Files
scylladb/locator/tablets.hh
Tomasz Grabiec ded9aca6ee repair: Exclude tablet migrations with tablet repair
We want to exclude repair with tablet migrations to avoid races
between repair reads and writes with replica movement. Repair is not
prepared to handle topology transitions in the middle.

One reason why it's not safe is that repair may successfully write to
a leaving replica post streaming phase and consider all replicas to be
repaired, but in fact they are not, the new replica would not be
repaired.

Other kinds of races could result in repair failures. If repair writes
to a leaving replica which was already cleaned up, such writes will
fail, causing repair to fail.

Excluding works by keeping effective_replication_map_ptr in a version
which doesn't have table's tablets in transitions. That prevents later
transitions from starting because topology coordinator's barrier will
wait for that erm before moving to a stage later than
allow_write_both_read_old, so before any requets start using the new
topology. Also, if transitions are already running, repair waits for
them to finish.

Fixes #17658.
Fixes #18561.

(cherry picked from commit 98323be296)
2024-06-08 16:31:18 +02:00

562 lines
20 KiB
C++

/*
* Copyright (C) 2023-present ScyllaDB
*/
/*
* SPDX-License-Identifier: AGPL-3.0-or-later
*/
#pragma once
#include "dht/token.hh"
#include "utils/small_vector.hh"
#include "locator/host_id.hh"
#include "service/session.hh"
#include "dht/i_partitioner_fwd.hh"
#include "dht/token-sharding.hh"
#include "dht/ring_position.hh"
#include "schema/schema_fwd.hh"
#include "utils/chunked_vector.hh"
#include "utils/hash.hh"
#include <boost/range/adaptor/transformed.hpp>
#include <seastar/core/reactor.hh>
#include <seastar/util/log.hh>
#include <seastar/core/coroutine.hh>
#include <seastar/util/noncopyable_function.hh>
#include <seastar/coroutine/maybe_yield.hh>
namespace locator {
class topology;
extern seastar::logger tablet_logger;
using token = dht::token;
// Identifies tablet within the scope of a single tablet_map,
// which has a scope of (table_id, token metadata version).
// Different tablets of different tables can have the same tablet_id.
// Different tablets in subsequent token metadata version can have the same tablet_id.
// When splitting a tablet, one of the new tablets (in the new token metadata version)
// will have the same tablet_id as the old one.
struct tablet_id {
size_t id;
explicit tablet_id(size_t id) : id(id) {}
size_t value() const { return id; }
explicit operator size_t() const { return id; }
bool operator<=>(const tablet_id&) const = default;
};
/// Identifies tablet (not be confused with tablet replica) in the scope of the whole cluster.
struct global_tablet_id {
table_id table;
tablet_id tablet;
bool operator<=>(const global_tablet_id&) const = default;
};
struct tablet_replica {
host_id host;
shard_id shard;
bool operator==(const tablet_replica&) const = default;
};
using tablet_replica_set = utils::small_vector<tablet_replica, 3>;
}
namespace std {
template<>
struct hash<locator::tablet_id> {
size_t operator()(const locator::tablet_id& id) const {
return std::hash<size_t>()(id.value());
}
};
template<>
struct hash<locator::tablet_replica> {
size_t operator()(const locator::tablet_replica& r) const {
return utils::hash_combine(
std::hash<locator::host_id>()(r.host),
std::hash<shard_id>()(r.shard));
}
};
template<>
struct hash<locator::global_tablet_id> {
size_t operator()(const locator::global_tablet_id& id) const {
return utils::hash_combine(
std::hash<table_id>()(id.table),
std::hash<locator::tablet_id>()(id.tablet));
}
};
}
namespace locator {
/// Creates a new replica set with old_replica replaced by new_replica.
