2016-02-10 00:12:00 +01:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
2017-07-16 01:03:42 +02:00
|
|
|
// This source code is licensed under both the GPLv2 (found in the
|
|
|
|
// COPYING file in the root directory) and Apache 2.0 License
|
|
|
|
// (found in the LICENSE.Apache file in the root directory).
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
|
|
|
|
#include "db/compaction_iterator.h"
|
Compaction Support for Range Deletion
Summary:
This diff introduces RangeDelAggregator, which takes ownership of iterators
provided to it via AddTombstones(). The tombstones are organized in a two-level
map (snapshot stripe -> begin key -> tombstone). Tombstone creation avoids data
copy by holding Slices returned by the iterator, which remain valid thanks to pinning.
For compaction, we create a hierarchical range tombstone iterator with structure
matching the iterator over compaction input data. An aggregator based on that
iterator is used by CompactionIterator to determine which keys are covered by
range tombstones. In case of merge operand, the same aggregator is used by
MergeHelper. Upon finishing each file in the compaction, relevant range tombstones
are added to the output file's range tombstone metablock and file boundaries are
updated accordingly.
To check whether a key is covered by range tombstone, RangeDelAggregator::ShouldDelete()
considers tombstones in the key's snapshot stripe. When this function is used outside of
compaction, it also checks newer stripes, which can contain covering tombstones. Currently
the intra-stripe check involves a linear scan; however, in the future we plan to collapse ranges
within a stripe such that binary search can be used.
RangeDelAggregator::AddToBuilder() adds all range tombstones in the table's key-range
to a new table's range tombstone meta-block. Since range tombstones may fall in the gap
between files, we may need to extend some files' key-ranges. The strategy is (1) first file
extends as far left as possible and other files do not extend left, (2) all files extend right
until either the start of the next file or the end of the last range tombstone in the gap,
whichever comes first.
One other notable change is adding release/move semantics to ScopedArenaIterator
such that it can be used to transfer ownership of an arena-allocated iterator, similar to
how unique_ptr is used for malloc'd data.
Depends on D61473
Test Plan: compaction_iterator_test, mock_table, end-to-end tests in D63927
Reviewers: sdong, IslamAbdelRahman, wanning, yhchiang, lightmark
Reviewed By: lightmark
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62205
2016-10-18 21:04:56 +02:00
|
|
|
|
|
|
|
#include <string>
|
|
|
|
#include <vector>
|
|
|
|
|
2017-02-06 23:43:55 +01:00
|
|
|
#include "port/port.h"
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
#include "util/testharness.h"
|
|
|
|
#include "util/testutil.h"
|
|
|
|
|
|
|
|
namespace rocksdb {
|
|
|
|
|
2017-01-12 00:01:21 +01:00
|
|
|
// Expects no merging attempts.
|
|
|
|
class NoMergingMergeOp : public MergeOperator {
|
|
|
|
public:
|
2017-07-22 03:13:59 +02:00
|
|
|
bool FullMergeV2(const MergeOperationInput& merge_in,
|
|
|
|
MergeOperationOutput* merge_out) const override {
|
2017-01-12 00:01:21 +01:00
|
|
|
ADD_FAILURE();
|
|
|
|
return false;
|
|
|
|
}
|
2017-07-22 03:13:59 +02:00
|
|
|
bool PartialMergeMulti(const Slice& key,
|
|
|
|
const std::deque<Slice>& operand_list,
|
|
|
|
std::string* new_value,
|
|
|
|
Logger* logger) const override {
|
2017-01-12 00:01:21 +01:00
|
|
|
ADD_FAILURE();
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
const char* Name() const override {
|
|
|
|
return "CompactionIteratorTest NoMergingMergeOp";
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
// Compaction filter that gets stuck when it sees a particular key,
|
|
|
|
// then gets unstuck when told to.
|
|
|
|
// Always returns Decition::kRemove.
|
|
|
|
class StallingFilter : public CompactionFilter {
|
|
|
|
public:
|
2017-07-22 03:13:59 +02:00
|
|
|
virtual Decision FilterV2(int level, const Slice& key, ValueType t,
|
|
|
|
const Slice& existing_value, std::string* new_value,
|
|
|
|
std::string* skip_until) const override {
|
2017-01-12 00:01:21 +01:00
|
|
|
int k = std::atoi(key.ToString().c_str());
|
|
|
|
last_seen.store(k);
|
|
|
|
while (k >= stall_at.load()) {
|
|
|
|
std::this_thread::yield();
|
|
|
|
}
|
|
|
|
return Decision::kRemove;
|
|
|
|
}
|
|
|
|
|
|
|
|
const char* Name() const override {
|
|
|
|
return "CompactionIteratorTest StallingFilter";
|
|
|
|
}
|
|
|
|
|
|
|
|
// Wait until the filter sees a key >= k and stalls at that key.
|
|
|
|
// If `exact`, asserts that the seen key is equal to k.
|
|
|
|
void WaitForStall(int k, bool exact = true) {
|
|
|
|
stall_at.store(k);
|
|
|
|
while (last_seen.load() < k) {
|
|
|
|
std::this_thread::yield();
|
|
|
|
}
|
|
|
|
if (exact) {
|
|
|
|
EXPECT_EQ(k, last_seen.load());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Filter will stall on key >= stall_at. Advance stall_at to unstall.
|
|
|
|
mutable std::atomic<int> stall_at{0};
|
|
|
|
// Last key the filter was called with.
