Files
scylladb/sstables/sstable_set.hh
Raphael S. Carvalho f588b46f55 table: Optimize creation of reader excluding staging for view building
View building from staging creates a reader from scratch (memtable
+ sstables - staging) for every partition, in order to calculate
the diff between new staging data and data in base sstable set,
and then pushes the result into the view replicas.

perf shows that the reader creation is very expensive:
+   12.15%    10.75%  reactor-3        scylla             [.] lexicographical_tri_compare<compound_type<(allow_prefixes)0>::iterator, compound_type<(allow_prefixes)0>::iterator, legacy_compound_view<compound_type<(allow_prefixes)0> >::tri_comparator::operator()(managed_bytes_basic_view<(mutable_view)0>, managed_bytes
+   10.01%     9.99%  reactor-3        scylla             [.] boost::icl::is_empty<boost::icl::continuous_interval<compatible_ring_position_or_view, std::less> >
+    8.95%     8.94%  reactor-3        scylla             [.] legacy_compound_view<compound_type<(allow_prefixes)0> >::tri_comparator::operator()
+    7.29%     7.28%  reactor-3        scylla             [.] dht::ring_position_tri_compare
+    6.28%     6.27%  reactor-3        scylla             [.] dht::tri_compare
+    4.11%     3.52%  reactor-3        scylla             [.] boost::icl::interval_base_map<boost::icl::interval_map<compatible_ring_position_or_view, std::unordered_set<seastar::lw_shared_ptr<sstables::sstable>, std::hash<seastar::lw_shared_ptr<sstables::sstable> >, std::equal_to<seastar::lw_shared_ptr<sstables::sst+    4.09%     4.07%  reactor-3        scylla             [.] sstables::index_consume_entry_context<sstables::index_consumer>::process_state
+    3.46%     0.93%  reactor-3        scylla             [.] sstables::sstable_run::will_introduce_overlapping
+    2.53%     2.53%  reactor-3        libstdc++.so.6     [.] std::_Rb_tree_increment
+    2.45%     2.45%  reactor-3        scylla             [.] boost::icl::non_empty::exclusive_less<boost::icl::continuous_interval<compatible_ring_position_or_view, std::less> >
+    2.14%     2.13%  reactor-3        scylla             [.] boost::icl::exclusive_less<boost::icl::continuous_interval<compatible_ring_position_or_view, std::less> >
+    2.07%     2.07%  reactor-3        scylla             [.] logalloc::region_impl::free
+    2.06%     1.91%  reactor-3        scylla             [.] sstables::index_consumer::consume_entry(sstables::parsed_partition_index_entry&&)::{lambda()#1}::operator()() const::{lambda()#1}::operator()
+    2.04%     2.04%  reactor-3        scylla             [.] boost::icl::interval_base_map<boost::icl::interval_map<compatible_ring_position_or_view, std::unordered_set<seastar::lw_shared_ptr<sstables::sstable>, std::hash<seastar::lw_shared_ptr<sstables::sstable> >, std::equal_to<seastar::lw_shared_ptr<sstables::sst+    1.87%     0.00%  reactor-3        [kernel.kallsyms]  [k] entry_SYSCALL_64_after_hwframe
+    1.86%     0.00%  reactor-3        [kernel.kallsyms]  [k] do_syscall_64
+    1.39%     1.38%  reactor-3        libc.so.6          [.] __memcmp_avx2_movbe
+    1.37%     0.92%  reactor-3        scylla             [.] boost::icl::segmental::join_left<boost::icl::interval_map<compatible_ring_position_or_view, std::unordered_set<seastar::lw_shared_ptr<sstables::sstable>, std::hash<seastar::lw_shared_ptr<sstables::sstable> >, std::equal_to<seastar::lw_shared_ptr<sstables::
+    1.34%     1.33%  reactor-3        scylla             [.] logalloc::region_impl::alloc_small
+    1.33%     1.33%  reactor-3        scylla             [.] seastar::memory::small_pool::add_more_objects
+    1.30%     0.35%  reactor-3        scylla             [.] seastar::reactor::do_run
+    1.29%     1.29%  reactor-3        scylla             [.] seastar::memory::allocate
+    1.19%     0.05%  reactor-3        libc.so.6          [.] syscall
+    1.16%     1.04%  reactor-3        scylla             [.] boost::icl::interval_base_map<boost::icl::interval_map<compatible_ring_position_or_view, std::unordered_set<seastar::lw_shared_ptr<sstables::sstable>, std::hash<seastar::lw_shared_ptr<sstables::sstable> >, std::equal_to<seastar::lw_shared_ptr<sstables::sst
+    1.07%     0.79%  reactor-3        scylla             [.] sstables::partitioned_sstable_set::insert

