Files
scylladb/counters.hh
Avi Kivity f698496ab2 Merge "Fix Scylla upgrades when counters are used" from Paweł
"Scylla 1.7.4 and older use incorrect ordering of counter shards, this
was fixed in 0d87f3dd7d ("utils::UUID:
operator< should behave as comparison of hex strings/bytes"). However,
that patch was not backported to 1.7 branch until very recently. This
means that versions 1.7.4 and older emit counter shards in an incorrect
order and expect them to be so. This is particularly bad when dealing
with imported correct sstables in which case some shards may become
duplicated.

The solution implemented in this patch is to allow any order of counter
shards and automaticly merge all duplicates. The code is written in a
way so that the correct ordering is expected in the fast path in order
not to excessively punish unaffected deployments.

A new feature flag CORRECT_COUNTER_ORDER is introduced to allow seamless
upgrade from 1.7.4 to later Scylla versions. If that feature is not
available Scylla still writes sstables and sends on-wire counters using
the old ordering so that it can be correctly understood by 1.7.4, once
the flag becomes available Scylla switches to the correct order.

Fixes #2752."

* tag 'fix-upgrade-with-counters/v2' of https://github.com/pdziepak/scylla:
  tests/counter: verify counter_id ordering
  counter: check that utils::UUID uses int64_t
  mutation_partition_serializer: use old counter ordering if necessary
  mutation_partition_view: do not expect counter shards to be sorted
  sstables: write counter shards in the order expected by the cluster
  tests/sstables: add storage_service_for_tests to counter write test
  tests/sstables: add test for reading wrong-order counter cells
  sstables: do not expect counter shards to be sorted
  storage_service: introduce CORRECT_COUNTER_ORDER feature
  tests/counter: test 1.7.4 compatible shard ordering
  counters: add helper for retrieving shards in 1.7.4 order
  tests/counter: add tests for 1.7.4 counter shard order
  counters: add counter id comparator compatible with Scylla 1.7.4
  tests/counter: verify order of counter shards
  tests/counter: add test for sorting and deduplicating shards
  counters: add function for sorting and deduplicating counter cells
  counters: add counter_id::operator>

