Class: PEROBS::FlatFile
- Inherits:
-
Object
- Object
- PEROBS::FlatFile
- Defined in:
- lib/perobs/FlatFile.rb
Overview
The FlatFile class manages the storage file of the FlatFileDB. It contains a sequence of blobs Each blob consists of header and the actual blob data bytes.
Constant Summary collapse
- INDEX_BTREE_ORDER =
The number of entries in a single BTree node of the index file.
65
Class Method Summary collapse
Instance Method Summary collapse
-
#check(repair = false) ⇒ Integer
Check (and repair) the FlatFile.
-
#clear_all_marks ⇒ Object
Clear alls marks.
-
#close ⇒ Object
Close the flat file.
-
#defragmentize ⇒ Object
Eliminate all the holes in the file.
-
#delete_obj_by_address(addr, id) ⇒ Object
Delete the blob that is stored at the specified address.
-
#delete_obj_by_id(id) ⇒ Boolean
Delete the blob for the specified ID.
-
#delete_unmarked_objects ⇒ Object
Delete all unmarked objects.
-
#find_obj_addr_by_id(id) ⇒ Integer
Find the address of the object with the given ID.
- #has_id_at?(id, address) ⇒ Boolean
- #has_space?(address, size) ⇒ Boolean
-
#initialize(dir, progressmeter) ⇒ FlatFile
constructor
Create a new FlatFile object for a database in the given path.
- #inspect ⇒ Object
-
#is_marked_by_id?(id) ⇒ Boolean
Return true if the object with the given ID is marked, false otherwise.
-
#item_counter ⇒ Integer
Number of items stored in the DB.
-
#mark_obj_by_id(id) ⇒ Object
Mark the object with the given ID.
-
#open ⇒ Object
Open the flat file for reading and writing.
-
#read_obj_by_address(addr, id) ⇒ String
Read the object at the specified address.
-
#read_obj_by_id(id) ⇒ String or nil
Read the object with the given ID.
-
#refresh ⇒ Object
This method iterates over all entries in the FlatFile and removes the entry and inserts it again.
-
#regenerate_index_and_spaces ⇒ Object
This method clears the index tree and the free space list and regenerates them from the FlatFile.
-
#sync ⇒ Object
Force outstanding data to be written to the filesystem.
-
#write_obj_by_id(id, raw_obj) ⇒ Integer
Write the given object into the file.
Constructor Details
#initialize(dir, progressmeter) ⇒ FlatFile
Create a new FlatFile object for a database in the given path.
48 49 50 51 52 53 54 55 |
# File 'lib/perobs/FlatFile.rb', line 48 def initialize(dir, progressmeter) @db_dir = dir @progressmeter = progressmeter @f = nil @marks = nil @index = BTree.new(@db_dir, 'index', INDEX_BTREE_ORDER, @progressmeter) @space_list = SpaceTree.new(@db_dir, @progressmeter) end |
Class Method Details
.insert_header_checksums(db_dir) ⇒ Object
638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 |
# File 'lib/perobs/FlatFile.rb', line 638 def FlatFile::insert_header_checksums(db_dir) old_file_name = File.join(db_dir, 'database.blobs') new_file_name = File.join(db_dir, 'database_v4.blobs') bak_file_name = File.join(db_dir, 'database_v3.blobs') old_file = File.open(old_file_name, 'rb') new_file = File.open(new_file_name, 'wb') entries = 0 while (buf = old_file.read(21)) flags, length, id, crc = *buf.unpack('CQQL') blob_data = old_file.read(length) # Some basic sanity checking to ensure all reserved bits are 0. Older # versions of PEROBS used to set bit 1 despite it being reserved now. unless flags & 0xF0 == 0 PEROBS.log.fatal "Blob file #{old_file_name} contains illegal " + "flag byte #{'%02x' % flags} at #{old_file.pos - 21}" end # Check if the blob is valid and current. if flags & 0x1 == 1 && flags & 0x8 == 0 # Make sure the bit 1 is not set anymore. flags = flags & 0x05 header_str = [ flags, length, id, crc ].pack('CQQL') header_crc = Zlib.crc32(header_str, 0) header_str += [ header_crc ].pack('L') new_file.write(header_str + blob_data) entries += 1 end end PEROBS.log.info "Header checksum added to #{entries} entries" old_file.close new_file.close File.rename(old_file_name, bak_file_name) File.rename(new_file_name, old_file_name) end |
Instance Method Details
#check(repair = false) ⇒ Integer
Check (and repair) the FlatFile.
