mirrors/scylladb
1
0
Fork 0
mirror of https://github.com/scylladb/scylladb.git synced 2026-05-31 03:56:42 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
c0ea98f922b2d6fa5aa95e27052d0579d7e1dafa
scylladb/transport/server.hh
Wojciech Mitros 13c043903d strong_consistency: cache leader location for non-replica nodes 
When a non-replica node handles a strongly consistent write, it must
forward the request to a replica. If the closest replica is not the
leader, the request gets redirected again, causing an extra roundtrip.

Add a leader location cache in groups_manager, keyed by raft group_id.
After a write request is forwarded, the CQL transport layer records the
final node as the leader in the cache. Subsequent write requests from
the same node for the same group are forwarded directly to the cached
leader, eliminating the extra roundtrip.

The cache is only used for writes. Reads can be served by any replica,
so they skip the cache and use proximity-based routing instead.

Cache entries are validated at use time: if the cached leader is no
longer a replica (e.g. after tablet migration), the entry is evicted
and the normal closest-replica path is taken. This prevents a scenario
where two nodes keep redirecting to each other because both think that
the other is the leader but actually both are non-replicas - such loop
is broken as soon as the tablet maps are updated.

On token_metadata updates, entries for groups that no longer exist
(e.g. table dropped, tablet merged) are evicted. Entries for groups
that still exist are kept — use-time validation handles staleness.

An on_node_resolved callback is propagated through the redirect/bounce
path so the transport layer can update the cache generically without
coupling to the strong-consistency coordinator. The coordinator creates
the callback only for writes (capturing the groups_manager and
group_id) and attaches it to the bounce message; the transport layer
invokes it once the final node is known, keeping the forwarding
infrastructure subsystem-agnostic.

We also add a test which verifies that after the initial redirect,
following requests to the same node avoid the extra redirect and
forward directly to the leader.

