mirrors/scylladb
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
82394debe68f450dffd19187a99655f4e433044b
scylladb/tests/memtable_test.cc
Glauber Costa 80440c0d79 database: rework dirty memory hierarchy 
Issue #1918 describes a problem, in which we are generating smaller
memtables than we could, and therefore not respecting the flush
criteria.

That happens because group sizes (and limits) for pressure purposes, and
the the soft threshold is currently at 40 %. This causes system group's
soft threshold to be way below regular's virtual dirty limit and close
to regular group's soft threshold. The system group was very likely to
become under soft pressure when regular was because writes to regular
group are not yet throttled when they cross both soft thresholds.

This is a direct consequence of the linear hierarchy between the regions
and to guarantee that it won't happen we would have acqire the semaphore
of all ancestor regions when flushing from a child region. While that
works, it can lead to problems on its own, like priority inversion if
the regions have different priorities - like streaming and regular, and
groups lower in the hierarchy, like user, blocking explicit flushes
from their ancestors

To fix that, this patch reorganizes the dirty memory region groups so
that groups are now completely independent. As a disadvantage, when
streaming happen we will draw some memory from the cache, but we will
live with it for the time being.

Fixes #1918

Signed-off-by: Glauber Costa <glauber@scylladb.com>
2016-12-13 14:07:53 -05:00

321 lines
12 KiB
C++
Raw Blame History

/*
* Copyright (C) 2015 ScyllaDB
*/
/*
* This file is part of Scylla.
*
* Scylla is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* Scylla is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with Scylla. If not, see <http://www.gnu.org/licenses/>.
*/
#include <boost/test/unit_test.hpp>
#include "service/priority_manager.hh"
#include "database.hh"
#include "utils/UUID_gen.hh"
#include "tests/test-utils.hh"
#include "schema_builder.hh"
#include "core/thread.hh"
#include "memtable.hh"
#include "mutation_source_test.hh"
#include "mutation_reader_assertions.hh"
#include "mutation_assertions.hh"
#include "disk-error-handler.hh"
thread_local disk_error_signal_type commit_error;
thread_local disk_error_signal_type general_disk_error;
static api::timestamp_type next_timestamp() {
static thread_local api::timestamp_type next_timestamp = 1;
return next_timestamp++;
}
static bytes make_unique_bytes() {
return to_bytes(utils::UUID_gen::get_time_UUID().to_sstring());
}
static void set_column(mutation& m, const sstring& column_name) {
assert(m.schema()->get_column_definition(to_bytes(column_name))->type == bytes_type);
auto value = data_value(make_unique_bytes());
m.set_clustered_cell(clustering_key::make_empty(), to_bytes(column_name), value, next_timestamp());
}
static
mutation make_unique_mutation(schema_ptr s) {
return mutation(partition_key::from_single_value(*s, make_unique_bytes()), s);
}
// Returns a vector of empty mutations in ring order
std::vector<mutation> make_ring(schema_ptr s, int n_mutations) {
std::vector<mutation> ring;
for (int i = 0; i < n_mutations; ++i) {
ring.push_back(make_unique_mutation(s));
}
std::sort(ring.begin(), ring.end(), mutation_decorated_key_less_comparator());
return ring;
}
SEASTAR_TEST_CASE(test_memtable_conforms_to_mutation_source) {
return seastar::async([] {
run_mutation_source_tests([](schema_ptr s, const std::vector<mutation>& partitions) {
auto mt = make_lw_shared<memtable>(s);
for (auto&& m : partitions) {
mt->apply(m);
}
logalloc::shard_tracker().full_compaction();
return mt->as_data_source();
});
});
}
SEASTAR_TEST_CASE(test_adding_a_column_during_reading_doesnt_affect_read_result) {
return seastar::async([] {
