rb', line 485 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 |
.with_each_shard(*args) ⇒ Object
Parameters
-
shards
- an array or relation of Shards to iterate over -
categories
- an array of categories to activate -
options
-:parallel - true/false to execute in parallel, or a integer of how many sub-processes per database server. Note that parallel invocation currently uses forking, so should be used sparingly because errors are not raised, and you cannot get results back :max_procs - only run this many parallel processes at a time :exception - :ignore, :raise, :defer (wait until the end and raise the first error), or a proc
158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 |
# File 'app/models/switchman/shard_internal.rb', line 158 def with_each_shard(*args) raise ArgumentError, "wrong number of arguments (#{args.length} for 0...3)" if args.length > 3 unless default.is_a?(Shard) return Array.wrap(yield) end = args. if args.length == 1 if Array === args.first && args.first.first.is_a?(Symbol) categories = args.first else scope = args.first end else scope, categories = args end parallel = case [:parallel] when true 1 when false, nil 0 else [:parallel] end .delete(:parallel) scope ||= Shard.all if ::ActiveRecord::Relation === scope && scope.order_values.empty? scope = scope.order("database_server_id IS NOT NULL, database_server_id, id") end if parallel > 0 max_procs = determine_max_procs(.delete(:max_procs), parallel) 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).uniq. map(&:database_server).compact.uniq parallel = [(max_procs.to_f / database_servers.count).ceil, parallel].min if max_procs scopes = Hash[database_servers.map do |server| server_scope = server.shards.merge(scope) if parallel == 1 subscopes = [server_scope] else subscopes = [] total = server_scope.count ranges = [] server_scope.find_ids_in_ranges(:batch_size => (total.to_f / parallel).ceil) do |min, max| ranges << [min, max] end # create a half-open range on the last one ranges.last[1] = nil ranges.each do |min, max| subscope = server_scope.where("id>=?", min) subscope = subscope.where("id<=?", max) if max subscopes << subscope end end [server, subscopes] end] else scopes = scope.group_by(&:database_server) if parallel > 1 parallel = [(max_procs.to_f / scopes.count).ceil, parallel].min if max_procs scopes = Hash[scopes.map do |(server, shards)| [server, shards.in_groups(parallel, false).compact] end] end end exception_pipes = [] pids = [] out_fds = [] err_fds = [] pid_to_name_map = {} fd_to_name_map = {} errors = [] wait_for_output = lambda do |out_fds, err_fds, fd_to_name_map| ready, _ = IO.select(out_fds + err_fds) ready.each do |fd| if fd.eof? fd.close out_fds.delete(fd) err_fds.delete(fd) next end line = fd.readline puts "#{fd_to_name_map[fd]}: #{line}" end end # only one process; don't bother forking if scopes.length == 1 && parallel == 1 return with_each_shard(subscopes.first, categories, ) { yield } 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) ::ActiveRecord::Base.clear_all_connections! scopes.each do |server, subscopes| if !(::ActiveRecord::Relation === subscopes.first) && subscopes.first.class != Array subscopes = [subscopes] end subscopes.each_with_index do |subscope, idx| if subscopes.length > 1 name = "#{server.id} #{idx + 1}" else name = server.id end exception_pipe = IO.pipe exception_pipes << exception_pipe pid, io_in, io_out, io_err = Open4.pfork4(lambda do begin Switchman.config[:on_fork_proc].try(:call) $0 = [$0, ARGV, name].flatten.join(' ') with_each_shard(subscope, categories, ) { yield } exception_pipe.last.close rescue => e begin dumped = Marshal.dump(e) rescue # couldn't dump the exception; create a copy with just # the message and the backtrace e2 = e.class.new(e.) e2.set_backtrace(e.backtrace) e2.instance_variable_set(:@active_shards, e.instance_variable_get(:@active_shards)) dumped = Marshal.dump(e2) end exception_pipe.last.set_encoding(dumped.encoding) exception_pipe.last.write(dumped) exception_pipe.last.flush exception_pipe.last.close exit! 1 end end) exception_pipe.last.close pids << pid io_in.close # don't care about writing to stdin out_fds << io_out err_fds << io_err pid_to_name_map[pid] = name fd_to_name_map[io_out] = name fd_to_name_map[io_err] = name while max_procs && pids.count >= max_procs while max_procs && out_fds.count >= max_procs # wait for output if we've hit the max_procs limit wait_for_output.call(out_fds, err_fds, fd_to_name_map) end # we've gotten all the output from one fd so wait for its child process to exit found_pid, status = Process.wait2 pids.delete(found_pid) errors << pid_to_name_map[found_pid] if status.exitstatus != 0 end end end while out_fds.any? || err_fds.any? wait_for_output.call(out_fds, err_fds, fd_to_name_map) end pids.each do |pid| _, status = Process.waitpid2(pid) errors << pid_to_name_map[pid] if status.exitstatus != 0 end # check for an exception; we only re-raise the first one exception_pipes.each do |exception_pipe| begin serialized_exception = exception_pipe.first.read next if serialized_exception.empty? exception = Marshal.load(serialized_exception) raise exception ensure exception_pipe.first.close end end unless errors.empty? raise ParallelShardExecError.new("The following subprocesses did not exit cleanly: #{errors.sort.join(", ")}") end return end categories ||= [] previous_shard = nil close_connections_if_needed = lambda do |shard| # prune the prior connection unless it happened to be the same if previous_shard && shard != previous_shard && (shard.database_server != previous_shard.database_server || !previous_shard.database_server.shareable?) previous_shard.activate do ::Shackles.activated_environments.each do |env| ::Shackles.activate(env) do if ::ActiveRecord::Base.connected? && ::ActiveRecord::Base.connection.open_transactions == 0 ::ActiveRecord::Base.connection_pool.current_pool.disconnect! end end end end end end result = [] exception = nil scope.each do |shard| # shard references a database server that isn't configured in this environment next unless shard.database_server close_connections_if_needed.call(shard) shard.activate(*categories) do begin result.concat Array.wrap(yield) rescue case [:exception] when :ignore when :defer exception ||= $! when Proc [:exception].call when :raise raise else raise end end end previous_shard = shard end close_connections_if_needed.call(Shard.current) raise exception if exception result end |
Instance Method Details
#_dump(depth) ⇒ Object
custom serialization, since shard is self-referential
599 600 601 |
# File 'app/models/switchman/shard_internal.rb', line 599 def _dump(depth) self.id.to_s end |
#activate(*categories) ⇒ Object
584 585 586 587 588 589 |
# File 'app/models/switchman/shard_internal.rb', line 584 def activate(*categories) shards = hashify_categories(categories) Shard.activate(shards) do yield end end |
#activate!(*categories) ⇒ Object
for use from console ONLY
592 593 594 595 596 |
# File 'app/models/switchman/shard_internal.rb', line 592 def activate!(*categories) shards = hashify_categories(categories) Shard.activate!(shards) nil end |
#database_server ⇒ Object
562 563 564 |
# File 'app/models/switchman/shard_internal.rb', line 562 def database_server @database_server ||= DatabaseServer.find(self.database_server_id) end |
#database_server=(database_server) ⇒ Object
566 567 568 569 |
# File 'app/models/switchman/shard_internal.rb', line 566 def database_server=(database_server) self.database_server_id = database_server.id @database_server = database_server end |
#description ⇒ Object
575 576 577 |
# File 'app/models/switchman/shard_internal.rb', line 575 def description [database_server.id, name].compact.join(':') end |
#destroy ⇒ Object
666 667 668 669 |
# File 'app/models/switchman/shard_internal.rb', line 666 def destroy raise("Cannot destroy the default shard") if self.default? super end |
#drop_database ⇒ Object
607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 |
# File 'app/models/switchman/shard_internal.rb', line 607 def drop_database raise("Cannot drop the database of the default shard") if self.default? return unless read_attribute(:name) begin adapter = self.database_server.config[:adapter] sharding_config = Switchman.config || {} drop_statement = sharding_config[adapter].try(:[], :drop_statement) drop_statement ||= sharding_config[:drop_statement] if drop_statement drop_statement = Array(drop_statement).dup. map { |statement| statement.gsub('%{name}', self.name) } end case adapter when 'mysql', 'mysql2' self.activate do ::Shackles.activate(:deploy) do drop_statement ||= "DROP DATABASE #{self.name}" Array(drop_statement).each do |stmt| ::ActiveRecord::Base.connection.execute(stmt) end end end when 'postgresql' self.activate do ::Shackles.activate(:deploy) do # Shut up, Postgres! conn = ::ActiveRecord::Base.connection old_proc = conn.raw_connection.set_notice_processor {} begin drop_statement ||= "DROP SCHEMA #{self.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 when 'sqlite3' File.delete(self.name) unless self.name == ':memory:' end rescue logger.info "Drop failed: #{$!}" end end |
#global_id_for(local_id) ⇒ Object
takes an id local to this shard, and returns a global id
656 657 658 659 |
# File 'app/models/switchman/shard_internal.rb', line 656 def global_id_for(local_id) return nil unless local_id local_id + self.id * IDS_PER_SHARD end |
#hash ⇒ Object
skip global_id.hash
662 663 664 |
# File 'app/models/switchman/shard_internal.rb', line 662 def hash id.hash end |
#name ⇒ Object
548 549 550 551 552 553 554 555 |
# File 'app/models/switchman/shard_internal.rb', line 548 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
557 558 559 560 |
# File 'app/models/switchman/shard_internal.rb', line 557 def name=(name) write_attribute(:name, @name = name) remove_instance_variable(:@name) if name == nil end |
#primary? ⇒ Boolean
571 572 573 |
# File 'app/models/switchman/shard_internal.rb', line 571 def primary? self == database_server.primary_shard end |