作者: 韩晨旭 10225101440 李畅 10225102463
您最多选择25个主题 主题必须以字母或数字开头,可以包含连字符 (-),并且长度不得超过35个字符

332 行
11 KiB

Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
Fix snapshot compaction bug Closes google/leveldb#320 During compaction it was possible that records from a block b1=(l1,u1) would be pushed down from level i to level i+1. If there is a block b2=(l2,u2) at level i with k1 = user_key(u1) = user_key(l2) then a subsequent search for k1 will yield the record l2 which has a smaller sequence number than u1 because the sort order for records sorts increasing by user key but decreaing by sequence number. This change add a call to a new function AddBoundaryInputs to SetupOtherInputs. AddBoundaryInputs searches for a block b2 matching the criteria above and adds it to the set of files to be compacted. Whenever AddBoundaryInputs is called it is important that the compaction fileset in level i+1 (known as c->inputs_[1] in the code) be recomputed. Each call to AddBoundaryInputs is followed by a call to GetOverlappingInputs. SetupOtherInputs is called on both manual and automated compaction passes. It is called for both level zero and for levels greater than 0. The original change posted in https://github.com/google/leveldb/pull/339 has been modified to also include changed made by Chris Mumford<cmumford@google.com> in https://github.com/cmumford/leveldb/commit/4b72cb14f8da2aab12451c24b8e205aff686e9dc 1. Releasing snapshots during test cleanup to avoid memory leak warnings. 2. Refactored test to use testutil.h to be in line with other issue tests and to create the test database in the correct temporary location. 3. Added copyright banner. Otherwise, just minor formatting and limiting character width to 80 characters. Additionally the change was rebased on top of current master and changes previously made to the Makefile were ported to the CMakeLists.txt. Testing Done: A test program (issue320_test) was constructed that performs mutations while snapshots are active. issue320_test fails without this bug fix after 64k writes. It passes with this bug fix. It was run with 200M writes and passed. Unit tests were written for the new function that was added to the code. Make test was run and seen to pass. Signed-off-by: Richard Cole <richcole@amazon.com>
9 年前
  1. // Copyright (c) 2011 The LevelDB Authors. All rights reserved.
  2. // Use of this source code is governed by a BSD-style license that can be
  3. // found in the LICENSE file. See the AUTHORS file for names of contributors.
  4. #include "db/version_set.h"
  5. #include "util/logging.h"
  6. #include "util/testharness.h"
  7. #include "util/testutil.h"
  8. namespace leveldb {
  9. class FindFileTest {
  10. public:
  11. FindFileTest() : disjoint_sorted_files_(true) {}
  12. ~FindFileTest() {
  13. for (int i = 0; i < files_.size(); i++) {
  14. delete files_[i];
  15. }
  16. }
  17. void Add(const char* smallest, const char* largest,
  18. SequenceNumber smallest_seq = 100,
  19. SequenceNumber largest_seq = 100) {
  20. FileMetaData* f = new FileMetaData;
  21. f->number = files_.size() + 1;
  22. f->smallest = InternalKey(smallest, smallest_seq, kTypeValue);
  23. f->largest = InternalKey(largest, largest_seq, kTypeValue);
  24. files_.push_back(f);
  25. }
  26. int Find(const char* key) {
  27. InternalKey target(key, 100, kTypeValue);
  28. InternalKeyComparator cmp(BytewiseComparator());
  29. return FindFile(cmp, files_, target.Encode());
  30. }