(cherry picked from commit 31706ba989)
2017-09-05 14:25:36 +03:00

436 lines
14 KiB
C++

/*
* Copyright (C) 2016 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/>.
*/
#pragma once
#include <boost/range/algorithm/find_if.hpp>
#include "atomic_cell_or_collection.hh"
#include "types.hh"
#include "stdx.hh"
class mutation;
class mutation;
class counter_id {
int64_t _least_significant;
int64_t _most_significant;
public:
static_assert(std::is_same<decltype(std::declval<utils::UUID>().get_least_significant_bits()), int64_t>::value
&& std::is_same<decltype(std::declval<utils::UUID>().get_most_significant_bits()), int64_t>::value,
"utils::UUID is expected to work with two signed 64-bit integers");
counter_id() = default;
explicit counter_id(utils::UUID uuid) noexcept
: _least_significant(uuid.get_least_significant_bits())
, _most_significant(uuid.get_most_significant_bits())
{ }
utils::UUID to_uuid() const {
return utils::UUID(_most_significant, _least_significant);
}
bool operator<(const counter_id& other) const {
return to_uuid() < other.to_uuid();
}
bool operator>(const counter_id& other) const {
return other.to_uuid() < to_uuid();
}
bool operator==(const counter_id& other) const {
return to_uuid() == other.to_uuid();
}
bool operator!=(const counter_id& other) const {
return !(*this == other);
}
public:
// (Wrong) Counter ID ordering used by Scylla 1.7.4 and earlier.
struct less_compare_1_7_4 {
bool operator()(const counter_id& a, const counter_id& b) const;
};
public:
static counter_id local();
// For tests.
static counter_id generate_random() {
return counter_id(utils::make_random_uuid());
}
};
static_assert(std::is_pod<counter_id>::value, "counter_id should be a POD type");
std::ostream& operator<<(std::ostream& os, const counter_id& id);
template<typename View>
class basic_counter_shard_view {
enum class offset : unsigned {
id = 0u,
value = unsigned(id) + sizeof(counter_id),
logical_clock = unsigned(value) + sizeof(int64_t),
total_size = unsigned(logical_clock) + sizeof(int64_t),
};
private:
typename View::pointer _base;
private:
template<typename T>
T read(offset off) const {
T value;
std::copy_n(_base + static_cast<unsigned>(off), sizeof(T), reinterpret_cast<signed char*>(&value));
return value;
}
public:
static constexpr auto size = size_t(offset::total_size);
public:
basic_counter_shard_view() = default;
explicit basic_counter_shard_view(typename View::pointer ptr) noexcept
: _base(ptr) { }
counter_id id() const { return read<counter_id>(offset::id); }
int64_t value() const { return read<int64_t>(offset::value); }
int64_t logical_clock() const { return read<int64_t>(offset::logical_clock); }
void swap_value_and_clock(basic_counter_shard_view& other) noexcept {
static constexpr size_t off = size_t(offset::value);
static constexpr size_t size = size_t(offset::total_size) - off;
typename View::value_type tmp[size];
std::copy_n(_base + off, size, tmp);
std::copy_n(other._base + off, size, _base + off);
std::copy_n(tmp, size, other._base + off);
}
void set_value_and_clock(const basic_counter_shard_view& other) noexcept {
static constexpr size_t off = size_t(offset::value);
static constexpr size_t size = size_t(offset::total_size) - off;
std::copy_n(other._base + off, size, _base + off);
}
bool operator==(const basic_counter_shard_view& other) const {
return id() == other.id() && value() == other.value()
&& logical_clock() == other.logical_clock();
}
bool operator!=(const basic_counter_shard_view& other) const {
return !(*this == other);
}
struct less_compare_by_id {
bool operator()(const basic_counter_shard_view& x, const basic_counter_shard_view& y) const {
return x.id() < y.id();
}
};
};
using counter_shard_view = basic_counter_shard_view<bytes_view>;
std::ostream& operator<<(std::ostream& os, counter_shard_view csv);
class counter_shard {
counter_id _id;
int64_t _value;
int64_t _logical_clock;
private:
template<typename T>
static void write(const T& value, bytes::iterator& out) {
out = std::copy_n(reinterpret_cast<const signed char*>(&value), sizeof(T), out);
}
private:
// Shared logic for applying counter_shards and counter_shard_views.
// T is either counter_shard or basic_counter_shard_view<U>.
template<typename T>
GCC6_CONCEPT(requires requires(T shard) {
{ shard.value() } -> int64_t;
{ shard.logical_clock() } -> int64_t;
})
counter_shard& do_apply(T&& other) noexcept {
auto other_clock = other.logical_clock();
if (_logical_clock < other_clock) {
_logical_clock = other_clock;
_value = other.value();
}
return *this;
}
public:
counter_shard(counter_id id, int64_t value, int64_t logical_clock) noexcept
: _id(id)
, _value(value)
, _logical_clock(logical_clock)
{ }
explicit counter_shard(counter_shard_view csv) noexcept
: _id(csv.id())
, _value(csv.value())
, _logical_clock(csv.logical_clock())
{ }
counter_id id() const { return _id; }
int64_t value() const { return _value; }
int64_t logical_clock() const { return _logical_clock; }
counter_shard& update(int64_t value_delta, int64_t clock_increment) noexcept {
_value += value_delta;
_logical_clock += clock_increment;
return *this;
}
counter_shard& apply(counter_shard_view other) noexcept {
return do_apply(other);
}
counter_shard& apply(const counter_shard& other) noexcept {
return do_apply(other);
}
static size_t serialized_size() {
return counter_shard_view::size;
}
void serialize(bytes::iterator& out) const {
write(_id, out);
write(_value, out);
write(_logical_clock, out);
}
};
class counter_cell_builder {
std::vector<counter_shard> _shards;
bool _sorted = true;
private:
void do_sort_and_remove_duplicates();
public:
counter_cell_builder() = default;
counter_cell_builder(size_t shard_count) {
_shards.reserve(shard_count);
}
void add_shard(const counter_shard& cs) {
_shards.emplace_back(cs);
}
void add_maybe_unsorted_shard(const counter_shard& cs) {
add_shard(cs);
if (_sorted && _shards.size() > 1) {
auto current = _shards.rbegin();
auto previous = std::next(current);
_sorted = current->id() > previous->id();
}
}
void sort_and_remove_duplicates() {