|
|
|
|
mutable std::atomic<int> last_seen{0};
|
|
|
|
};
|
|
|
|
|
2016-12-01 16:00:17 +01:00
|
|
|
class LoggingForwardVectorIterator : public InternalIterator {
|
|
|
|
public:
|
|
|
|
struct Action {
|
|
|
|
enum class Type {
|
|
|
|
SEEK_TO_FIRST,
|
|
|
|
SEEK,
|
|
|
|
NEXT,
|
|
|
|
};
|
|
|
|
|
|
|
|
Type type;
|
|
|
|
std::string arg;
|
|
|
|
|
|
|
|
explicit Action(Type _type, std::string _arg = "")
|
|
|
|
: type(_type), arg(_arg) {}
|
|
|
|
|
|
|
|
bool operator==(const Action& rhs) const {
|
|
|
|
return std::tie(type, arg) == std::tie(rhs.type, rhs.arg);
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
LoggingForwardVectorIterator(const std::vector<std::string>& keys,
|
|
|
|
const std::vector<std::string>& values)
|
|
|
|
: keys_(keys), values_(values), current_(keys.size()) {
|
|
|
|
assert(keys_.size() == values_.size());
|
|
|
|
}
|
|
|
|
|
|
|
|
virtual bool Valid() const override { return current_ < keys_.size(); }
|
|
|
|
|
|
|
|
virtual void SeekToFirst() override {
|
|
|
|
log.emplace_back(Action::Type::SEEK_TO_FIRST);
|
|
|
|
current_ = 0;
|
|
|
|
}
|
|
|
|
virtual void SeekToLast() override { assert(false); }
|
|
|
|
|
|
|
|
virtual void Seek(const Slice& target) override {
|
|
|
|
log.emplace_back(Action::Type::SEEK, target.ToString());
|
|
|
|
current_ = std::lower_bound(keys_.begin(), keys_.end(), target.ToString()) -
|
|
|
|
keys_.begin();
|
|
|
|
}
|
|
|
|
|
2017-07-22 03:13:59 +02:00
|
|
|
virtual void SeekForPrev(const Slice& target) override { assert(false); }
|
2016-12-01 16:00:17 +01:00
|
|
|
|
|
|
|
virtual void Next() override {
|
|
|
|
assert(Valid());
|
|
|
|
log.emplace_back(Action::Type::NEXT);
|
|
|
|
current_++;
|
|
|
|
}
|
|
|
|
virtual void Prev() override { assert(false); }
|
|
|
|
|
|
|
|
virtual Slice key() const override {
|
|
|
|
assert(Valid());
|
|
|
|
return Slice(keys_[current_]);
|
|
|
|
}
|
|
|
|
virtual Slice value() const override {
|
|
|
|
assert(Valid());
|
|
|
|
return Slice(values_[current_]);
|
|
|
|
}
|
|
|
|
|
|
|
|
virtual Status status() const override { return Status::OK(); }
|
|
|
|
|
|
|
|
std::vector<Action> log;
|
|
|
|
|
|
|
|
private:
|
|
|
|
std::vector<std::string> keys_;
|
|
|
|
std::vector<std::string> values_;
|
|
|
|
size_t current_;
|
|
|
|
};
|
|
|
|
|
|
|
|
class FakeCompaction : public CompactionIterator::CompactionProxy {
|
|
|
|
public:
|
|
|
|
FakeCompaction() = default;
|
|
|
|
|
2017-07-22 03:13:59 +02:00
|
|
|
virtual int level(size_t compaction_input_level) const { return 0; }
|
2016-12-01 16:00:17 +01:00
|
|
|
virtual bool KeyNotExistsBeyondOutputLevel(
|
2017-07-22 03:13:59 +02:00
|
|
|
const Slice& user_key, std::vector<size_t>* level_ptrs) const {
|
2017-01-12 00:01:21 +01:00
|
|
|
return key_not_exists_beyond_output_level;
|
2016-12-01 16:00:17 +01:00
|
|
|
}
|
|
|
|
virtual bool bottommost_level() const { return false; }
|
|
|
|
virtual int number_levels() const { return 1; }
|
|
|
|
virtual Slice GetLargestUserKey() const {
|
|
|
|
return "\xff\xff\xff\xff\xff\xff\xff\xff\xff";
|
|
|
|
}
|
2017-05-17 20:32:26 +02:00
|
|
|
virtual bool allow_ingest_behind() const { return false; }
|
2017-01-12 00:01:21 +01:00
|
|
|
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
2017-11-02 02:43:29 +01:00
|
|
|
virtual bool preserve_deletes() const {return false; }
|
|
|
|
|
2017-01-12 00:01:21 +01:00
|
|
|
bool key_not_exists_beyond_output_level = false;
|
2016-12-01 16:00:17 +01:00
|
|
|
};
|
|
|
|
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
class CompactionIteratorTest : public testing::Test {
|
|
|
|
public:
|
Compaction Support for Range Deletion
Summary:
This diff introduces RangeDelAggregator, which takes ownership of iterators
provided to it via AddTombstones(). The tombstones are organized in a two-level
map (snapshot stripe -> begin key -> tombstone). Tombstone creation avoids data
copy by holding Slices returned by the iterator, which remain valid thanks to pinning.
For compaction, we create a hierarchical range tombstone iterator with structure
matching the iterator over compaction input data. An aggregator based on that
iterator is used by CompactionIterator to determine which keys are covered by
range tombstones. In case of merge operand, the same aggregator is used by
MergeHelper. Upon finishing each file in the compaction, relevant range tombstones
are added to the output file's range tombstone metablock and file boundaries are
updated accordingly.
To check whether a key is covered by range tombstone, RangeDelAggregator::ShouldDelete()
considers tombstones in the key's snapshot stripe. When this function is used outside of
compaction, it also checks newer stripes, which can contain covering tombstones. Currently
the intra-stripe check involves a linear scan; however, in the future we plan to collapse ranges
within a stripe such that binary search can be used.
RangeDelAggregator::AddToBuilder() adds all range tombstones in the table's key-range
to a new table's range tombstone meta-block. Since range tombstones may fall in the gap
between files, we may need to extend some files' key-ranges. The strategy is (1) first file
extends as far left as possible and other files do not extend left, (2) all files extend right
until either the start of the next file or the end of the last range tombstone in the gap,
whichever comes first.
One other notable change is adding release/move semantics to ScopedArenaIterator
such that it can be used to transfer ownership of an arena-allocated iterator, similar to
how unique_ptr is used for malloc'd data.