Fixes: SCYLLADB-1064

Closes scylladb/scylladb#29392
2026-05-21 10:32:56 +02:00

461 lines
23 KiB
C++
Raw Blame History

/*
* Copyright (C) 2015-present ScyllaDB
*/
/*
* SPDX-License-Identifier: LicenseRef-ScyllaDB-Source-Available-1.1
*/
#pragma once
#include "auth/service.hh"
#include <seastar/core/seastar.hh>
#include <seastar/core/scheduling.hh>
#include "cql3/prepared_statements_cache.hh"
#include "service/endpoint_lifecycle_subscriber.hh"
#include "service/migration_listener.hh"
#include "auth/authenticator.hh"
#include <seastar/core/sharded.hh>
#include "service/qos/qos_configuration_change_subscriber.hh"
#include "timeout_config.hh"
#include <seastar/core/semaphore.hh>
#include <memory>
#include <type_traits>
#include <boost/intrusive/list.hpp>
#include <seastar/net/tls.hh>
#include <seastar/core/metrics_registration.hh>
#include "utils/fragmented_temporary_buffer.hh"
#include "utils/result.hh"
#include "service_permit.hh"
#include <seastar/core/sharded.hh>
#include <seastar/core/execution_stage.hh>
#include "utils/updateable_value.hh"
#include "generic_server.hh"
#include "service/query_state.hh"
#include "cql3/query_options.hh"
#include "cql3/dialect.hh"
#include "transport/messages/result_message.hh"
#include "utils/chunked_vector.hh"
#include "exceptions/coordinator_result.hh"
#include "exceptions/exceptions.hh"
#include "db/operation_type.hh"
#include "service/maintenance_mode.hh"
#include "service/client_routes.hh"
#include "utils/estimated_histogram.hh"
#include "transport/forward.hh"
namespace cql3 {
class query_processor;
}
namespace scollectd {
class registrations;
}
namespace service {
class memory_limiter;
}
enum class client_type;
struct client_data;
namespace qos {
class service_level_controller;
} // namespace qos
namespace gms {
class gossiper;
}
namespace netw {
class messaging_service;
}
namespace cql_transport {
class request_reader;
class response;
enum class cql_binary_opcode : uint8_t;
enum class cql_compression {
none,
lz4,
snappy,
};
enum cql_frame_flags {
compression = 0x01,
tracing = 0x02,
custom_payload = 0x04,
warning = 0x08,
};
struct [[gnu::packed]] cql_binary_frame_v3 {
uint8_t version;
uint8_t flags;
net::packed<uint16_t> stream;
uint8_t opcode;
net::packed<uint32_t> length;
template <typename Adjuster>
void adjust_endianness(Adjuster a) {
return a(stream, length);
}
};
struct cql_query_state {
service::query_state query_state;
std::unique_ptr<cql3::query_options> options;
cql_query_state(service::client_state& client_state, tracing::trace_state_ptr trace_state_ptr, service_permit permit)
: query_state(client_state, std::move(trace_state_ptr), std::move(permit))
{ }
};
struct cql_server_config {
updateable_timeout_config& timeout_config;
size_t max_request_size;
sstring partitioner_name;
unsigned sharding_ignore_msb;
std::optional<uint16_t> shard_aware_transport_port;
std::optional<uint16_t> shard_aware_transport_port_ssl;
bool allow_shard_aware_drivers = true;
smp_service_group bounce_request_smp_service_group = default_smp_service_group();
utils::updateable_value<uint32_t> max_concurrent_requests;
utils::updateable_value<bool> cql_duplicate_bind_variable_names_refer_to_same_variable;
utils::updateable_value<uint32_t> uninitialized_connections_semaphore_cpu_concurrency;
utils::updateable_value<uint32_t> request_timeout_on_shutdown_in_seconds;
};
/**
* CQL op-code stats collected for each scheduling group
*/
struct cql_sg_stats {
struct request_kind_stats {
utils::bytes_histogram request_size;
utils::bytes_histogram response_size;
};
cql_sg_stats(maintenance_socket_enabled);
request_kind_stats& get_cql_opcode_stats(cql_binary_opcode op) { return _cql_requests_stats[static_cast<uint8_t>(op)]; }
void register_metrics();
void rename_metrics();
// Track total memory consumed by responses waiting to be sent.
// Incremented when a response is queued, decremented when the write completes.
int64_t _pending_response_memory = 0;
// Track the number of CQL requests currently being processed in this
// scheduling group. Incremented when a request starts processing,
// decremented after the response is sent.
uint32_t _requests_serving = 0;
// Latency histogram tracking the transport-level request lifetime:
// from the start of request processing until the response is written
// to the socket. This captures processing time and time waiting in the
// response write queue, complementing storage-proxy-level latency.
utils::time_estimated_histogram _request_latency;
private:
bool _use_metrics = false;
seastar::metrics::metric_groups _metrics;
std::vector<request_kind_stats> _cql_requests_stats;
};
struct connection_service_level_params {
sstring role_name;
timeout_config timeout_config;
qos::service_level_options::workload_type workload_type;
sstring scheduling_group_name;
};
class cql_metadata_id_wrapper {
private:
std::optional<cql3::cql_metadata_id_type> _request_metadata_id;
std::optional<cql3::cql_metadata_id_type> _response_metadata_id;
public:
cql_metadata_id_wrapper()
: _request_metadata_id(std::nullopt)
, _response_metadata_id(std::nullopt)
{ }
explicit cql_metadata_id_wrapper(cql3::cql_metadata_id_type&& response_metadata_id)
: _request_metadata_id(std::nullopt)