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 |
# File 'lib/perobs/FlatFile.rb', line 447 def check(repair = false) errors = 0 return errors unless @f t = Time.now PEROBS.log.info "Checking FlatFile database" + "#{repair ? ' in repair mode' : ''}..." # First check the database blob file. Each entry should be readable and # correct and all IDs must be unique. We use a shadow index to keep # track of the already found IDs. new_index = BTree.new(@db_dir, 'new-index', INDEX_BTREE_ORDER, @progressmeter) new_index.erase new_index.open corrupted_blobs = 0 @progressmeter.start('Checking blobs file', @f.size) do |pm| corrupted_blobs = each_blob_header do |header| if header.is_valid? # We have a non-deleted entry. begin @f.seek(header.addr + FlatFileBlobHeader::LENGTH) buf = @f.read(header.length) if buf.bytesize != header.length PEROBS.log.error "Premature end of file in blob with ID " + "#{header.id}." discard_damaged_blob(header) if repair errors += 1 next end # Uncompress the data if the compression bit is set in the mark # byte. if header.is_compressed? begin buf = Zlib.inflate(buf) rescue Zlib::BufError, Zlib::DataError PEROBS.log.error "Corrupted compressed block with ID " + "#{header.id} found." discard_damaged_blob(header) if repair errors += 1 next end end if header.crc && checksum(buf) != header.crc PEROBS.log.error "Checksum failure while checking blob " + "with ID #{header.id}" discard_damaged_blob(header) if repair errors += 1 next end rescue IOError => e PEROBS.log.fatal "Check of blob with ID #{header.id} failed: " + e. end # Check if the ID has already been found in the file. if (previous_address = new_index.get(header.id)) PEROBS.log.error "Multiple blobs for ID #{header.id} found. " + "Addresses: #{previous_address}, #{header.addr}" errors += 1 previous_header = FlatFileBlobHeader.read(@f, previous_address, header.id) if repair # We have two blobs with the same ID and we must discard one of # them. if header.is_outdated? discard_damaged_blob(header) elsif previous_header.is_outdated? discard_damaged_blob(previous_header) else PEROBS.log.error "None of the blobs with same ID have " + "the outdated flag set. Deleting the smaller one." errors += 1 discard_damaged_blob(header.length < previous_header.length ? header : previous_header) end next end else # ID is unique so far. Add it to the shadow index. new_index.insert(header.id, header.addr) end end pm.update(header.addr) end errors += corrupted_blobs end # We no longer need the new index. new_index.close new_index.erase if repair && corrupted_blobs > 0 erase_index_files defragmentize regenerate_index_and_spaces else # Now we check the index data. It must be correct and the entries must # match the blob file. All entries in the index must be in the blob file # and vise versa. begin index_ok = @index.check do |id, address| has_id_at?(id, address) end x_check_errs = 0 space_check_ok = true unless index_ok && (space_check_ok = @space_list.check(self)) && (x_check_errs = cross_check_entries) == 0 errors += 1 unless index_ok && space_check_ok errors += x_check_errs regenerate_index_and_spaces if repair end rescue PEROBS::FatalError errors += 1 regenerate_index_and_spaces if repair end end sync if repair PEROBS.log.info "check_db completed in #{Time.now - t} seconds. " + "#{errors} errors found." errors end |
#clear_all_marks ⇒ Object
Clear alls marks.
327 328 329 330 331 332 333 |
# File 'lib/perobs/FlatFile.rb', line 327 def clear_all_marks if @marks @marks.clear else @marks = IDList.new(@db_dir, 'marks', 8) end end |
#close ⇒ Object
Close the flat file. This method must be called to ensure that all data is really written into the filesystem.
84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 |
# File 'lib/perobs/FlatFile.rb', line 84 def close @space_list.close if @space_list.is_open? @index.close if @index.is_open? if @marks @marks.erase @marks = nil end if @f @f.flush @f.flock(File::LOCK_UN) @f.fsync @f.close @f = nil end end |
#defragmentize ⇒ Object
Eliminate all the holes in the file. This is an in-place implementation. No additional space will be needed on the file system.