Depends on D61473
Test Plan: compaction_iterator_test, mock_table, end-to-end tests in D63927
Reviewers: sdong, IslamAbdelRahman, wanning, yhchiang, lightmark
Reviewed By: lightmark
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62205
2016-10-18 21:04:56 +02:00
|
|
|
CompactionIteratorTest()
|
|
|
|
: cmp_(BytewiseComparator()), icmp_(cmp_), snapshots_({}) {}
|
|
|
|
|
|
|
|
void InitIterators(const std::vector<std::string>& ks,
|
|
|
|
const std::vector<std::string>& vs,
|
|
|
|
const std::vector<std::string>& range_del_ks,
|
|
|
|
const std::vector<std::string>& range_del_vs,
|
2016-12-01 16:00:17 +01:00
|
|
|
SequenceNumber last_sequence,
|
|
|
|
MergeOperator* merge_op = nullptr,
|
|
|
|
CompactionFilter* filter = nullptr) {
|
Compaction Support for Range Deletion
Summary:
This diff introduces RangeDelAggregator, which takes ownership of iterators
provided to it via AddTombstones(). The tombstones are organized in a two-level
map (snapshot stripe -> begin key -> tombstone). Tombstone creation avoids data
copy by holding Slices returned by the iterator, which remain valid thanks to pinning.
For compaction, we create a hierarchical range tombstone iterator with structure
matching the iterator over compaction input data. An aggregator based on that
iterator is used by CompactionIterator to determine which keys are covered by
range tombstones. In case of merge operand, the same aggregator is used by
MergeHelper. Upon finishing each file in the compaction, relevant range tombstones
are added to the output file's range tombstone metablock and file boundaries are
updated accordingly.
To check whether a key is covered by range tombstone, RangeDelAggregator::ShouldDelete()
considers tombstones in the key's snapshot stripe. When this function is used outside of
compaction, it also checks newer stripes, which can contain covering tombstones. Currently
the intra-stripe check involves a linear scan; however, in the future we plan to collapse ranges
within a stripe such that binary search can be used.
RangeDelAggregator::AddToBuilder() adds all range tombstones in the table's key-range
to a new table's range tombstone meta-block. Since range tombstones may fall in the gap
between files, we may need to extend some files' key-ranges. The strategy is (1) first file
extends as far left as possible and other files do not extend left, (2) all files extend right
until either the start of the next file or the end of the last range tombstone in the gap,
whichever comes first.
One other notable change is adding release/move semantics to ScopedArenaIterator
such that it can be used to transfer ownership of an arena-allocated iterator, similar to
how unique_ptr is used for malloc'd data.
Depends on D61473
Test Plan: compaction_iterator_test, mock_table, end-to-end tests in D63927
Reviewers: sdong, IslamAbdelRahman, wanning, yhchiang, lightmark
Reviewed By: lightmark
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62205
2016-10-18 21:04:56 +02:00
|
|
|
std::unique_ptr<InternalIterator> range_del_iter(
|
|
|
|
new test::VectorIterator(range_del_ks, range_del_vs));
|
|
|
|
range_del_agg_.reset(new RangeDelAggregator(icmp_, snapshots_));
|
2016-10-19 19:59:46 +02:00
|
|
|
ASSERT_OK(range_del_agg_->AddTombstones(std::move(range_del_iter)));
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
|
2016-12-01 16:00:17 +01:00
|
|
|
std::unique_ptr<CompactionIterator::CompactionProxy> compaction;
|
|
|
|
if (filter) {
|
2017-01-12 00:01:21 +01:00
|
|
|
compaction_proxy_ = new FakeCompaction();
|
|
|
|
compaction.reset(compaction_proxy_);
|
2016-12-01 16:00:17 +01:00
|
|
|
}
|
2017-10-06 19:26:38 +02:00
|
|
|
// TODO(yiwu) add a mock snapshot checker and add test for it.
|
|
|
|
SnapshotChecker* snapshot_checker = nullptr;
|
2016-12-01 16:00:17 +01:00
|
|
|
|
|
|
|
merge_helper_.reset(new MergeHelper(Env::Default(), cmp_, merge_op, filter,
|
2017-02-23 23:53:03 +01:00
|
|
|
nullptr, false, 0, 0, nullptr,
|
2017-01-12 00:01:21 +01:00
|
|
|
&shutting_down_));
|
2016-12-01 16:00:17 +01:00
|
|
|
iter_.reset(new LoggingForwardVectorIterator(ks, vs));
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
iter_->SeekToFirst();
|
2015-12-08 21:25:48 +01:00
|
|
|
c_iter_.reset(new CompactionIterator(
|
|
|
|
iter_.get(), cmp_, merge_helper_.get(), last_sequence, &snapshots_,
|
2017-10-06 19:26:38 +02:00
|
|
|
kMaxSequenceNumber, snapshot_checker, Env::Default(), false,
|
|
|
|
range_del_agg_.get(), std::move(compaction), filter, nullptr,
|
|
|
|
&shutting_down_));
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
}
|
|
|
|
|
Compaction Support for Range Deletion
Summary:
This diff introduces RangeDelAggregator, which takes ownership of iterators
provided to it via AddTombstones(). The tombstones are organized in a two-level
map (snapshot stripe -> begin key -> tombstone). Tombstone creation avoids data
copy by holding Slices returned by the iterator, which remain valid thanks to pinning.
For compaction, we create a hierarchical range tombstone iterator with structure
matching the iterator over compaction input data. An aggregator based on that
iterator is used by CompactionIterator to determine which keys are covered by
range tombstones. In case of merge operand, the same aggregator is used by
MergeHelper. Upon finishing each file in the compaction, relevant range tombstones
are added to the output file's range tombstone metablock and file boundaries are
updated accordingly.
To check whether a key is covered by range tombstone, RangeDelAggregator::ShouldDelete()
considers tombstones in the key's snapshot stripe. When this function is used outside of
compaction, it also checks newer stripes, which can contain covering tombstones. Currently
the intra-stripe check involves a linear scan; however, in the future we plan to collapse ranges
within a stripe such that binary search can be used.
RangeDelAggregator::AddToBuilder() adds all range tombstones in the table's key-range
to a new table's range tombstone meta-block. Since range tombstones may fall in the gap
between files, we may need to extend some files' key-ranges. The strategy is (1) first file
extends as far left as possible and other files do not extend left, (2) all files extend right
until either the start of the next file or the end of the last range tombstone in the gap,
whichever comes first.