auto common_builder = schema_builder("ks", "cf")
.with_column("pk", bytes_type, column_kind::partition_key);
auto s1 = common_builder
.with_column("v2", bytes_type, column_kind::regular_column)
.build();
auto s2 = common_builder
.with_column("v1", bytes_type, column_kind::regular_column) // new column
.with_column("v2", bytes_type, column_kind::regular_column)
.build();
auto mt = make_lw_shared<memtable>(s1);
std::vector<mutation> ring = make_ring(s1, 3);
for (auto&& m : ring) {
set_column(m, "v2");
mt->apply(m);
}
auto check_rd_s1 = assert_that(mt->make_reader(s1));
auto check_rd_s2 = assert_that(mt->make_reader(s2));
check_rd_s1.next_mutation().has_schema(s1).is_equal_to(ring[0]);
check_rd_s2.next_mutation().has_schema(s2).is_equal_to(ring[0]);
mt->set_schema(s2);
check_rd_s1.next_mutation().has_schema(s1).is_equal_to(ring[1]);
check_rd_s2.next_mutation().has_schema(s2).is_equal_to(ring[1]);
check_rd_s1.next_mutation().has_schema(s1).is_equal_to(ring[2]);
check_rd_s2.next_mutation().has_schema(s2).is_equal_to(ring[2]);
check_rd_s1.produces_end_of_stream();
check_rd_s2.produces_end_of_stream();
assert_that(mt->make_reader(s1))
.produces(ring[0])
.produces(ring[1])
.produces(ring[2])
.produces_end_of_stream();
assert_that(mt->make_reader(s2))
.produces(ring[0])
.produces(ring[1])
.produces(ring[2])
.produces_end_of_stream();
});
}
SEASTAR_TEST_CASE(test_virtual_dirty_accounting_on_flush) {
return seastar::async([] {
schema_ptr s = schema_builder("ks", "cf")
.with_column("pk", bytes_type, column_kind::partition_key)
.with_column("col", bytes_type, column_kind::regular_column)
.build();
dirty_memory_manager mgr;
auto mt = make_lw_shared<memtable>(s, mgr);
std::vector<mutation> ring = make_ring(s, 3);
std::vector<mutation> current_ring;
for (auto&& m : ring) {
auto m_with_cell = m;
m_with_cell.set_clustered_cell(clustering_key::make_empty(), to_bytes("col"),
data_value(bytes(bytes::initialized_later(), 4096)), next_timestamp());
mt->apply(m_with_cell);
current_ring.push_back(m_with_cell);
}
// Create a reader which will cause many partition versions to be created
auto rd1 = mt->make_reader(s);
streamed_mutation_opt part0_stream = rd1().get0();
// Override large cell value with a short one
{
auto part0_update = ring[0];
part0_update.set_clustered_cell(clustering_key::make_empty(), to_bytes("col"),
data_value(bytes(bytes::initialized_later(), 8)), next_timestamp());
mt->apply(std::move(part0_update));
current_ring[0] = part0_update;
}
std::vector<size_t> virtual_dirty_values;
virtual_dirty_values.push_back(mgr.virtual_dirty_memory());
auto flush_reader_check = assert_that(mt->make_flush_reader(s, service::get_local_priority_manager().memtable_flush_priority()));
flush_reader_check.produces(current_ring[0]);
virtual_dirty_values.push_back(mgr.virtual_dirty_memory());
flush_reader_check.produces(current_ring[1]);
virtual_dirty_values.push_back(mgr.virtual_dirty_memory());
part0_stream = {};
while (rd1().get0()) ;
logalloc::shard_tracker().full_compaction();
flush_reader_check.produces(current_ring[2]);
virtual_dirty_values.push_back(mgr.virtual_dirty_memory());
flush_reader_check.produces_end_of_stream();
virtual_dirty_values.push_back(mgr.virtual_dirty_memory());
std::reverse(virtual_dirty_values.begin(), virtual_dirty_values.end());
BOOST_REQUIRE(std::is_sorted(virtual_dirty_values.begin(), virtual_dirty_values.end()));
});
}
// Reproducer for #1753
SEASTAR_TEST_CASE(test_partition_version_consistency_after_lsa_compaction_happens) {
return seastar::async([] {
schema_ptr s = schema_builder("ks", "cf")
.with_column("pk", bytes_type, column_kind::partition_key)
.with_column("ck", bytes_type, column_kind::clustering_key)
.with_column("col", bytes_type, column_kind::regular_column)
.build();
auto mt = make_lw_shared<memtable>(s);