, _response_metadata_id(std::move(response_metadata_id))
{ }
cql_metadata_id_wrapper(cql3::cql_metadata_id_type&& request_metadata_id, cql3::cql_metadata_id_type&& response_metadata_id)
: _request_metadata_id(std::move(request_metadata_id))
, _response_metadata_id(std::move(response_metadata_id))
{ }
bool has_request_metadata_id() const;
bool has_response_metadata_id() const;
const cql3::cql_metadata_id_type& get_request_metadata_id() const;
const cql3::cql_metadata_id_type& get_response_metadata_id() const;
};
class cql_server : public seastar::peering_sharded_service<cql_server>, public generic_server::server {
private:
struct transport_stats {
// server stats
uint64_t connects = 0;
uint64_t connections = 0;
uint64_t requests_served = 0;
uint32_t requests_serving = 0;
uint64_t requests_blocked_memory = 0;
uint64_t requests_shed = 0;
// forwarding stats
uint64_t requests_forwarded_successfully = 0;
uint64_t requests_forwarded_failed = 0;
uint64_t requests_forwarded_redirected = 0;
uint64_t requests_forwarded_prepared_not_found = 0;
std::unordered_map<exceptions::exception_code, uint64_t> errors;
};
private:
class event_notifier;
static constexpr cql_protocol_version_type current_version = cql_serialization_format::latest_version;
sharded<cql3::query_processor>& _query_processor;
netw::messaging_service& _ms;
cql_server_config _config;
semaphore& _memory_available;
const uint32_t _total_memory;
seastar::metrics::metric_groups _metrics;
std::unique_ptr<event_notifier> _notifier;
private:
client_options_cache_type _connection_options_keys_and_values;
transport_stats _stats;
auth::service& _auth_service;
qos::service_level_controller& _sl_controller;
gms::gossiper& _gossiper;
scheduling_group_key _stats_key;
maintenance_socket_enabled _used_by_maintenance_socket;
public:
cql_server(sharded<cql3::query_processor>& qp, auth::service&,
service::memory_limiter& ml,
cql_server_config config,
qos::service_level_controller& sl_controller,
gms::gossiper& g,
scheduling_group_key stats_key,
maintenance_socket_enabled used_by_maintenance_socket,
netw::messaging_service& ms);
~cql_server();
future<> stop();
public:
using response = cql_transport::response;
using result_with_foreign_response_ptr = exceptions::coordinator_result<foreign_ptr<std::unique_ptr<cql_server::response>>>;
using result_with_bounce = foreign_ptr<seastar::shared_ptr<messages::result_message::bounce>>;
using process_fn_return_type = std::variant<result_with_foreign_response_ptr, result_with_bounce>;
service::endpoint_lifecycle_subscriber* get_lifecycle_listener() const noexcept;
service::migration_listener* get_migration_listener() const noexcept;
qos::qos_configuration_change_subscriber* get_qos_configuration_listener() const noexcept;
cql_sg_stats& get_cql_sg_stats() {
return scheduling_group_get_specific<cql_sg_stats>(_stats_key);
}
cql_sg_stats::request_kind_stats& get_cql_opcode_stats(cql_binary_opcode op) {
return get_cql_sg_stats().get_cql_opcode_stats(op);
}
future<utils::chunked_vector<foreign_ptr<std::unique_ptr<client_data>>>> get_client_data();
future<> update_connections_scheduling_group();
future<> update_connections_service_level_params();
future<std::vector<connection_service_level_params>> get_connections_service_level_params();
private:
class fmt_visitor;
friend class connection;
friend std::unique_ptr<cql_server::response> make_result(int16_t stream, messages::result_message& msg,
const tracing::trace_state_ptr& tr_state, cql_protocol_version_type version, cql_metadata_id_wrapper&& metadata_id, bool skip_metadata);
static std::unique_ptr<cql_server::response> make_unavailable_error(int16_t stream, exceptions::exception_code err, sstring msg, db::consistency_level cl, int32_t required, int32_t alive, const tracing::trace_state_ptr& tr_state);
static std::unique_ptr<cql_server::response> make_read_timeout_error(int16_t stream, exceptions::exception_code err, sstring msg, db::consistency_level cl, int32_t received, int32_t blockfor, bool data_present, const tracing::trace_state_ptr& tr_state);
static std::unique_ptr<cql_server::response> make_read_failure_error(int16_t stream, exceptions::exception_code err, sstring msg, db::consistency_level cl, int32_t received, int32_t numfailures, int32_t blockfor, bool data_present, const tracing::trace_state_ptr& tr_state, cql_protocol_version_type version);
static std::unique_ptr<cql_server::response> make_mutation_write_timeout_error(int16_t stream, exceptions::exception_code err, sstring msg, db::consistency_level cl, int32_t received, int32_t blockfor, db::write_type type, const tracing::trace_state_ptr& tr_state);
static std::unique_ptr<cql_server::response> make_mutation_write_failure_error(int16_t stream, exceptions::exception_code err, sstring msg, db::consistency_level cl, int32_t received, int32_t numfailures, int32_t blockfor, db::write_type type, const tracing::trace_state_ptr& tr_state, cql_protocol_version_type version);
static std::unique_ptr<cql_server::response> make_already_exists_error(int16_t stream, exceptions::exception_code err, sstring msg, sstring ks_name, sstring cf_name, const tracing::trace_state_ptr& tr_state);