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 |
# File 'lib/perobs/FlatFile.rb', line 337 def defragmentize distance = 0 new_file_size = 0 deleted_blobs = 0 corrupted_blobs = 0 valid_blobs = 0 # Iterate over all entries. @progressmeter.start('Defragmentizing blobs file', @f.size) do |pm| each_blob_header do |header| # If we have stumbled over a corrupted blob we treat it similar to a # deleted blob and reuse the space. if header.corruption_start distance += header.addr - header.corruption_start corrupted_blobs += 1 end # Total size of the current entry entry_bytes = FlatFileBlobHeader::LENGTH + header.length if header.is_valid? # We have found a valid entry. valid_blobs += 1 if distance > 0 begin # Read current entry into a buffer @f.seek(header.addr) buf = @f.read(entry_bytes) # Write the buffer right after the end of the previous entry. @f.seek(header.addr - distance) @f.write(buf) # Mark the space between the relocated current entry and the # next valid entry as deleted space. FlatFileBlobHeader.new(@f, @f.pos, 0, distance - FlatFileBlobHeader::LENGTH, 0, 0).write @f.flush rescue IOError => e PEROBS.log.fatal "Error while moving blob for ID " + "#{header.id}: #{e.}" end end new_file_size = header.addr - distance + FlatFileBlobHeader::LENGTH + header.length else deleted_blobs += 1 distance += entry_bytes end pm.update(header.addr) end end PEROBS.log.info "#{distance / 1000} KiB/#{deleted_blobs} blobs of " + "#{@f.size / 1000} KiB/#{valid_blobs} blobs or " + "#{'%.1f' % (distance.to_f / @f.size * 100.0)}% reclaimed" if corrupted_blobs > 0 PEROBS.log.info "#{corrupted_blobs} corrupted blob(s) found. Space " + "was recycled." end @f.flush @f.truncate(new_file_size) @f.flush sync end |
#delete_obj_by_address(addr, id) ⇒ Object
Delete the blob that is stored at the specified address.
129 130 131 132 133 134 |
# File 'lib/perobs/FlatFile.rb', line 129 def delete_obj_by_address(addr, id) @index.remove(id) if @index.is_open? header = FlatFileBlobHeader.read(@f, addr, id) header.clear_flags @space_list.add_space(addr, header.length) if @space_list.is_open? end |
#delete_obj_by_id(id) ⇒ Boolean
Delete the blob for the specified ID.
117 118 119 120 121 122 123 124 |
# File 'lib/perobs/FlatFile.rb', line 117 def delete_obj_by_id(id) if (pos = find_obj_addr_by_id(id)) delete_obj_by_address(pos, id) return true end return false end |
#delete_unmarked_objects ⇒ Object
Delete all unmarked objects.
137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 |
# File 'lib/perobs/FlatFile.rb', line 137 def delete_unmarked_objects # We don't update the index and the space list during this operation as # we defragmentize the blob file at the end. We'll end the operation # with an empty space list. clear_index_files deleted_objects_count = 0 @progressmeter.start('Sweeping unmarked objects', @f.size) do |pm| each_blob_header do |header| if header.is_valid? && !@marks.include?(header.id) delete_obj_by_address(header.addr, header.id) deleted_objects_count += 1 end pm.update(header.addr) end end defragmentize # Update the index file and create a new, empty space list. regenerate_index_and_spaces deleted_objects_count end |
#find_obj_addr_by_id(id) ⇒ Integer
Find the address of the object with the given ID.
257 258 259 |
# File 'lib/perobs/FlatFile.rb', line 257 def find_obj_addr_by_id(id) @index.get(id) end |
#has_id_at?(id, address) ⇒ Boolean
619 620 621 622 |
# File 'lib/perobs/FlatFile.rb', line 619 def has_id_at?(id, address) header = FlatFileBlobHeader.read(@f, address) header.is_valid? && header.id == id end |
#has_space?(address, size) ⇒ Boolean
614 615 616 617 |
# File 'lib/perobs/FlatFile.rb', line 614 def has_space?(address, size) header = FlatFileBlobHeader.read(@f, address) !header.is_valid? && header.length == size end |
#inspect ⇒ Object
624 625 626 627 628 629 630 631 632 633 634 635 636 |
# File 'lib/perobs/FlatFile.rb', line 624 def inspect s = '[' each_blob_header do |header| s << "{ :pos => #{header.addr}, :flags => #{header.flags}, " + ":length => #{header.length}, :id => #{header.id}, " + ":crc => #{header.crc}" if header.is_valid? s << ", :value => #{@f.read(header.length)}" end s << " }\n" end s + ']' end |
#is_marked_by_id?(id) ⇒ Boolean
Return true if the object with the given ID is marked, false otherwise.
