ClickHouse/src/Storages/MergeTree/MergeTreeDataMergerMutator.h
2020-12-04 17:01:59 +03:00

273 lines
11 KiB
C++

#pragma once
#include <Storages/MergeTree/MergeTreeData.h>
#include <Storages/MutationCommands.h>
#include <atomic>
#include <functional>
#include <Common/ActionBlocker.h>
#include <Storages/MergeTree/TTLMergeSelector.h>
#include <Storages/MergeTree/MergeAlgorithm.h>
#include <Storages/MergeTree/MergeType.h>
namespace DB
{
class MergeProgressCallback;
enum class SelectPartsDecision
{
SELECTED = 0,
CANNOT_SELECT = 1,
NOTHING_TO_MERGE = 2,
};
/// Auxiliary struct holding metainformation for the future merged or mutated part.
struct FutureMergedMutatedPart
{
String name;
UUID uuid = UUIDHelpers::Nil;
String path;
MergeTreeDataPartType type;
MergeTreePartInfo part_info;
MergeTreeData::DataPartsVector parts;
MergeType merge_type = MergeType::REGULAR;
const MergeTreePartition & getPartition() const { return parts.front()->partition; }
FutureMergedMutatedPart() = default;
explicit FutureMergedMutatedPart(MergeTreeData::DataPartsVector parts_)
{
assign(std::move(parts_));
}
FutureMergedMutatedPart(MergeTreeData::DataPartsVector parts_, MergeTreeDataPartType future_part_type)
{
assign(std::move(parts_), future_part_type);
}
void assign(MergeTreeData::DataPartsVector parts_);
void assign(MergeTreeData::DataPartsVector parts_, MergeTreeDataPartType future_part_type);
void updatePath(const MergeTreeData & storage, const ReservationPtr & reservation);
};
/** Can select parts for background processes and do them.
* Currently helps with merges, mutations and moves
*/
class MergeTreeDataMergerMutator
{
public:
using AllowedMergingPredicate = std::function<bool (const MergeTreeData::DataPartPtr &, const MergeTreeData::DataPartPtr &, String *)>;
MergeTreeDataMergerMutator(MergeTreeData & data_, size_t background_pool_size);
/** Get maximum total size of parts to do merge, at current moment of time.
* It depends on number of free threads in background_pool and amount of free space in disk.
*/
UInt64 getMaxSourcePartsSizeForMerge() const;
/** For explicitly passed size of pool and number of used tasks.
* This method could be used to calculate threshold depending on number of tasks in replication queue.
*/
UInt64 getMaxSourcePartsSizeForMerge(size_t pool_size, size_t pool_used) const;
/** Get maximum total size of parts to do mutation, at current moment of time.
* It depends only on amount of free space in disk.
*/
UInt64 getMaxSourcePartSizeForMutation() const;
/** Selects which parts to merge. Uses a lot of heuristics.
*
* can_merge - a function that determines if it is possible to merge a pair of adjacent parts.
* This function must coordinate merge with inserts and other merges, ensuring that
* - Parts between which another part can still appear can not be merged. Refer to METR-7001.
* - A part that already merges with something in one place, you can not start to merge into something else in another place.
*/
SelectPartsDecision selectPartsToMerge(
FutureMergedMutatedPart & future_part,
bool aggressive,
size_t max_total_size_to_merge,
const AllowedMergingPredicate & can_merge,
bool merge_with_ttl_allowed,
String * out_disable_reason = nullptr);
/** Select all the parts in the specified partition for merge, if possible.
* final - choose to merge even a single part - that is, allow to merge one part "with itself",
* but if setting optimize_skip_merged_partitions is true than single part with level > 0
* and without expired TTL won't be merged with itself.
*/
SelectPartsDecision selectAllPartsToMergeWithinPartition(
FutureMergedMutatedPart & future_part,
UInt64 & available_disk_space,
const AllowedMergingPredicate & can_merge,
const String & partition_id,
bool final,
const StorageMetadataPtr & metadata_snapshot,
String * out_disable_reason = nullptr,
bool optimize_skip_merged_partitions = false);
/** Merge the parts.
* If `reservation != nullptr`, now and then reduces the size of the reserved space
* is approximately proportional to the amount of data already written.
*
* Creates and returns a temporary part.
* To end the merge, call the function renameMergedTemporaryPart.
*
* time_of_merge - the time when the merge was assigned.
* Important when using ReplicatedGraphiteMergeTree to provide the same merge on replicas.
*/
MergeTreeData::MutableDataPartPtr mergePartsToTemporaryPart(
const FutureMergedMutatedPart & future_part,
const StorageMetadataPtr & metadata_snapshot,
MergeListEntry & merge_entry,
TableLockHolder & table_lock_holder,
time_t time_of_merge,
const Context & context,
const ReservationPtr & space_reservation,
bool deduplicate);
/// Mutate a single data part with the specified commands. Will create and return a temporary part.