That shows some significant amount of work for inserting sstables
into the interval map and maintaining the sstable run (which sorts
fragments by first key and checks for overlapping).

The interval map is known for having issues with L0 sstables, as
it will have to be replicated almost to every single interval
stored by the map, causing terrible space and time complexity.
With enough L0 sstables, it can fall into quadratic behavior.

This overhead is fixed by not building a new fresh sstable set
when recreating the reader, but rather supplying a predicate
to sstable set that will filter out staging sstables when
creating either a single-key or range scan reader.

This could have another benefit over today's approach which
may incorrectly consider a staging sstable as non-staging, if
the staging sst wasn't included in the current batch for view
building.

With this improvement, view building was measured to be 3x faster.

from
INFO  2023-06-16 12:36:40,014 [shard 0] view_update_generator - Processed keyspace1.standard1: 5 sstables in 963957ms = 50kB/s

to
INFO  2023-06-16 14:47:12,129 [shard 0] view_update_generator - Processed keyspace1.standard1: 5 sstables in 319899ms = 150kB/s

Refs #14089.
Fixes #14244.

Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>

Closes #14476
2023-07-06 10:32:31 +03:00

177 lines
7.0 KiB
C++

/*
* Copyright (C) 2016-present ScyllaDB
*/
/*
* SPDX-License-Identifier: AGPL-3.0-or-later
*/
#pragma once
#include "readers/flat_mutation_reader_fwd.hh"
#include "readers/flat_mutation_reader_v2.hh"
#include "sstables/progress_monitor.hh"
#include "shared_sstable.hh"
#include "dht/i_partitioner.hh"
#include <seastar/core/shared_ptr.hh>
#include <seastar/core/io_priority_class.hh>
#include <vector>
namespace utils {
class estimated_histogram;
}
namespace sstables {
using sstable_predicate = noncopyable_function<bool(const sstable&)>;
// Default predicate includes everything
const sstable_predicate& default_sstable_predicate();
class sstable_set_impl;
class incremental_selector_impl;
struct sstable_first_key_less_comparator {
bool operator()(const shared_sstable& s1, const shared_sstable& s2) const;
};
// Structure holds all sstables (a.k.a. fragments) that belong to same run identifier, which is an UUID.
// SStables in that same run will not overlap with one another.
class sstable_run {
public:
using sstable_set = std::set<shared_sstable, sstable_first_key_less_comparator>;
private:
sstable_set _all;
private:
bool will_introduce_overlapping(const shared_sstable& sst) const;
public:
// Returns false if sstable being inserted cannot satisfy the disjoint invariant. Then caller should pick another run for it.
bool insert(shared_sstable sst);
void erase(shared_sstable sst);
// Data size of the whole run, meaning it's a sum of the data size of all its fragments.
uint64_t data_size() const;
const sstable_set& all() const { return _all; }
double estimate_droppable_tombstone_ratio(gc_clock::time_point gc_before) const;
};
class sstable_set : public enable_lw_shared_from_this<sstable_set> {
std::unique_ptr<sstable_set_impl> _impl;
schema_ptr _schema;
public:
~sstable_set();
sstable_set(std::unique_ptr<sstable_set_impl> impl, schema_ptr s);
sstable_set(const sstable_set&);
sstable_set(sstable_set&&) noexcept;
sstable_set& operator=(const sstable_set&);
sstable_set& operator=(sstable_set&&) noexcept;
std::vector<shared_sstable> select(const dht::partition_range& range) const;
// Return all runs which contain any of the input sstables.
std::vector<sstable_run> select_sstable_runs(const std::vector<shared_sstable>& sstables) const;
// Return all sstables. It's not guaranteed that sstable_set will keep a reference to the returned list, so user should keep it.
lw_shared_ptr<sstable_list> all() const;
// Prefer for_each_sstable() over all() for iteration purposes, as the latter may have to copy all sstables into a temporary