One other notable change is adding release/move semantics to ScopedArenaIterator
such that it can be used to transfer ownership of an arena-allocated iterator, similar to
how unique_ptr is used for malloc'd data.
Depends on D61473
Test Plan: compaction_iterator_test, mock_table, end-to-end tests in D63927
Reviewers: sdong, IslamAbdelRahman, wanning, yhchiang, lightmark
Reviewed By: lightmark
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62205
2016-10-18 21:04:56 +02:00
|
|
|
void AddSnapshot(SequenceNumber snapshot) { snapshots_.push_back(snapshot); }
|
|
|
|
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
const Comparator* cmp_;
|
Compaction Support for Range Deletion
Summary:
This diff introduces RangeDelAggregator, which takes ownership of iterators
provided to it via AddTombstones(). The tombstones are organized in a two-level
map (snapshot stripe -> begin key -> tombstone). Tombstone creation avoids data
copy by holding Slices returned by the iterator, which remain valid thanks to pinning.
For compaction, we create a hierarchical range tombstone iterator with structure
matching the iterator over compaction input data. An aggregator based on that
iterator is used by CompactionIterator to determine which keys are covered by
range tombstones. In case of merge operand, the same aggregator is used by
MergeHelper. Upon finishing each file in the compaction, relevant range tombstones
are added to the output file's range tombstone metablock and file boundaries are
updated accordingly.
To check whether a key is covered by range tombstone, RangeDelAggregator::ShouldDelete()
considers tombstones in the key's snapshot stripe. When this function is used outside of
compaction, it also checks newer stripes, which can contain covering tombstones. Currently
the intra-stripe check involves a linear scan; however, in the future we plan to collapse ranges
within a stripe such that binary search can be used.
RangeDelAggregator::AddToBuilder() adds all range tombstones in the table's key-range
to a new table's range tombstone meta-block. Since range tombstones may fall in the gap
between files, we may need to extend some files' key-ranges. The strategy is (1) first file
extends as far left as possible and other files do not extend left, (2) all files extend right
until either the start of the next file or the end of the last range tombstone in the gap,
whichever comes first.
One other notable change is adding release/move semantics to ScopedArenaIterator
such that it can be used to transfer ownership of an arena-allocated iterator, similar to
how unique_ptr is used for malloc'd data.
Depends on D61473
Test Plan: compaction_iterator_test, mock_table, end-to-end tests in D63927
Reviewers: sdong, IslamAbdelRahman, wanning, yhchiang, lightmark
Reviewed By: lightmark
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62205
2016-10-18 21:04:56 +02:00
|
|
|
const InternalKeyComparator icmp_;
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
std::vector<SequenceNumber> snapshots_;
|
|
|
|
std::unique_ptr<MergeHelper> merge_helper_;
|
2016-12-01 16:00:17 +01:00
|
|
|
std::unique_ptr<LoggingForwardVectorIterator> iter_;
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
std::unique_ptr<CompactionIterator> c_iter_;
|
Compaction Support for Range Deletion
Summary:
This diff introduces RangeDelAggregator, which takes ownership of iterators
provided to it via AddTombstones(). The tombstones are organized in a two-level
map (snapshot stripe -> begin key -> tombstone). Tombstone creation avoids data
copy by holding Slices returned by the iterator, which remain valid thanks to pinning.
For compaction, we create a hierarchical range tombstone iterator with structure
matching the iterator over compaction input data. An aggregator based on that
iterator is used by CompactionIterator to determine which keys are covered by
range tombstones. In case of merge operand, the same aggregator is used by
MergeHelper. Upon finishing each file in the compaction, relevant range tombstones
are added to the output file's range tombstone metablock and file boundaries are
updated accordingly.
To check whether a key is covered by range tombstone, RangeDelAggregator::ShouldDelete()
considers tombstones in the key's snapshot stripe. When this function is used outside of
compaction, it also checks newer stripes, which can contain covering tombstones. Currently
the intra-stripe check involves a linear scan; however, in the future we plan to collapse ranges
within a stripe such that binary search can be used.
RangeDelAggregator::AddToBuilder() adds all range tombstones in the table's key-range
to a new table's range tombstone meta-block. Since range tombstones may fall in the gap
between files, we may need to extend some files' key-ranges. The strategy is (1) first file
extends as far left as possible and other files do not extend left, (2) all files extend right
until either the start of the next file or the end of the last range tombstone in the gap,
whichever comes first.
One other notable change is adding release/move semantics to ScopedArenaIterator
such that it can be used to transfer ownership of an arena-allocated iterator, similar to
how unique_ptr is used for malloc'd data.
Depends on D61473
Test Plan: compaction_iterator_test, mock_table, end-to-end tests in D63927
Reviewers: sdong, IslamAbdelRahman, wanning, yhchiang, lightmark
Reviewed By: lightmark
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62205
2016-10-18 21:04:56 +02:00
|
|
|
std::unique_ptr<RangeDelAggregator> range_del_agg_;
|
2017-01-12 00:01:21 +01:00
|
|
|
std::atomic<bool> shutting_down_{false};
|
|
|
|
FakeCompaction* compaction_proxy_;
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
// It is possible that the output of the compaction iterator is empty even if
|
|
|
|
// the input is not.
|
|
|
|
TEST_F(CompactionIteratorTest, EmptyResult) {
|
Compaction Support for Range Deletion
Summary:
This diff introduces RangeDelAggregator, which takes ownership of iterators
provided to it via AddTombstones(). The tombstones are organized in a two-level
map (snapshot stripe -> begin key -> tombstone). Tombstone creation avoids data
copy by holding Slices returned by the iterator, which remain valid thanks to pinning.
For compaction, we create a hierarchical range tombstone iterator with structure
matching the iterator over compaction input data. An aggregator based on that
iterator is used by CompactionIterator to determine which keys are covered by
range tombstones. In case of merge operand, the same aggregator is used by
MergeHelper. Upon finishing each file in the compaction, relevant range tombstones
are added to the output file's range tombstone metablock and file boundaries are
updated accordingly.