/// If there is no old_replica, the set is returned unchanged.
inline
tablet_replica_set replace_replica(const tablet_replica_set& rs, tablet_replica old_replica, tablet_replica new_replica) {
tablet_replica_set result;
result.reserve(rs.size());
for (auto&& r : rs) {
if (r == old_replica) {
result.push_back(new_replica);
} else {
result.push_back(r);
}
}
return result;
}
/// Subtracts 'sub' from 'rs' and returns the result
/// Replicas from 'sub' that are missing in 'rs' are ignored
inline
std::unordered_set<tablet_replica> substract_sets(const tablet_replica_set& rs, const tablet_replica_set& sub) {
std::unordered_set<tablet_replica> result(rs.begin(), rs.end());
for (auto&& r : sub) {
result.erase(r);
}
return result;
}
inline
bool contains(const tablet_replica_set& rs, host_id host) {
for (auto replica : rs) {
if (replica.host == host) {
return true;
}
}
return false;
}
inline
bool contains(const tablet_replica_set& rs, const tablet_replica& r) {
return std::ranges::any_of(rs, [&] (auto&& r_) { return r_ == r; });
}
/// Stores information about a single tablet.
struct tablet_info {
tablet_replica_set replicas;
bool operator==(const tablet_info&) const = default;
};
/// Represents states of the tablet migration state machine.
///
/// The stage serves two major purposes:
///
/// Firstly, it determines which action should be taken by the topology change coordinator on behalf
/// of the tablet before it can move to the next step. When stage is advanced, it means that
/// expected invariants about cluster-wide state relevant to the tablet, associated with the new stage, hold.
///
/// Also, stage affects which replicas are used by the coordinator for reads and writes.
/// Replica selectors kept in tablet_transition_info::writes and tablet_transition_info::reads,
/// are directly derived from the stage stored in group0.
///
/// See "Tablet migration" in docs/dev/topology-over-raft.md
enum class tablet_transition_stage {
allow_write_both_read_old,
write_both_read_old,
streaming,
write_both_read_new,
use_new,
cleanup,
cleanup_target,
revert_migration,
end_migration,
};
enum class tablet_transition_kind {
// Tablet replica is migrating from one shard to another.
// The new replica is (tablet_transition_info::next - tablet_info::replicas).
// The leaving replica is (tablet_info::replicas - tablet_transition_info::next).
migration,
// Like migration, but the new pending replica is on the same host as leaving replica.
intranode_migration,
// New tablet replica is replacing a dead one.
// The new replica is (tablet_transition_info::next - tablet_info::replicas).
// The leaving replica is (tablet_info::replicas - tablet_transition_info::next).
rebuild,
};
sstring tablet_transition_stage_to_string(tablet_transition_stage);
tablet_transition_stage tablet_transition_stage_from_string(const sstring&);
sstring tablet_transition_kind_to_string(tablet_transition_kind);
tablet_transition_kind tablet_transition_kind_from_string(const sstring&);
using write_replica_set_selector = dht::write_replica_set_selector;
enum class read_replica_set_selector {
previous, next
};
/// Used for storing tablet state transition during topology changes.
/// Describes transition of a single tablet.
struct tablet_transition_info {
tablet_transition_stage stage;
tablet_transition_kind transition;
tablet_replica_set next;
std::optional<tablet_replica> pending_replica; // Optimization (next - tablet_info::replicas)
service::session_id session_id;
write_replica_set_selector writes;
read_replica_set_selector reads;
tablet_transition_info(tablet_transition_stage stage,
tablet_transition_kind kind,
tablet_replica_set next,
std::optional<tablet_replica> pending_replica,
service::session_id session_id = {});
bool operator==(const tablet_transition_info&) const = default;
};
// Returns the leaving replica for a given transition.
std::optional<tablet_replica> get_leaving_replica(const tablet_info&, const tablet_transition_info&);
/// Represents intention to move a single tablet replica from src to dst.
struct tablet_migration_info {
locator::tablet_transition_kind kind;
locator::global_tablet_id tablet;
locator::tablet_replica src;
locator::tablet_replica dst;
};
/// Returns the replica set which will become the replica set of the tablet after executing a given tablet transition.