if (!_sorted) {
do_sort_and_remove_duplicates();
}
}
size_t serialized_size() const {
return _shards.size() * counter_shard::serialized_size();
}
void serialize(bytes::iterator& out) const {
for (auto&& cs : _shards) {
cs.serialize(out);
}
}
bool empty() const {
return _shards.empty();
}
atomic_cell build(api::timestamp_type timestamp) const {
return atomic_cell::make_live_from_serializer(timestamp, serialized_size(), [this] (bytes::iterator out) {
serialize(out);
});
}
static atomic_cell from_single_shard(api::timestamp_type timestamp, const counter_shard& cs) {
return atomic_cell::make_live_from_serializer(timestamp, counter_shard::serialized_size(), [&cs] (bytes::iterator out) {
cs.serialize(out);
});
}
class inserter_iterator : public std::iterator<std::output_iterator_tag, counter_shard> {
counter_cell_builder* _builder;
public:
explicit inserter_iterator(counter_cell_builder& b) : _builder(&b) { }
inserter_iterator& operator=(const counter_shard& cs) {
_builder->add_shard(cs);
return *this;
}
inserter_iterator& operator=(const counter_shard_view& csv) {
return operator=(counter_shard(csv));
}
inserter_iterator& operator++() { return *this; }
inserter_iterator& operator++(int) { return *this; }
inserter_iterator& operator*() { return *this; };
};
inserter_iterator inserter() {
return inserter_iterator(*this);
}
};
// <counter_id> := <int64_t><int64_t>
// <shard> := <counter_id><int64_t:value><int64_t:logical_clock>
// <counter_cell> := <shard>*
template<typename View>
class basic_counter_cell_view {
protected:
atomic_cell_base<View> _cell;
private:
class shard_iterator : public std::iterator<std::input_iterator_tag, basic_counter_shard_view<View>> {
typename View::pointer _current;
basic_counter_shard_view<View> _current_view;
public:
shard_iterator() = default;
shard_iterator(typename View::pointer ptr) noexcept
: _current(ptr), _current_view(ptr) { }
basic_counter_shard_view<View>& operator*() noexcept {
return _current_view;
}
basic_counter_shard_view<View>* operator->() noexcept {
return &_current_view;
}
shard_iterator& operator++() noexcept {
_current += counter_shard_view::size;
_current_view = basic_counter_shard_view<View>(_current);
return *this;
}
shard_iterator operator++(int) noexcept {
auto it = *this;
operator++();
return it;
}
shard_iterator& operator--() noexcept {
_current -= counter_shard_view::size;
_current_view = basic_counter_shard_view<View>(_current);
return *this;
}
shard_iterator operator--(int) noexcept {
auto it = *this;
operator--();
return it;
}
bool operator==(const shard_iterator& other) const noexcept {
return _current == other._current;
}
bool operator!=(const shard_iterator& other) const noexcept {
return !(*this == other);
}
};
public:
boost::iterator_range<shard_iterator> shards() const {
auto bv = _cell.value();
auto begin = shard_iterator(bv.data());
auto end = shard_iterator(bv.data() + bv.size());
return boost::make_iterator_range(begin, end);
}
size_t shard_count() const {
return _cell.value().size() / counter_shard_view::size;
}
public:
// ac must be a live counter cell
explicit basic_counter_cell_view(atomic_cell_base<View> ac) noexcept : _cell(ac) {
assert(_cell.is_live());
assert(!_cell.is_counter_update());
}
api::timestamp_type timestamp() const { return _cell.timestamp(); }
static data_type total_value_type() { return long_type; }
int64_t total_value() const {
return boost::accumulate(shards(), int64_t(0), [] (int64_t v, counter_shard_view cs) {
return v + cs.value();
});
}
stdx::optional<counter_shard_view> get_shard(const counter_id& id) const {
auto it = boost::range::find_if(shards(), [&id] (counter_shard_view csv) {
return csv.id() == id;
});
if (it == shards().end()) {
return { };
}
return *it;
}
stdx::optional<counter_shard_view> local_shard() const {
// TODO: consider caching local shard position
return get_shard(counter_id::local());
}
bool operator==(const basic_counter_cell_view& other) const {
return timestamp() == other.timestamp() && boost::equal(shards(), other.shards());
}
};
struct counter_cell_view : basic_counter_cell_view<bytes_view> {
using basic_counter_cell_view::basic_counter_cell_view;
// Returns counter shards in an order that is compatible with Scylla 1.7.4.
std::vector<counter_shard> shards_compatible_with_1_7_4() const;
// Reversibly applies two counter cells, at least one of them must be live.
// Returns true iff dst was modified.
static bool apply_reversibly(atomic_cell_or_collection& dst, atomic_cell_or_collection& src);
// Reverts apply performed by apply_reversible().
static void revert_apply(atomic_cell_or_collection& dst, atomic_cell_or_collection& src);
// Computes a counter cell containing minimal amount of data which, when
// applied to 'b' returns the same cell as 'a' and 'b' applied together.
static stdx::optional<atomic_cell> difference(atomic_cell_view a, atomic_cell_view b);
friend std::ostream& operator<<(std::ostream& os, counter_cell_view ccv);
};
struct counter_cell_mutable_view : basic_counter_cell_view<bytes_mutable_view> {
using basic_counter_cell_view::basic_counter_cell_view;
void set_timestamp(api::timestamp_type ts) { _cell.set_timestamp(ts); }
};
// Transforms mutation dst from counter updates to counter shards using state
// stored in current_state.
// If current_state is present it has to be in the same schema as dst.
void transform_counter_updates_to_shards(mutation& dst, const mutation* current_state, uint64_t clock_offset);
template<>
struct appending_hash<counter_shard_view> {
template<typename Hasher>
void operator()(Hasher& h, const counter_shard_view& cshard) const {
::feed_hash(h, cshard.id().to_uuid());
::feed_hash(h, cshard.value());
::feed_hash(h, cshard.logical_clock());
}
};
template<>
struct appending_hash<counter_cell_view> {
template<typename Hasher>
void operator()(Hasher& h, const counter_cell_view& cell) const {
::feed_hash(h, true); // is_live
::feed_hash(h, cell.timestamp());
for (auto&& csv : cell.shards()) {
::feed_hash(h, csv);
}
}
};