To check whether a key is covered by range tombstone, RangeDelAggregator::ShouldDelete()
considers tombstones in the key's snapshot stripe. When this function is used outside of
compaction, it also checks newer stripes, which can contain covering tombstones. Currently
the intra-stripe check involves a linear scan; however, in the future we plan to collapse ranges
within a stripe such that binary search can be used.
RangeDelAggregator::AddToBuilder() adds all range tombstones in the table's key-range
to a new table's range tombstone meta-block. Since range tombstones may fall in the gap
between files, we may need to extend some files' key-ranges. The strategy is (1) first file
extends as far left as possible and other files do not extend left, (2) all files extend right
until either the start of the next file or the end of the last range tombstone in the gap,
whichever comes first.
One other notable change is adding release/move semantics to ScopedArenaIterator
such that it can be used to transfer ownership of an arena-allocated iterator, similar to
how unique_ptr is used for malloc'd data.
Depends on D61473
Test Plan: compaction_iterator_test, mock_table, end-to-end tests in D63927
Reviewers: sdong, IslamAbdelRahman, wanning, yhchiang, lightmark
Reviewed By: lightmark
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62205
2016-10-18 21:04:56 +02:00
|
|
|
InitIterators({test::KeyStr("a", 5, kTypeSingleDeletion),
|
|
|
|
test::KeyStr("a", 3, kTypeValue)},
|
|
|
|
{"", "val"}, {}, {}, 5);
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
c_iter_->SeekToFirst();
|
|
|
|
ASSERT_FALSE(c_iter_->Valid());
|
|
|
|
}
|
|
|
|
|
|
|
|
// If there is a corruption after a single deletion, the corrupted key should
|
|
|
|
// be preserved.
|
|
|
|
TEST_F(CompactionIteratorTest, CorruptionAfterSingleDeletion) {
|
Compaction Support for Range Deletion
Summary:
This diff introduces RangeDelAggregator, which takes ownership of iterators
provided to it via AddTombstones(). The tombstones are organized in a two-level
map (snapshot stripe -> begin key -> tombstone). Tombstone creation avoids data
copy by holding Slices returned by the iterator, which remain valid thanks to pinning.
For compaction, we create a hierarchical range tombstone iterator with structure
matching the iterator over compaction input data. An aggregator based on that
iterator is used by CompactionIterator to determine which keys are covered by
range tombstones. In case of merge operand, the same aggregator is used by
MergeHelper. Upon finishing each file in the compaction, relevant range tombstones
are added to the output file's range tombstone metablock and file boundaries are
updated accordingly.
To check whether a key is covered by range tombstone, RangeDelAggregator::ShouldDelete()
considers tombstones in the key's snapshot stripe. When this function is used outside of
compaction, it also checks newer stripes, which can contain covering tombstones. Currently
the intra-stripe check involves a linear scan; however, in the future we plan to collapse ranges
within a stripe such that binary search can be used.
RangeDelAggregator::AddToBuilder() adds all range tombstones in the table's key-range
to a new table's range tombstone meta-block. Since range tombstones may fall in the gap
between files, we may need to extend some files' key-ranges. The strategy is (1) first file
extends as far left as possible and other files do not extend left, (2) all files extend right
until either the start of the next file or the end of the last range tombstone in the gap,
whichever comes first.
One other notable change is adding release/move semantics to ScopedArenaIterator
such that it can be used to transfer ownership of an arena-allocated iterator, similar to
how unique_ptr is used for malloc'd data.
Depends on D61473
Test Plan: compaction_iterator_test, mock_table, end-to-end tests in D63927
Reviewers: sdong, IslamAbdelRahman, wanning, yhchiang, lightmark
Reviewed By: lightmark
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62205
2016-10-18 21:04:56 +02:00
|
|
|
InitIterators({test::KeyStr("a", 5, kTypeSingleDeletion),
|
|
|
|
test::KeyStr("a", 3, kTypeValue, true),
|
|
|
|
test::KeyStr("b", 10, kTypeValue)},
|
|
|
|
{"", "val", "val2"}, {}, {}, 10);
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
c_iter_->SeekToFirst();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("a", 5, kTypeSingleDeletion),
|
|
|
|
c_iter_->key().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("a", 3, kTypeValue, true), c_iter_->key().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("b", 10, kTypeValue), c_iter_->key().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_FALSE(c_iter_->Valid());
|
|
|
|
}
|
|
|
|
|
Compaction Support for Range Deletion
Summary:
This diff introduces RangeDelAggregator, which takes ownership of iterators
provided to it via AddTombstones(). The tombstones are organized in a two-level
map (snapshot stripe -> begin key -> tombstone). Tombstone creation avoids data
copy by holding Slices returned by the iterator, which remain valid thanks to pinning.
For compaction, we create a hierarchical range tombstone iterator with structure
matching the iterator over compaction input data. An aggregator based on that
iterator is used by CompactionIterator to determine which keys are covered by
range tombstones. In case of merge operand, the same aggregator is used by
MergeHelper. Upon finishing each file in the compaction, relevant range tombstones
are added to the output file's range tombstone metablock and file boundaries are
updated accordingly.
To check whether a key is covered by range tombstone, RangeDelAggregator::ShouldDelete()
considers tombstones in the key's snapshot stripe. When this function is used outside of
compaction, it also checks newer stripes, which can contain covering tombstones. Currently
the intra-stripe check involves a linear scan; however, in the future we plan to collapse ranges
within a stripe such that binary search can be used.
RangeDelAggregator::AddToBuilder() adds all range tombstones in the table's key-range
to a new table's range tombstone meta-block. Since range tombstones may fall in the gap
between files, we may need to extend some files' key-ranges. The strategy is (1) first file
extends as far left as possible and other files do not extend left, (2) all files extend right
until either the start of the next file or the end of the last range tombstone in the gap,
whichever comes first.
One other notable change is adding release/move semantics to ScopedArenaIterator
such that it can be used to transfer ownership of an arena-allocated iterator, similar to
how unique_ptr is used for malloc'd data.