tablet_replica_set get_new_replicas(const tablet_info&, const tablet_migration_info&);
tablet_replica_set get_primary_replicas(const tablet_info&, const tablet_transition_info*);
tablet_transition_info migration_to_transition_info(const tablet_info&, const tablet_migration_info&);
/// Describes streaming required for a given tablet transition.
struct tablet_migration_streaming_info {
std::unordered_set<tablet_replica> read_from;
std::unordered_set<tablet_replica> written_to;
};
tablet_migration_streaming_info get_migration_streaming_info(const locator::topology&, const tablet_info&, const tablet_transition_info&);
tablet_migration_streaming_info get_migration_streaming_info(const locator::topology&, const tablet_info&, const tablet_migration_info&);
// Describes if a given token is located at either left or right side of a tablet's range
enum tablet_range_side {
left = 0,
right = 1,
};
// The decision of whether tablets of a given should be split, merged, or none, is made
// by the load balancer. This decision is recorded in the tablet_map and stored in group0.
struct resize_decision {
struct none {};
struct split {};
struct merge {};
using seq_number_t = int64_t;
std::variant<none, split, merge> way;
// The sequence number globally identifies a resize decision.
// It's monotonically increasing, globally.
// Needed to distinguish stale decision from latest one, in case coordinator
// revokes the current decision and signal it again later.
seq_number_t sequence_number = 0;
resize_decision() = default;
resize_decision(sstring decision, uint64_t seq_number);
bool split_or_merge() const {
return !std::holds_alternative<resize_decision::none>(way);
}
bool operator==(const resize_decision&) const;
sstring type_name() const;
seq_number_t next_sequence_number() const;
};
struct table_load_stats {
uint64_t size_in_bytes = 0;
// Stores the minimum seq number among all replicas, as coordinator wants to know if
// all replicas have completed splitting, which happens when they all store the
// seq number of the current split decision.
resize_decision::seq_number_t split_ready_seq_number = std::numeric_limits<resize_decision::seq_number_t>::max();
table_load_stats& operator+=(const table_load_stats& s) noexcept;
friend table_load_stats operator+(table_load_stats a, const table_load_stats& b) {
return a += b;
}
};
struct load_stats {
std::unordered_map<table_id, table_load_stats> tables;
load_stats& operator+=(const load_stats& s);
friend load_stats operator+(load_stats a, const load_stats& b) {
return a += b;
}
};
using load_stats_ptr = lw_shared_ptr<const load_stats>;
/// Stores information about tablets of a single table.
///
/// The map contains a constant number of tablets, tablet_count().
/// Each tablet has an associated tablet_info, and an optional tablet_transition_info.
/// Any given token is owned by exactly one tablet in this map.
///
/// A tablet map describes the whole ring, it cannot contain a partial mapping.
/// This means that the following sequence is always valid:
///
/// tablet_map& tmap = ...;
/// dht::token t = ...;
/// tablet_id id = tmap.get_tablet_id(t);
/// tablet_info& info = tmap.get_tablet_info(id);
///
/// A tablet_id obtained from an instance of tablet_map is valid for that instance only.
class tablet_map {
public:
using tablet_container = utils::chunked_vector<tablet_info>;
private:
// The implementation assumes that _tablets.size() is a power of 2:
//
// _tablets.size() == 1 << _log2_tablets
//
tablet_container _tablets;
size_t _log2_tablets; // log_2(_tablets.size())
std::unordered_map<tablet_id, tablet_transition_info> _transitions;
resize_decision _resize_decision;
public:
/// Constructs a tablet map.
///
/// \param tablet_count The desired tablets to allocate. Must be a power of two.
explicit tablet_map(size_t tablet_count);
/// Returns tablet_id of a tablet which owns a given token.
tablet_id get_tablet_id(token) const;
// Returns tablet_id and also the side of the tablet's range that a given token belongs to.
std::pair<tablet_id, tablet_range_side> get_tablet_id_and_range_side(token) const;
/// Returns tablet_info associated with a given tablet.