322 323 324 |
# File 'lib/perobs/FlatFile.rb', line 322 def is_marked_by_id?(id) @marks.include?(id) end |
#item_counter ⇒ Integer
Returns Number of items stored in the DB.
273 274 275 |
# File 'lib/perobs/FlatFile.rb', line 273 def item_counter @index.entries_count end |
#mark_obj_by_id(id) ⇒ Object
Mark the object with the given ID.
316 317 318 |
# File 'lib/perobs/FlatFile.rb', line 316 def mark_obj_by_id(id) @marks.insert(id) end |
#open ⇒ Object
Open the flat file for reading and writing.
58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 |
# File 'lib/perobs/FlatFile.rb', line 58 def open file_name = File.join(@db_dir, 'database.blobs') new_db_created = false begin if File.exist?(file_name) @f = File.open(file_name, 'rb+') else PEROBS.log.info "New FlatFile database '#{file_name}' created" @f = File.open(file_name, 'wb+') new_db_created = true end rescue IOError => e PEROBS.log.fatal "Cannot open FlatFile database #{file_name}: " + e. end unless @f.flock(File::LOCK_NB | File::LOCK_EX) PEROBS.log.fatal "FlatFile database '#{file_name}' is locked by " + "another process" end @f.sync = true open_index_files(!new_db_created) end |
#read_obj_by_address(addr, id) ⇒ String
Read the object at the specified address.
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 |
# File 'lib/perobs/FlatFile.rb', line 281 def read_obj_by_address(addr, id) header = FlatFileBlobHeader.read(@f, addr, id) if header.id != id PEROBS.log.fatal "Database index corrupted: Index for object " + "#{id} points to object with ID #{header.id}" end buf = nil begin @f.seek(addr + FlatFileBlobHeader::LENGTH) buf = @f.read(header.length) rescue IOError => e PEROBS.log.fatal "Cannot read blob for ID #{id}: #{e.}" end # Uncompress the data if the compression bit is set in the flags byte. if header.is_compressed? begin buf = Zlib.inflate(buf) rescue Zlib::BufError, Zlib::DataError PEROBS.log.fatal "Corrupted compressed block with ID " + "#{header.id} found." end end if checksum(buf) != header.crc PEROBS.log.fatal "Checksum failure while reading blob ID #{id}" end buf end |
#read_obj_by_id(id) ⇒ String or nil
Read the object with the given ID.
264 265 266 267 268 269 270 |
# File 'lib/perobs/FlatFile.rb', line 264 def read_obj_by_id(id) if (addr = find_obj_addr_by_id(id)) return read_obj_by_address(addr, id) end nil end |
#refresh ⇒ Object
This method iterates over all entries in the FlatFile and removes the entry and inserts it again. This is useful to update all entries in case the storage format has changed.
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 |
# File 'lib/perobs/FlatFile.rb', line 407 def refresh # This iteration might look scary as we iterate over the entries while # while we are rearranging them. Re-inserted items may be inserted # before or at the current entry and this is fine. They also may be # inserted after the current entry and will be re-read again unless they # are inserted after the original file end. file_size = @f.size # We don't update the index and the space list during this operation as # we defragmentize the blob file at the end. We'll end the operation # with an empty space list. clear_index_files @progressmeter.start('Converting objects to new storage format', @f.size) do |pm| each_blob_header do |header| if header.is_valid? buf = read_obj_by_address(header.addr, header.id) delete_obj_by_address(header.addr, header.id) write_obj_by_id(header.id, buf) end # Some re-inserted blobs may be inserted after the original file end. # No need to process those blobs again. break if header.addr >= file_size pm.update(header.addr) end end # Reclaim the space saved by compressing entries. defragmentize # Recreate the index file and create an empty space list. regenerate_index_and_spaces end |
#regenerate_index_and_spaces ⇒ Object
This method clears the index tree and the free space list and regenerates them from the FlatFile.