Depends on D61473
Test Plan: compaction_iterator_test, mock_table, end-to-end tests in D63927
Reviewers: sdong, IslamAbdelRahman, wanning, yhchiang, lightmark
Reviewed By: lightmark
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D62205
2016-10-18 21:04:56 +02:00
|
|
|
TEST_F(CompactionIteratorTest, SimpleRangeDeletion) {
|
|
|
|
InitIterators({test::KeyStr("morning", 5, kTypeValue),
|
|
|
|
test::KeyStr("morning", 2, kTypeValue),
|
|
|
|
test::KeyStr("night", 3, kTypeValue)},
|
|
|
|
{"zao", "zao", "wan"},
|
|
|
|
{test::KeyStr("ma", 4, kTypeRangeDeletion)}, {"mz"}, 5);
|
|
|
|
c_iter_->SeekToFirst();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("morning", 5, kTypeValue), c_iter_->key().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("night", 3, kTypeValue), c_iter_->key().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_FALSE(c_iter_->Valid());
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(CompactionIteratorTest, RangeDeletionWithSnapshots) {
|
|
|
|
AddSnapshot(10);
|
|
|
|
std::vector<std::string> ks1;
|
|
|
|
ks1.push_back(test::KeyStr("ma", 28, kTypeRangeDeletion));
|
|
|
|
std::vector<std::string> vs1{"mz"};
|
|
|
|
std::vector<std::string> ks2{test::KeyStr("morning", 15, kTypeValue),
|
|
|
|
test::KeyStr("morning", 5, kTypeValue),
|
|
|
|
test::KeyStr("night", 40, kTypeValue),
|
|
|
|
test::KeyStr("night", 20, kTypeValue)};
|
|
|
|
std::vector<std::string> vs2{"zao 15", "zao 5", "wan 40", "wan 20"};
|
|
|
|
InitIterators(ks2, vs2, ks1, vs1, 40);
|
|
|
|
c_iter_->SeekToFirst();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("morning", 5, kTypeValue), c_iter_->key().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("night", 40, kTypeValue), c_iter_->key().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_FALSE(c_iter_->Valid());
|
|
|
|
}
|
|
|
|
|
2016-12-01 16:00:17 +01:00
|
|
|
TEST_F(CompactionIteratorTest, CompactionFilterSkipUntil) {
|
|
|
|
class Filter : public CompactionFilter {
|
2017-07-22 03:13:59 +02:00
|
|
|
virtual Decision FilterV2(int level, const Slice& key, ValueType t,
|
2016-12-01 16:00:17 +01:00
|
|
|
const Slice& existing_value,
|
2017-07-22 03:13:59 +02:00
|
|
|
std::string* new_value,
|
2016-12-02 03:24:13 +01:00
|
|
|
std::string* skip_until) const override {
|
2016-12-01 16:00:17 +01:00
|
|
|
std::string k = key.ToString();
|
|
|
|
std::string v = existing_value.ToString();
|
|
|
|
// See InitIterators() call below for the sequence of keys and their
|
|
|
|
// filtering decisions. Here we closely assert that compaction filter is
|
|
|
|
// called with the expected keys and only them, and with the right values.
|
|
|
|
if (k == "a") {
|
|
|
|
EXPECT_EQ(ValueType::kValue, t);
|
|
|
|
EXPECT_EQ("av50", v);
|
|
|
|
return Decision::kKeep;
|
|
|
|
}
|
|
|
|
if (k == "b") {
|
|
|
|
EXPECT_EQ(ValueType::kValue, t);
|
|
|
|
EXPECT_EQ("bv60", v);
|
|
|
|
*skip_until = "d+";
|
|
|
|
return Decision::kRemoveAndSkipUntil;
|
|
|
|
}
|
|
|
|
if (k == "e") {
|
|
|
|
EXPECT_EQ(ValueType::kMergeOperand, t);
|
|
|
|
EXPECT_EQ("em71", v);
|
|
|
|
return Decision::kKeep;
|
|
|
|
}
|
|
|
|
if (k == "f") {
|
|
|
|
if (v == "fm65") {
|
|
|
|
EXPECT_EQ(ValueType::kMergeOperand, t);
|
|
|
|
*skip_until = "f";
|
|
|
|
} else {
|
|
|
|
EXPECT_EQ("fm30", v);
|
|
|
|
EXPECT_EQ(ValueType::kMergeOperand, t);
|
|
|
|
*skip_until = "g+";
|
|
|
|
}
|
|
|
|
return Decision::kRemoveAndSkipUntil;
|
|
|
|
}
|
|
|
|
if (k == "h") {
|
|
|
|
EXPECT_EQ(ValueType::kValue, t);
|
|
|
|
EXPECT_EQ("hv91", v);
|
|
|
|
return Decision::kKeep;
|
|
|
|
}
|
|
|
|
if (k == "i") {
|
2016-12-06 00:07:01 +01:00
|
|
|
EXPECT_EQ(ValueType::kMergeOperand, t);
|
|
|
|
EXPECT_EQ("im95", v);
|
2016-12-01 16:00:17 +01:00
|
|
|
*skip_until = "z";
|
|
|
|
return Decision::kRemoveAndSkipUntil;
|
|
|
|
}
|
|
|
|
ADD_FAILURE();
|
|
|
|
return Decision::kKeep;
|
|
|
|
}
|
|
|
|
|
|
|
|
const char* Name() const override {
|
|
|
|
return "CompactionIteratorTest.CompactionFilterSkipUntil::Filter";
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
2017-01-12 00:01:21 +01:00
|
|
|
NoMergingMergeOp merge_op;
|
2016-12-01 16:00:17 +01:00
|
|
|
Filter filter;
|
|
|
|
InitIterators(
|
|
|
|
{test::KeyStr("a", 50, kTypeValue), // keep
|
|
|
|
test::KeyStr("a", 45, kTypeMerge),
|
|
|
|
test::KeyStr("b", 60, kTypeValue), // skip to "d+"
|
|
|
|
test::KeyStr("b", 40, kTypeValue), test::KeyStr("c", 35, kTypeValue),
|
|
|
|
test::KeyStr("d", 70, kTypeMerge),
|
|
|
|
test::KeyStr("e", 71, kTypeMerge), // keep
|
|
|
|
test::KeyStr("f", 65, kTypeMerge), // skip to "f", aka keep
|
|
|
|
test::KeyStr("f", 30, kTypeMerge), // skip to "g+"
|
|
|
|
test::KeyStr("f", 25, kTypeValue), test::KeyStr("g", 90, kTypeValue),
|
|
|
|
test::KeyStr("h", 91, kTypeValue), // keep
|
2016-12-06 00:07:01 +01:00
|
|
|
test::KeyStr("i", 95, kTypeMerge), // skip to "z"
|
2016-12-01 16:00:17 +01:00
|
|
|
test::KeyStr("j", 99, kTypeValue)},
|
|
|
|
{"av50", "am45", "bv60", "bv40", "cv35", "dm70", "em71", "fm65", "fm30",
|
2016-12-06 00:07:01 +01:00
|
|
|
"fv25", "gv90", "hv91", "im95", "jv99"},
|
2016-12-01 16:00:17 +01:00
|
|
|
{}, {}, kMaxSequenceNumber, &merge_op, &filter);
|
|
|
|
|
|
|
|
// Compaction should output just "a", "e" and "h" keys.