auto empty_m = make_unique_mutation(s);
auto ck1 = clustering_key::from_single_value(*s, data_value(make_unique_bytes()).serialize());
auto ck2 = clustering_key::from_single_value(*s, data_value(make_unique_bytes()).serialize());
auto ck3 = clustering_key::from_single_value(*s, data_value(make_unique_bytes()).serialize());
auto m1 = empty_m;
m1.set_clustered_cell(ck1, to_bytes("col"), data_value(bytes(bytes::initialized_later(), 8)), next_timestamp());
auto m2 = empty_m;
m2.set_clustered_cell(ck2, to_bytes("col"), data_value(bytes(bytes::initialized_later(), 8)), next_timestamp());
auto m3 = empty_m;
m3.set_clustered_cell(ck3, to_bytes("col"), data_value(bytes(bytes::initialized_later(), 8)), next_timestamp());
mt->apply(m1);
auto rd1 = mt->make_reader(s);
streamed_mutation_opt stream1 = rd1().get0();
mt->apply(m2);
auto rd2 = mt->make_reader(s);
streamed_mutation_opt stream2 = rd2().get0();
mt->apply(m3);
auto rd3 = mt->make_reader(s);
streamed_mutation_opt stream3 = rd3().get0();
logalloc::shard_tracker().full_compaction();
auto rd4 = mt->make_reader(s);
streamed_mutation_opt stream4 = rd4().get0();
auto rd5 = mt->make_reader(s);
streamed_mutation_opt stream5 = rd5().get0();
auto rd6 = mt->make_reader(s);
streamed_mutation_opt stream6 = rd6().get0();
assert_that(mutation_from_streamed_mutation(std::move(stream1)).get0()).has_mutation().is_equal_to(m1);
assert_that(mutation_from_streamed_mutation(std::move(stream2)).get0()).has_mutation().is_equal_to(m1 + m2);
assert_that(mutation_from_streamed_mutation(std::move(stream3)).get0()).has_mutation().is_equal_to(m1 + m2 + m3);
rd3 = {};
assert_that(mutation_from_streamed_mutation(std::move(stream4)).get0()).has_mutation().is_equal_to(m1 + m2 + m3);
rd1 = {};
assert_that(mutation_from_streamed_mutation(std::move(stream5)).get0()).has_mutation().is_equal_to(m1 + m2 + m3);
rd2 = {};
assert_that(mutation_from_streamed_mutation(std::move(stream6)).get0()).has_mutation().is_equal_to(m1 + m2 + m3);
});
}
struct function_invoking_consumer {
std::function<void()> func;
template<typename T>
stop_iteration consume(T t) {
func();
return stop_iteration::no;
}
void consume_end_of_stream() { }
};
// Reproducer for #1746
SEASTAR_TEST_CASE(test_segment_migration_during_flush) {
return seastar::async([] {
schema_ptr s = schema_builder("ks", "cf")
.with_column("pk", bytes_type, column_kind::partition_key)
.with_column("ck", bytes_type, column_kind::clustering_key)
.with_column("col", bytes_type, column_kind::regular_column)
.build();
dirty_memory_manager mgr;
auto mt = make_lw_shared<memtable>(s, mgr);
const int rows_per_partition = 300;
const int partitions = 3;
std::vector<mutation> ring = make_ring(s, partitions);
for (auto& m : ring) {
for (int i = 0; i < rows_per_partition; ++i) {
auto ck = clustering_key::from_single_value(*s, data_value(make_unique_bytes()).serialize());
auto col_value = data_value(bytes(bytes::initialized_later(), 8));
m.set_clustered_cell(ck, to_bytes("col"), col_value, next_timestamp());
}
mt->apply(m);
}
std::vector<size_t> virtual_dirty_values;
virtual_dirty_values.push_back(mgr.virtual_dirty_memory());
auto rd = mt->make_flush_reader(s, service::get_local_priority_manager().memtable_flush_priority());
auto consume_mutation = [] (streamed_mutation_opt part) {
assert(part);
consume(*part, function_invoking_consumer{[] {
logalloc::shard_tracker().full_compaction();
}}).get();
};
for (int i = 0; i < partitions; ++i) {
consume_mutation(rd().get0());
virtual_dirty_values.push_back(mgr.virtual_dirty_memory());
}
BOOST_REQUIRE(!rd().get0());
std::reverse(virtual_dirty_values.begin(), virtual_dirty_values.end());
BOOST_REQUIRE(std::is_sorted(virtual_dirty_values.begin(), virtual_dirty_values.end()));
});
}
Reference in New Issue View Git Blame Copy Permalink