  31. bool Overlaps(const char* smallest, const char* largest) {
  32. InternalKeyComparator cmp(BytewiseComparator());
  33. Slice s(smallest != nullptr ? smallest : "");
  34. Slice l(largest != nullptr ? largest : "");
  35. return SomeFileOverlapsRange(cmp, disjoint_sorted_files_, files_,
  36. (smallest != nullptr ? &s : nullptr),
  37. (largest != nullptr ? &l : nullptr));
  38. }
  39. bool disjoint_sorted_files_;
  40. private:
  41. std::vector<FileMetaData*> files_;
  42. };
  43. TEST(FindFileTest, Empty) {
  44. ASSERT_EQ(0, Find("foo"));
  45. ASSERT_TRUE(!Overlaps("a", "z"));
  46. ASSERT_TRUE(!Overlaps(nullptr, "z"));
  47. ASSERT_TRUE(!Overlaps("a", nullptr));
  48. ASSERT_TRUE(!Overlaps(nullptr, nullptr));
  49. }
  50. TEST(FindFileTest, Single) {
  51. Add("p", "q");
  52. ASSERT_EQ(0, Find("a"));
  53. ASSERT_EQ(0, Find("p"));
  54. ASSERT_EQ(0, Find("p1"));
  55. ASSERT_EQ(0, Find("q"));
  56. ASSERT_EQ(1, Find("q1"));
  57. ASSERT_EQ(1, Find("z"));
  58. ASSERT_TRUE(!Overlaps("a", "b"));
  59. ASSERT_TRUE(!Overlaps("z1", "z2"));
  60. ASSERT_TRUE(Overlaps("a", "p"));
  61. ASSERT_TRUE(Overlaps("a", "q"));
  62. ASSERT_TRUE(Overlaps("a", "z"));
  63. ASSERT_TRUE(Overlaps("p", "p1"));
  64. ASSERT_TRUE(Overlaps("p", "q"));
  65. ASSERT_TRUE(Overlaps("p", "z"));
  66. ASSERT_TRUE(Overlaps("p1", "p2"));
  67. ASSERT_TRUE(Overlaps("p1", "z"));
  68. ASSERT_TRUE(Overlaps("q", "q"));
  69. ASSERT_TRUE(Overlaps("q", "q1"));
  70. ASSERT_TRUE(!Overlaps(nullptr, "j"));
  71. ASSERT_TRUE(!Overlaps("r", nullptr));
  72. ASSERT_TRUE(Overlaps(nullptr, "p"));
  73. ASSERT_TRUE(Overlaps(nullptr, "p1"));
  74. ASSERT_TRUE(Overlaps("q", nullptr));
  75. ASSERT_TRUE(Overlaps(nullptr, nullptr));
  76. }
  77. TEST(FindFileTest, Multiple) {
  78. Add("150", "200");
  79. Add("200", "250");
  80. Add("300", "350");
  81. Add("400", "450");
  82. ASSERT_EQ(0, Find("100"));
  83. ASSERT_EQ(0, Find("150"));
  84. ASSERT_EQ(0, Find("151"));
  85. ASSERT_EQ(0, Find("199"));
  86. ASSERT_EQ(0, Find("200"));
  87. ASSERT_EQ(1, Find("201"));
  88. ASSERT_EQ(1, Find("249"));
  89. ASSERT_EQ(1, Find("250"));
  90. ASSERT_EQ(2, Find("251"));
  91. ASSERT_EQ(2, Find("299"));
  92. ASSERT_EQ(2, Find("300"));
  93. ASSERT_EQ(2, Find("349"));
  94. ASSERT_EQ(2, Find("350"));
  95. ASSERT_EQ(3, Find("351"));
  96. ASSERT_EQ(3, Find("400"));
  97. ASSERT_EQ(3, Find("450"));
  98. ASSERT_EQ(4, Find("451"));
  99. ASSERT_TRUE(!Overlaps("100", "149"));
  100. ASSERT_TRUE(!Overlaps("251", "299"));
  101. ASSERT_TRUE(!Overlaps("451", "500"));
  102. ASSERT_TRUE(!Overlaps("351", "399"));
  103. ASSERT_TRUE(Overlaps("100", "150"));
  104. ASSERT_TRUE(Overlaps("100", "200"));
  105. ASSERT_TRUE(Overlaps("100", "300"));
  106. ASSERT_TRUE(Overlaps("100", "400"));
  107. ASSERT_TRUE(Overlaps("100", "500"));
  108. ASSERT_TRUE(Overlaps("375", "400"));
  109. ASSERT_TRUE(Overlaps("450", "450"));
  110. ASSERT_TRUE(Overlaps("450", "500"));
  111. }
  112. TEST(FindFileTest, MultipleNullBoundaries) {
  113. Add("150", "200");
  114. Add("200", "250");
  115. Add("300", "350");
  116. Add("400", "450");
  117. ASSERT_TRUE(!Overlaps(nullptr, "149"));
  118. ASSERT_TRUE(!Overlaps("451", nullptr));
  119. ASSERT_TRUE(Overlaps(nullptr, nullptr));
  120. ASSERT_TRUE(Overlaps(nullptr, "150"));
  121. ASSERT_TRUE(Overlaps(nullptr, "199"));
  122. ASSERT_TRUE(Overlaps(nullptr, "200"));
  123. ASSERT_TRUE(Overlaps(nullptr, "201"));
  124. ASSERT_TRUE(Overlaps(nullptr, "400"));
  125. ASSERT_TRUE(Overlaps(nullptr, "800"));