|
|
|
|
c_iter_->SeekToFirst();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("a", 50, kTypeValue), c_iter_->key().ToString());
|
|
|
|
ASSERT_EQ("av50", c_iter_->value().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("e", 71, kTypeMerge), c_iter_->key().ToString());
|
|
|
|
ASSERT_EQ("em71", c_iter_->value().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("h", 91, kTypeValue), c_iter_->key().ToString());
|
|
|
|
ASSERT_EQ("hv91", c_iter_->value().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_FALSE(c_iter_->Valid());
|
|
|
|
|
|
|
|
// Check that the compaction iterator did the correct sequence of calls on
|
|
|
|
// the underlying iterator.
|
|
|
|
using A = LoggingForwardVectorIterator::Action;
|
|
|
|
using T = A::Type;
|
|
|
|
std::vector<A> expected_actions = {
|
|
|
|
A(T::SEEK_TO_FIRST),
|
|
|
|
A(T::NEXT),
|
|
|
|
A(T::NEXT),
|
|
|
|
A(T::SEEK, test::KeyStr("d+", kMaxSequenceNumber, kValueTypeForSeek)),
|
|
|
|
A(T::NEXT),
|
|
|
|
A(T::NEXT),
|
|
|
|
A(T::SEEK, test::KeyStr("g+", kMaxSequenceNumber, kValueTypeForSeek)),
|
|
|
|
A(T::NEXT),
|
|
|
|
A(T::SEEK, test::KeyStr("z", kMaxSequenceNumber, kValueTypeForSeek))};
|
|
|
|
ASSERT_EQ(expected_actions, iter_->log);
|
|
|
|
}
|
|
|
|
|
2017-01-12 00:01:21 +01:00
|
|
|
TEST_F(CompactionIteratorTest, ShuttingDownInFilter) {
|
|
|
|
NoMergingMergeOp merge_op;
|
|
|
|
StallingFilter filter;
|
|
|
|
InitIterators(
|
|
|
|
{test::KeyStr("1", 1, kTypeValue), test::KeyStr("2", 2, kTypeValue),
|
|
|
|
test::KeyStr("3", 3, kTypeValue), test::KeyStr("4", 4, kTypeValue)},
|
|
|
|
{"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, &merge_op, &filter);
|
|
|
|
// Don't leave tombstones (kTypeDeletion) for filtered keys.
|
|
|
|
compaction_proxy_->key_not_exists_beyond_output_level = true;
|
|
|
|
|
|
|
|
std::atomic<bool> seek_done{false};
|
2017-02-06 23:43:55 +01:00
|
|
|
rocksdb::port::Thread compaction_thread([&] {
|
2017-01-12 00:01:21 +01:00
|
|
|
c_iter_->SeekToFirst();
|
|
|
|
EXPECT_FALSE(c_iter_->Valid());
|
|
|
|
EXPECT_TRUE(c_iter_->status().IsShutdownInProgress());
|
|
|
|
seek_done.store(true);
|
|
|
|
});
|
|
|
|
|
|
|
|
// Let key 1 through.
|
|
|
|
filter.WaitForStall(1);
|
|
|
|
|
|
|
|
// Shutdown during compaction filter call for key 2.
|
|
|
|
filter.WaitForStall(2);
|
|
|
|
shutting_down_.store(true);
|
|
|
|
EXPECT_FALSE(seek_done.load());
|
|
|
|
|
|
|
|
// Unstall filter and wait for SeekToFirst() to return.
|
|
|
|
filter.stall_at.store(3);
|
|
|
|
compaction_thread.join();
|
|
|
|
assert(seek_done.load());
|
|
|
|
|
|
|
|
// Check that filter was never called again.
|
|
|
|
EXPECT_EQ(2, filter.last_seen.load());
|
|
|
|
}
|
|
|
|
|
|
|
|
// Same as ShuttingDownInFilter, but shutdown happens during filter call for
|
|
|
|
// a merge operand, not for a value.
|
|
|
|
TEST_F(CompactionIteratorTest, ShuttingDownInMerge) {
|
|
|
|
NoMergingMergeOp merge_op;
|
|
|
|
StallingFilter filter;
|
|
|
|
InitIterators(
|
|
|
|
{test::KeyStr("1", 1, kTypeValue), test::KeyStr("2", 2, kTypeMerge),
|
|
|
|
test::KeyStr("3", 3, kTypeMerge), test::KeyStr("4", 4, kTypeValue)},
|
|
|
|
{"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, &merge_op, &filter);
|
|
|
|
compaction_proxy_->key_not_exists_beyond_output_level = true;
|
|
|
|
|
|
|
|
std::atomic<bool> seek_done{false};
|
2017-02-06 23:43:55 +01:00
|
|
|
rocksdb::port::Thread compaction_thread([&] {
|
2017-01-12 00:01:21 +01:00
|
|
|
c_iter_->SeekToFirst();
|
|
|
|
ASSERT_FALSE(c_iter_->Valid());
|
|
|
|
ASSERT_TRUE(c_iter_->status().IsShutdownInProgress());
|
|
|
|
seek_done.store(true);
|
|
|
|
});
|
|
|
|
|
|
|
|
// Let key 1 through.