void for_each_sstable(std::function<void(const shared_sstable&)> func) const;
void insert(shared_sstable sst);
void erase(shared_sstable sst);
// Used to incrementally select sstables from sstable set using ring-position.
// sstable set must be alive during the lifetime of the selector.
class incremental_selector {
std::unique_ptr<incremental_selector_impl> _impl;
dht::ring_position_comparator _cmp;
mutable std::optional<dht::partition_range> _current_range;
mutable std::optional<nonwrapping_range<dht::ring_position_view>> _current_range_view;
mutable std::vector<shared_sstable> _current_sstables;
mutable dht::ring_position_ext _current_next_position = dht::ring_position_view::min();
public:
~incremental_selector();
incremental_selector(std::unique_ptr<incremental_selector_impl> impl, const schema& s);
incremental_selector(incremental_selector&&) noexcept;
struct selection {
const std::vector<shared_sstable>& sstables;
dht::ring_position_view next_position;
};
// Return the sstables that intersect with `pos` and the next
// position where the intersecting sstables change.
// To walk through the token range incrementally call `select()`
// with `dht::ring_position_view::min()` and then pass back the
// returned `next_position` on each next call until
// `next_position` becomes `dht::ring_position::max()`.
//
// Successive calls to `select()' have to pass weakly monotonic
// positions (incrementability).
//
// NOTE: both `selection.sstables` and `selection.next_position`
// are only guaranteed to be valid until the next call to
// `select()`.
selection select(const dht::ring_position_view& pos) const;
};
incremental_selector make_incremental_selector() const;
flat_mutation_reader_v2 create_single_key_sstable_reader(
replica::column_family*,
schema_ptr,
reader_permit,
utils::estimated_histogram&,
const dht::partition_range&, // must be singular and contain a key
const query::partition_slice&,
const io_priority_class&,
tracing::trace_state_ptr,
streamed_mutation::forwarding,
mutation_reader::forwarding,
const sstable_predicate& p = default_sstable_predicate()) const;
/// Read a range from the sstable set.
///
/// The reader is unrestricted, but will account its resource usage on the
/// semaphore belonging to the passed-in permit.
flat_mutation_reader_v2 make_range_sstable_reader(
schema_ptr,
reader_permit,
const dht::partition_range&,
const query::partition_slice&,
const io_priority_class&,
tracing::trace_state_ptr,
streamed_mutation::forwarding,
mutation_reader::forwarding,
read_monitor_generator& rmg = default_read_monitor_generator()) const;
// Filters out mutations that don't belong to the current shard.
flat_mutation_reader_v2 make_local_shard_sstable_reader(
schema_ptr,
reader_permit,
const dht::partition_range&,
const query::partition_slice&,
const io_priority_class&,
tracing::trace_state_ptr,
streamed_mutation::forwarding,
mutation_reader::forwarding,
read_monitor_generator& rmg = default_read_monitor_generator(),
const sstable_predicate& p = default_sstable_predicate()) const;
flat_mutation_reader_v2 make_crawling_reader(
schema_ptr,
reader_permit,
const io_priority_class&,
tracing::trace_state_ptr,
read_monitor_generator& rmg = default_read_monitor_generator()) const;
friend class compound_sstable_set;
};
sstable_set make_partitioned_sstable_set(schema_ptr schema, bool use_level_metadata = true);
sstable_set make_compound_sstable_set(schema_ptr schema, std::vector<lw_shared_ptr<sstable_set>> sets);
std::ostream& operator<<(std::ostream& os, const sstables::sstable_run& run);
using offstrategy = bool_class<class offstrategy_tag>;
/// Return the amount of overlapping in a set of sstables. 0 is returned if set is disjoint.
///
/// The 'sstables' parameter must be a set of sstables sorted by first key.
unsigned sstable_set_overlapping_count(const schema_ptr& schema, const std::vector<shared_sstable>& sstables);
}