  126. ASSERT_TRUE(Overlaps("100", nullptr));
  127. ASSERT_TRUE(Overlaps("200", nullptr));
  128. ASSERT_TRUE(Overlaps("449", nullptr));
  129. ASSERT_TRUE(Overlaps("450", nullptr));
  130. }
  131. TEST(FindFileTest, OverlapSequenceChecks) {
  132. Add("200", "200", 5000, 3000);
  133. ASSERT_TRUE(!Overlaps("199", "199"));
  134. ASSERT_TRUE(!Overlaps("201", "300"));
  135. ASSERT_TRUE(Overlaps("200", "200"));
  136. ASSERT_TRUE(Overlaps("190", "200"));
  137. ASSERT_TRUE(Overlaps("200", "210"));
  138. }
  139. TEST(FindFileTest, OverlappingFiles) {
  140. Add("150", "600");
  141. Add("400", "500");
  142. disjoint_sorted_files_ = false;
  143. ASSERT_TRUE(!Overlaps("100", "149"));
  144. ASSERT_TRUE(!Overlaps("601", "700"));
  145. ASSERT_TRUE(Overlaps("100", "150"));
  146. ASSERT_TRUE(Overlaps("100", "200"));
  147. ASSERT_TRUE(Overlaps("100", "300"));
  148. ASSERT_TRUE(Overlaps("100", "400"));
  149. ASSERT_TRUE(Overlaps("100", "500"));
  150. ASSERT_TRUE(Overlaps("375", "400"));
  151. ASSERT_TRUE(Overlaps("450", "450"));
  152. ASSERT_TRUE(Overlaps("450", "500"));
  153. ASSERT_TRUE(Overlaps("450", "700"));
  154. ASSERT_TRUE(Overlaps("600", "700"));
  155. }
  156. void AddBoundaryInputs(const InternalKeyComparator& icmp,
  157. const std::vector<FileMetaData*>& level_files,
  158. std::vector<FileMetaData*>* compaction_files);
  159. class AddBoundaryInputsTest {
  160. public:
  161. std::vector<FileMetaData*> level_files_;
  162. std::vector<FileMetaData*> compaction_files_;
  163. std::vector<FileMetaData*> all_files_;
  164. InternalKeyComparator icmp_;
  165. AddBoundaryInputsTest() : icmp_(BytewiseComparator()) {}
  166. ~AddBoundaryInputsTest() {
  167. for (size_t i = 0; i < all_files_.size(); ++i) {
  168. delete all_files_[i];
  169. }
  170. all_files_.clear();
  171. }
  172. FileMetaData* CreateFileMetaData(uint64_t number, InternalKey smallest,
  173. InternalKey largest) {
  174. FileMetaData* f = new FileMetaData();
  175. f->number = number;
  176. f->smallest = smallest;
  177. f->largest = largest;
  178. all_files_.push_back(f);
  179. return f;
  180. }
  181. };
  182. TEST(AddBoundaryInputsTest, TestEmptyFileSets) {
  183. AddBoundaryInputs(icmp_, level_files_, &compaction_files_);
  184. ASSERT_TRUE(compaction_files_.empty());
  185. ASSERT_TRUE(level_files_.empty());
  186. }
  187. TEST(AddBoundaryInputsTest, TestEmptyLevelFiles) {
  188. FileMetaData* f1 =
  189. CreateFileMetaData(1, InternalKey("100", 2, kTypeValue),
  190. InternalKey(InternalKey("100", 1, kTypeValue)));
  191. compaction_files_.push_back(f1);
  192. AddBoundaryInputs(icmp_, level_files_, &compaction_files_);
  193. ASSERT_EQ(1, compaction_files_.size());
  194. ASSERT_EQ(f1, compaction_files_[0]);
  195. ASSERT_TRUE(level_files_.empty());
  196. }
  197. TEST(AddBoundaryInputsTest, TestEmptyCompactionFiles) {
  198. FileMetaData* f1 =
  199. CreateFileMetaData(1, InternalKey("100", 2, kTypeValue),
  200. InternalKey(InternalKey("100", 1, kTypeValue)));
  201. level_files_.push_back(f1);
  202. AddBoundaryInputs(icmp_, level_files_, &compaction_files_);
  203. ASSERT_TRUE(compaction_files_.empty());
  204. ASSERT_EQ(1, level_files_.size());
  205. ASSERT_EQ(f1, level_files_[0]);
  206. }
  207. TEST(AddBoundaryInputsTest, TestNoBoundaryFiles) {
  208. FileMetaData* f1 =
  209. CreateFileMetaData(1, InternalKey("100", 2, kTypeValue),
  210. InternalKey(InternalKey("100", 1, kTypeValue)));