|
|
|
|
filter.WaitForStall(1);
|
|
|
|
|
|
|
|
// Shutdown during compaction filter call for key 2.
|
|
|
|
filter.WaitForStall(2);
|
|
|
|
shutting_down_.store(true);
|
|
|
|
EXPECT_FALSE(seek_done.load());
|
|
|
|
|
|
|
|
// Unstall filter and wait for SeekToFirst() to return.
|
|
|
|
filter.stall_at.store(3);
|
|
|
|
compaction_thread.join();
|
|
|
|
assert(seek_done.load());
|
|
|
|
|
|
|
|
// Check that filter was never called again.
|
|
|
|
EXPECT_EQ(2, filter.last_seen.load());
|
|
|
|
}
|
|
|
|
|
2017-08-17 08:38:04 +02:00
|
|
|
TEST_F(CompactionIteratorTest, SingleMergeOperand) {
|
|
|
|
class Filter : public CompactionFilter {
|
|
|
|
virtual Decision FilterV2(int level, const Slice& key, ValueType t,
|
|
|
|
const Slice& existing_value,
|
|
|
|
std::string* new_value,
|
|
|
|
std::string* skip_until) const override {
|
|
|
|
std::string k = key.ToString();
|
|
|
|
std::string v = existing_value.ToString();
|
|
|
|
|
|
|
|
// See InitIterators() call below for the sequence of keys and their
|
|
|
|
// filtering decisions. Here we closely assert that compaction filter is
|
|
|
|
// called with the expected keys and only them, and with the right values.
|
|
|
|
if (k == "a") {
|
|
|
|
EXPECT_EQ(ValueType::kMergeOperand, t);
|
|
|
|
EXPECT_EQ("av1", v);
|
|
|
|
return Decision::kKeep;
|
|
|
|
} else if (k == "b") {
|
|
|
|
EXPECT_EQ(ValueType::kMergeOperand, t);
|
|
|
|
return Decision::kKeep;
|
|
|
|
} else if (k == "c") {
|
|
|
|
return Decision::kKeep;
|
|
|
|
}
|
|
|
|
|
|
|
|
ADD_FAILURE();
|
|
|
|
return Decision::kKeep;
|
|
|
|
}
|
|
|
|
|
|
|
|
const char* Name() const override {
|
|
|
|
return "CompactionIteratorTest.SingleMergeOperand::Filter";
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
class SingleMergeOp : public MergeOperator {
|
|
|
|
public:
|
|
|
|
bool FullMergeV2(const MergeOperationInput& merge_in,
|
|
|
|
MergeOperationOutput* merge_out) const override {
|
|
|
|
// See InitIterators() call below for why "c" is the only key for which
|
|
|
|
// FullMergeV2 should be called.
|
|
|
|
EXPECT_EQ("c", merge_in.key.ToString());
|
|
|
|
|
|
|
|
std::string temp_value;
|
|
|
|
if (merge_in.existing_value != nullptr) {
|
|
|
|
temp_value = merge_in.existing_value->ToString();
|
|
|
|
}
|
|
|
|
|
|
|
|
for (auto& operand : merge_in.operand_list) {
|
|
|
|
temp_value.append(operand.ToString());
|
|
|
|
}
|
|
|
|
merge_out->new_value = temp_value;
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool PartialMergeMulti(const Slice& key,
|
|
|
|
const std::deque<Slice>& operand_list,
|
|
|
|
std::string* new_value,
|
|
|
|
Logger* logger) const override {
|
|
|
|
std::string string_key = key.ToString();
|
|
|
|
EXPECT_TRUE(string_key == "a" || string_key == "b");
|
|
|
|
|
|
|
|
if (string_key == "a") {
|
|
|
|
EXPECT_EQ(1, operand_list.size());
|
|
|
|
} else if (string_key == "b") {
|
|
|
|
EXPECT_EQ(2, operand_list.size());
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string temp_value;
|
|
|
|
for (auto& operand : operand_list) {
|
|
|
|
temp_value.append(operand.ToString());
|
|
|
|
}
|
|
|
|
swap(temp_value, *new_value);
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
const char* Name() const override {
|
|
|
|
return "CompactionIteratorTest SingleMergeOp";
|
|
|
|
}
|
|
|
|
|
|
|
|
bool AllowSingleOperand() const override { return true; }
|
|
|
|
};
|
|
|
|
|
|
|
|
SingleMergeOp merge_op;
|
|
|
|
Filter filter;
|
|
|
|
InitIterators(
|
|
|
|
// a should invoke PartialMergeMulti with a single merge operand.
|
|
|
|
{test::KeyStr("a", 50, kTypeMerge),
|
|
|
|
// b should invoke PartialMergeMulti with two operands.
|
|
|
|
test::KeyStr("b", 70, kTypeMerge), test::KeyStr("b", 60, kTypeMerge),
|
|
|
|
// c should invoke FullMerge due to kTypeValue at the beginning.
|
|
|
|
test::KeyStr("c", 90, kTypeMerge), test::KeyStr("c", 80, kTypeValue)},
|
|
|
|
{"av1", "bv2", "bv1", "cv2", "cv1"}, {}, {}, kMaxSequenceNumber,
|
|
|
|
&merge_op, &filter);
|
|
|
|
|
|
|
|
c_iter_->SeekToFirst();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ(test::KeyStr("a", 50, kTypeMerge), c_iter_->key().ToString());
|
|
|
|
ASSERT_EQ("av1", c_iter_->value().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_TRUE(c_iter_->Valid());
|
|
|
|
ASSERT_EQ("bv1bv2", c_iter_->value().ToString());
|
|
|
|
c_iter_->Next();
|
|
|
|
ASSERT_EQ("cv1cv2", c_iter_->value().ToString());
|
|
|
|
}
|
|
|
|
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
} // namespace rocksdb
|
|
|
|
|
|
|
|
int main(int argc, char** argv) {
|
|
|
|
::testing::InitGoogleTest(&argc, argv);
|
|
|
|
return RUN_ALL_TESTS();
|
|
|
|
}
|