/// The given id must belong to this instance.
const tablet_info& get_tablet_info(tablet_id) const;
/// Returns a pointer to tablet_transition_info associated with a given tablet.
/// If there is no transition for a given tablet, returns nullptr.
/// \throws std::logic_error If the given id does not belong to this instance.
const tablet_transition_info* get_tablet_transition_info(tablet_id) const;
/// Returns the largest token owned by a given tablet.
/// \throws std::logic_error If the given id does not belong to this instance.
dht::token get_last_token(tablet_id id) const;
/// Returns the smallest token owned by a given tablet.
/// \throws std::logic_error If the given id does not belong to this instance.
dht::token get_first_token(tablet_id id) const;
/// Returns token_range which contains all tokens owned by a given tablet and only such tokens.
/// \throws std::logic_error If the given id does not belong to this instance.
dht::token_range get_token_range(tablet_id id) const;
/// Returns the primary replica for the tablet
tablet_replica get_primary_replica(tablet_id id) const;
tablet_replica get_primary_replica_within_dc(tablet_id id, const topology& topo, sstring dc) const;
/// Returns a vector of sorted last tokens for tablets.
future<std::vector<token>> get_sorted_tokens() const;
/// Returns the id of the first tablet.
tablet_id first_tablet() const {
return tablet_id(0);
}
/// Returns the id of the last tablet.
tablet_id last_tablet() const {
return tablet_id(tablet_count() - 1);
}
/// Returns the id of a tablet which follows a given tablet in the ring,
/// or disengaged optional if the given tablet is the last one.
std::optional<tablet_id> next_tablet(tablet_id t) const {
if (t == last_tablet()) {
return std::nullopt;
}
return tablet_id(size_t(t) + 1);
}
/// Returns true iff tablet has a given replica.
/// If tablet is in transition, considers both previous and next replica set.
bool has_replica(tablet_id, tablet_replica) const;
const tablet_container& tablets() const {
return _tablets;
}
/// Calls a given function for each tablet in the map in token ownership order.
future<> for_each_tablet(seastar::noncopyable_function<future<>(tablet_id, const tablet_info&)> func) const;
const auto& transitions() const {
return _transitions;
}
bool has_transitions() const {
return !_transitions.empty();
}
/// Returns an iterable range over tablet_id:s which includes all tablets in token ring order.
auto tablet_ids() const {
return boost::irange<size_t>(0, tablet_count()) | boost::adaptors::transformed([] (size_t i) {
return tablet_id(i);
});
}
size_t tablet_count() const {
return _tablets.size();
}
/// Returns tablet_info associated with the tablet which owns a given token.
const tablet_info& get_tablet_info(token t) const {
return get_tablet_info(get_tablet_id(t));
}
size_t external_memory_usage() const;
bool operator==(const tablet_map&) const = default;
bool needs_split() const;
const locator::resize_decision& resize_decision() const;
public:
void set_tablet(tablet_id, tablet_info);
void set_tablet_transition_info(tablet_id, tablet_transition_info);
void set_resize_decision(locator::resize_decision);
void clear_transitions();
// Destroys gently.
// The tablet map is not usable after this call and should be destroyed.
future<> clear_gently();
friend fmt::formatter<tablet_map>;
private:
void check_tablet_id(tablet_id) const;
};
/// Holds information about all tablets in the cluster.
///
/// When this instance is obtained via token_metadata_ptr, it is immutable
/// (represents a snapshot) and references obtained through this are guaranteed
/// to remain valid as long as the containing token_metadata_ptr is held.
///
/// Copy constructor can be invoked across shards.
class tablet_metadata {
public:
// FIXME: Make cheap to copy.
// We want both immutability and cheap updates, so we should use
// hierarchical data structure with shared pointers and copy-on-write.
// Currently we have immutability but updates require full copy.