  211. FileMetaData* f2 =
  212. CreateFileMetaData(1, InternalKey("200", 2, kTypeValue),
  213. InternalKey(InternalKey("200", 1, kTypeValue)));
  214. FileMetaData* f3 =
  215. CreateFileMetaData(1, InternalKey("300", 2, kTypeValue),
  216. InternalKey(InternalKey("300", 1, kTypeValue)));
  217. level_files_.push_back(f3);
  218. level_files_.push_back(f2);
  219. level_files_.push_back(f1);
  220. compaction_files_.push_back(f2);
  221. compaction_files_.push_back(f3);
  222. AddBoundaryInputs(icmp_, level_files_, &compaction_files_);
  223. ASSERT_EQ(2, compaction_files_.size());
  224. }
  225. TEST(AddBoundaryInputsTest, TestOneBoundaryFiles) {
  226. FileMetaData* f1 =
  227. CreateFileMetaData(1, InternalKey("100", 3, kTypeValue),
  228. InternalKey(InternalKey("100", 2, kTypeValue)));
  229. FileMetaData* f2 =
  230. CreateFileMetaData(1, InternalKey("100", 1, kTypeValue),
  231. InternalKey(InternalKey("200", 3, kTypeValue)));
  232. FileMetaData* f3 =
  233. CreateFileMetaData(1, InternalKey("300", 2, kTypeValue),
  234. InternalKey(InternalKey("300", 1, kTypeValue)));
  235. level_files_.push_back(f3);
  236. level_files_.push_back(f2);
  237. level_files_.push_back(f1);
  238. compaction_files_.push_back(f1);
  239. AddBoundaryInputs(icmp_, level_files_, &compaction_files_);
  240. ASSERT_EQ(2, compaction_files_.size());
  241. ASSERT_EQ(f1, compaction_files_[0]);
  242. ASSERT_EQ(f2, compaction_files_[1]);
  243. }
  244. TEST(AddBoundaryInputsTest, TestTwoBoundaryFiles) {
  245. FileMetaData* f1 =
  246. CreateFileMetaData(1, InternalKey("100", 6, kTypeValue),
  247. InternalKey(InternalKey("100", 5, kTypeValue)));
  248. FileMetaData* f2 =
  249. CreateFileMetaData(1, InternalKey("100", 2, kTypeValue),
  250. InternalKey(InternalKey("300", 1, kTypeValue)));
  251. FileMetaData* f3 =
  252. CreateFileMetaData(1, InternalKey("100", 4, kTypeValue),
  253. InternalKey(InternalKey("100", 3, kTypeValue)));
  254. level_files_.push_back(f2);
  255. level_files_.push_back(f3);
  256. level_files_.push_back(f1);
  257. compaction_files_.push_back(f1);
  258. AddBoundaryInputs(icmp_, level_files_, &compaction_files_);
  259. ASSERT_EQ(3, compaction_files_.size());
  260. ASSERT_EQ(f1, compaction_files_[0]);
  261. ASSERT_EQ(f3, compaction_files_[1]);
  262. ASSERT_EQ(f2, compaction_files_[2]);
  263. }
  264. TEST(AddBoundaryInputsTest, TestDisjoinFilePointers) {
  265. FileMetaData* f1 =
  266. CreateFileMetaData(1, InternalKey("100", 6, kTypeValue),
  267. InternalKey(InternalKey("100", 5, kTypeValue)));
  268. FileMetaData* f2 =
  269. CreateFileMetaData(1, InternalKey("100", 6, kTypeValue),
  270. InternalKey(InternalKey("100", 5, kTypeValue)));
  271. FileMetaData* f3 =
  272. CreateFileMetaData(1, InternalKey("100", 2, kTypeValue),
  273. InternalKey(InternalKey("300", 1, kTypeValue)));
  274. FileMetaData* f4 =
  275. CreateFileMetaData(1, InternalKey("100", 4, kTypeValue),
  276. InternalKey(InternalKey("100", 3, kTypeValue)));
  277. level_files_.push_back(f2);
  278. level_files_.push_back(f3);
  279. level_files_.push_back(f4);
  280. compaction_files_.push_back(f1);
  281. AddBoundaryInputs(icmp_, level_files_, &compaction_files_);
  282. ASSERT_EQ(3, compaction_files_.size());
  283. ASSERT_EQ(f1, compaction_files_[0]);
  284. ASSERT_EQ(f4, compaction_files_[1]);
  285. ASSERT_EQ(f3, compaction_files_[2]);
  286. }
  287. } // namespace leveldb
  288. int main(int argc, char** argv) { return leveldb::test::RunAllTests(); }