MergeTreeData::MutableDataPartPtr mutatePartToTemporaryPart(
const FutureMergedMutatedPart & future_part,
const StorageMetadataPtr & metadata_snapshot,
const MutationCommands & commands,
MergeListEntry & merge_entry,
time_t time_of_mutation,
const Context & context,
const ReservationPtr & space_reservation,
TableLockHolder & table_lock_holder);
MergeTreeData::DataPartPtr renameMergedTemporaryPart(
MergeTreeData::MutableDataPartPtr & new_data_part,
const MergeTreeData::DataPartsVector & parts,
MergeTreeData::Transaction * out_transaction = nullptr);
/// The approximate amount of disk space needed for merge or mutation. With a surplus.
static size_t estimateNeededDiskSpace(const MergeTreeData::DataPartsVector & source_parts);
private:
/** Select all parts belonging to the same partition.
*/
MergeTreeData::DataPartsVector selectAllPartsFromPartition(const String & partition_id);
/** Split mutation commands into two parts:
* First part should be executed by mutations interpreter.
* Other is just simple drop/renames, so they can be executed without interpreter.
*/
static void splitMutationCommands(
MergeTreeData::DataPartPtr part,
const MutationCommands & commands,
MutationCommands & for_interpreter,
MutationCommands & for_file_renames);
/// Apply commands to source_part i.e. remove and rename some columns in
/// source_part and return set of files, that have to be removed or renamed
/// from filesystem and in-memory checksums. Ordered result is important,
/// because we can apply renames that affects each other: x -> z, y -> x.
static NameToNameVector collectFilesForRenames(MergeTreeData::DataPartPtr source_part, const MutationCommands & commands_for_removes, const String & mrk_extension);
/// Files, that we don't need to remove and don't need to hardlink, for example columns.txt and checksums.txt.
/// Because we will generate new versions of them after we perform mutation.
static NameSet collectFilesToSkip(
const MergeTreeDataPartPtr & source_part,
const Block & updated_header,
const std::set<MergeTreeIndexPtr> & indices_to_recalc,
const String & mrk_extension);
/// Get the columns list of the resulting part in the same order as storage_columns.
static NamesAndTypesList getColumnsForNewDataPart(
MergeTreeData::DataPartPtr source_part,
const Block & updated_header,
NamesAndTypesList storage_columns,
const MutationCommands & commands_for_removes);
/// Get skip indices, that should exists in the resulting data part.
static MergeTreeIndices getIndicesForNewDataPart(
const IndicesDescription & all_indices,
const MutationCommands & commands_for_removes);
static bool shouldExecuteTTL(const StorageMetadataPtr & metadata_snapshot, const Names & columns, const MutationCommands & commands);
/// Return set of indices which should be recalculated during mutation also
/// wraps input stream into additional expression stream
static std::set<MergeTreeIndexPtr> getIndicesToRecalculate(
BlockInputStreamPtr & input_stream,
const NamesAndTypesList & updated_columns,
const StorageMetadataPtr & metadata_snapshot,
const Context & context);
/// Override all columns of new part using mutating_stream
void mutateAllPartColumns(
MergeTreeData::MutableDataPartPtr new_data_part,
const StorageMetadataPtr & metadata_snapshot,
const MergeTreeIndices & skip_indices,
BlockInputStreamPtr mutating_stream,
time_t time_of_mutation,
const CompressionCodecPtr & codec,
MergeListEntry & merge_entry,
bool need_remove_expired_values,
bool need_sync) const;
/// Mutate some columns of source part with mutation_stream
void mutateSomePartColumns(
const MergeTreeDataPartPtr & source_part,
const StorageMetadataPtr & metadata_snapshot,
const std::set<MergeTreeIndexPtr> & indices_to_recalc,
const Block & mutation_header,
MergeTreeData::MutableDataPartPtr new_data_part,
BlockInputStreamPtr mutating_stream,
time_t time_of_mutation,
const CompressionCodecPtr & codec,
MergeListEntry & merge_entry,
bool need_remove_expired_values,
bool need_sync) const;
/// Initialize and write to disk new part fields like checksums, columns,
/// etc.
static void finalizeMutatedPart(
const MergeTreeDataPartPtr & source_part,
MergeTreeData::MutableDataPartPtr new_data_part,
bool need_remove_expired_values,
const CompressionCodecPtr & codec);
public :
/** Is used to cancel all merges and mutations. On cancel() call all currently running actions will throw exception soon.
* All new attempts to start a merge or mutation will throw an exception until all 'LockHolder' objects will be destroyed.
*/
ActionBlocker merges_blocker;
ActionBlocker ttl_merges_blocker;
private:
MergeAlgorithm chooseMergeAlgorithm(
const MergeTreeData::DataPartsVector & parts,
size_t rows_upper_bound, const NamesAndTypesList & gathering_columns, bool deduplicate, bool need_remove_expired_values) const;
bool checkOperationIsNotCanceled(const MergeListEntry & merge_entry) const;
private:
MergeTreeData & data;
const size_t background_pool_size;
Poco::Logger * log;
/// When the last time you wrote to the log that the disk space was running out (not to write about this too often).
time_t disk_space_warning_time = 0;
/// Stores the next TTL delete merge due time for each partition (used only by TTLDeleteMergeSelector)
ITTLMergeSelector::PartitionIdToTTLs next_delete_ttl_merge_times_by_partition;
/// Stores the next TTL recompress merge due time for each partition (used only by TTLRecompressionMergeSelector)
ITTLMergeSelector::PartitionIdToTTLs next_recompress_ttl_merge_times_by_partition;
/// Performing TTL merges independently for each partition guarantees that
/// there is only a limited number of TTL merges and no partition stores data, that is too stale
};
}