580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 |
# File 'lib/perobs/FlatFile.rb', line 580 def regenerate_index_and_spaces PEROBS.log.warn "Re-generating FlatFileDB index and space files" @index.open unless @index.is_open? @index.clear @space_list.open unless @space_list.is_open? @space_list.clear @progressmeter.start('Re-generating database index', @f.size) do |pm| each_blob_header do |header| if header.is_valid? if (duplicate_pos = @index.get(header.id)) PEROBS.log.error "FlatFile contains multiple blobs for ID " + "#{header.id}. First blob is at address #{duplicate_pos}. " + "Other blob found at address #{header.addr}." if header.length > 0 @space_list.add_space(header.addr, header.length) end discard_damaged_blob(header) else @index.insert(header.id, header.addr) end else if header.length > 0 @space_list.add_space(header.addr, header.length) end end pm.update(header.addr) end end sync end |
#sync ⇒ Object
Force outstanding data to be written to the filesystem.
103 104 105 106 107 108 109 110 111 112 |
# File 'lib/perobs/FlatFile.rb', line 103 def sync begin @f.flush @f.fsync rescue IOError => e PEROBS.log.fatal "Cannot sync flat file database: #{e.}" end @index.sync @space_list.sync end |
#write_obj_by_id(id, raw_obj) ⇒ Integer
Write the given object into the file. This method never uses in-place updates for existing objects. A new copy is inserted first and only when the insert was successful, the old copy is deleted and the index updated.
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 |
# File 'lib/perobs/FlatFile.rb', line 169 def write_obj_by_id(id, raw_obj) # Check if we have already an object with the given ID. We'll mark it as # outdated and save the header for later deletion. In case this # operation is aborted or interrupted we ensure that we either have the # old or the new version available. if (old_addr = find_obj_addr_by_id(id)) old_header = FlatFileBlobHeader.read(@f, old_addr) old_header.set_outdated_flag end crc = checksum(raw_obj) # If the raw_obj is larger then 256 characters we will compress it to # safe some space in the database file. For smaller strings the # performance impact of compression is not compensated by writing # less data to the storage. compressed = false if raw_obj.bytesize > 256 raw_obj = Zlib.deflate(raw_obj) compressed = true end addr, length = find_free_blob(raw_obj.bytesize) begin if length != -1 # Just a safeguard so we don't overwrite current data. header = FlatFileBlobHeader.read(@f, addr) if header.length != length PEROBS.log.fatal "Length in free list (#{length}) and header " + "(#{header.length}) for address #{addr} don't match." end if raw_obj.bytesize > header.length PEROBS.log.fatal "Object (#{raw_obj.bytesize}) is longer than " + "blob space (#{header.length})." end if header.is_valid? PEROBS.log.fatal "Entry at address #{addr} with flags: " + "#{header.flags} is already used for ID #{header.id}." end end flags = 1 << FlatFileBlobHeader::VALID_FLAG_BIT flags |= (1 << FlatFileBlobHeader::COMPRESSED_FLAG_BIT) if compressed FlatFileBlobHeader.new(@f, addr, flags, raw_obj.bytesize, id, crc).write @f.write(raw_obj) if length != -1 && raw_obj.bytesize < length # The new object was not appended and it did not completely fill the # free space. So we have to write a new header to mark the remaining # empty space. unless length - raw_obj.bytesize >= FlatFileBlobHeader::LENGTH PEROBS.log.fatal "Not enough space to append the empty space " + "header (space: #{length} bytes, object: #{raw_obj.bytesize} " + "bytes)." end space_address = @f.pos space_length = length - FlatFileBlobHeader::LENGTH - raw_obj.bytesize FlatFileBlobHeader.new(@f, space_address, 0, space_length, 0, 0).write # Register the new space with the space list. if @space_list.is_open? && space_length > 0 @space_list.add_space(space_address, space_length) end end # Once the blob has been written we can update the index as well. @index.insert(id, addr) if @index.is_open? if old_addr # If we had an existing object stored for the ID we have to mark # this entry as deleted now. old_header.clear_flags # And register the newly freed space with the space list. if @space_list.is_open? @space_list.add_space(old_addr, old_header.length) end else @f.flush end rescue IOError => e PEROBS.log.fatal "Cannot write blob for ID #{id} to FlatFileDB: " + e. end addr end |