//
// Also, currently the copy constructor is invoked across shards, which precludes
// using shared pointers. We should change that and use a foreign_ptr<> to
// hold immutable tablet_metadata which lives on shard 0 only.
// See storage_service::replicate_to_all_cores().
using table_to_tablet_map = std::unordered_map<table_id, tablet_map>;
private:
table_to_tablet_map _tablets;
// When false, tablet load balancer will not try to rebalance tablets.
bool _balancing_enabled = true;
public:
bool balancing_enabled() const { return _balancing_enabled; }
const tablet_map& get_tablet_map(table_id id) const;
const table_to_tablet_map& all_tables() const { return _tablets; }
table_to_tablet_map& all_tables() { return _tablets; }
size_t external_memory_usage() const;
bool has_replica_on(host_id) const;
public:
void set_balancing_enabled(bool value) { _balancing_enabled = value; }
void set_tablet_map(table_id, tablet_map);
tablet_map& get_tablet_map(table_id id);
future<> clear_gently();
public:
bool operator==(const tablet_metadata&) const = default;
friend fmt::formatter<tablet_metadata>;
};
// Check that all tablets which have replicas on this host, have a valid replica shard (< smp::count).
future<bool> check_tablet_replica_shards(const tablet_metadata& tm, host_id this_host);
struct tablet_routing_info {
tablet_replica_set tablet_replicas;
std::pair<dht::token, dht::token> token_range;
};
/// Split a list of ranges, such that conceptually each input range is
/// intersected with each tablet range.
/// Tablets are pre-filtered, selecting only tablets that have a replica on the
/// given host.
/// Return the resulting intersections, in order.
/// The ranges are generated lazily (one at a time).
///
/// Note: the caller is expected to pin tablets, by keeping an
/// effective-replication-map alive.
class tablet_range_splitter {
public:
struct range_split_result {
shard_id shard; // shard where the tablet owning this range lives
dht::partition_range range;
};
private:
schema_ptr _schema;
const dht::partition_range_vector& _ranges;
dht::partition_range_vector::const_iterator _ranges_it;
std::vector<range_split_result> _tablet_ranges;
std::vector<range_split_result>::iterator _tablet_ranges_it;
public:
tablet_range_splitter(schema_ptr schema, const tablet_map& tablets, host_id host, const dht::partition_range_vector& ranges);
/// Returns nullopt when there are no more ranges.
std::optional<range_split_result> operator()();
};
}
template <>
struct fmt::formatter<locator::tablet_transition_stage> : fmt::formatter<string_view> {
auto format(const locator::tablet_transition_stage&, fmt::format_context& ctx) const -> decltype(ctx.out());
};
template <>
struct fmt::formatter<locator::tablet_transition_kind> : fmt::formatter<string_view> {
auto format(const locator::tablet_transition_kind&, fmt::format_context& ctx) const -> decltype(ctx.out());
};
template <>
struct fmt::formatter<locator::global_tablet_id> : fmt::formatter<string_view> {
auto format(const locator::global_tablet_id&, fmt::format_context& ctx) const -> decltype(ctx.out());
};
template <>
struct fmt::formatter<locator::tablet_id> : fmt::formatter<string_view> {
auto format(locator::tablet_id id, fmt::format_context& ctx) const {
return fmt::format_to(ctx.out(), "{}", id.value());
}
};
template <>
struct fmt::formatter<locator::tablet_replica> : fmt::formatter<string_view> {
auto format(const locator::tablet_replica& r, fmt::format_context& ctx) const {
return fmt::format_to(ctx.out(), "{}:{}", r.host, r.shard);
}
};
template <>
struct fmt::formatter<locator::tablet_map> : fmt::formatter<string_view> {
auto format(const locator::tablet_map&, fmt::format_context& ctx) const -> decltype(ctx.out());
};
template <>
struct fmt::formatter<locator::tablet_metadata> : fmt::formatter<string_view> {
auto format(const locator::tablet_metadata&, fmt::format_context& ctx) const -> decltype(ctx.out());
};