static std::unique_ptr<cql_server::response> make_unprepared_error(int16_t stream, exceptions::exception_code err, sstring msg, bytes id, const tracing::trace_state_ptr& tr_state);
static std::unique_ptr<cql_server::response> make_function_failure_error(int16_t stream, exceptions::exception_code err, sstring msg, sstring ks_name, sstring func_name, std::vector<sstring> args, const tracing::trace_state_ptr& tr_state);
static std::unique_ptr<cql_server::response> make_rate_limit_error(int16_t stream, exceptions::exception_code err, sstring msg, db::operation_type op_type, bool rejected_by_coordinator, const tracing::trace_state_ptr& tr_state, bool has_rate_limit_extension);
static std::unique_ptr<cql_server::response> make_error(int16_t stream, exceptions::exception_code err, sstring msg, const tracing::trace_state_ptr& tr_state);
std::unique_ptr<cql_server::response> handle_exception(int16_t stream, std::exception_ptr eptr, const tracing::trace_state_ptr& trace_state, cql_protocol_version_type version, const service::client_state& client_state);
std::optional<seastar::lowres_clock::time_point> timeout_for_sleep(std::exception_ptr eptr) const;
future<foreign_ptr<std::unique_ptr<cql_server::response>>> sleep_until_timeout_passes(const seastar::lowres_clock::time_point& timeout, foreign_ptr<std::unique_ptr<cql_server::response>>&& resp);
class connection : public generic_server::connection {
cql_server& _server;
socket_address _server_addr;
fragmented_temporary_buffer::reader _buffer_reader;
cql_protocol_version_type _version = 0;
cql_compression _compression = cql_compression::none;
service::client_state _client_state;
timer<lowres_clock> _shedding_timer;
scheduling_group _current_scheduling_group;
bool _shed_incoming_requests = false;
bool _ready = false;
bool _authenticating = false;
bool _tenant_switch = false;
enum class tracing_request_type : uint8_t {
not_requested,
no_write_on_close,
write_on_close
};
private:
using execution_stage_type = inheriting_concrete_execution_stage<
future<foreign_ptr<std::unique_ptr<cql_server::response>>>,
cql_server::connection*,
fragmented_temporary_buffer::istream,
uint8_t,
uint16_t,
service::client_state&,
tracing_request_type,
service_permit>;
static thread_local execution_stage_type _process_request_stage;
public:
connection(cql_server& server, socket_address server_addr, connected_socket&& fd, socket_address addr, named_semaphore& sem, semaphore_units<named_semaphore_exception_factory> initial_sem_units);
virtual ~connection();
future<> process_request() override;
void handle_error(future<>&& f) override;
client_data make_client_data() const;
const service::client_state& get_client_state() const { return _client_state; }
void update_scheduling_group();
service::client_state& get_client_state() { return _client_state; }
scheduling_group get_scheduling_group() const { return _current_scheduling_group; }
private:
friend class process_request_executor;
future<foreign_ptr<std::unique_ptr<cql_server::response>>> process_request_one(fragmented_temporary_buffer::istream buf, uint8_t op, uint16_t stream, service::client_state& client_state, tracing_request_type tracing_request, service_permit permit);
unsigned frame_size() const;
unsigned pick_request_cpu();
utils::result_with_exception<cql_binary_frame_v3, exceptions::protocol_exception, class cql_frame_error> parse_frame(temporary_buffer<char> buf) const;
future<fragmented_temporary_buffer> read_and_decompress_frame(size_t length, uint8_t flags);
future<std::optional<cql_binary_frame_v3>> read_frame();
future<std::unique_ptr<cql_server::response>> process_startup(uint16_t stream, request_reader in, service::client_state& client_state, tracing::trace_state_ptr trace_state);
future<std::unique_ptr<cql_server::response>> process_auth_response(uint16_t stream, request_reader in, service::client_state& client_state, tracing::trace_state_ptr trace_state);
future<std::unique_ptr<cql_server::response>> process_options(uint16_t stream, request_reader in, service::client_state& client_state, tracing::trace_state_ptr trace_state);
future<std::unique_ptr<cql_server::response>> process_prepare(uint16_t stream, request_reader in, service::client_state& client_state, tracing::trace_state_ptr trace_state);
future<std::unique_ptr<cql_server::response>> process_register(uint16_t stream, request_reader in, service::client_state& client_state, tracing::trace_state_ptr trace_state);
std::unique_ptr<cql_server::response> make_ready(int16_t stream, const tracing::trace_state_ptr& tr_state) const;
std::unique_ptr<cql_server::response> make_supported(int16_t stream, const tracing::trace_state_ptr& tr_state) const;
std::unique_ptr<cql_server::response> make_topology_change_event(const cql_transport::event::topology_change& event) const;
std::unique_ptr<cql_server::response> make_status_change_event(const cql_transport::event::status_change& event) const;
std::unique_ptr<cql_server::response> make_schema_change_event(const cql_transport::event::schema_change& event) const;
std::unique_ptr<cql_server::response> make_client_routes_change_event(const cql_transport::event::client_routes_change& event) const;
std::unique_ptr<cql_server::response> make_autheticate(int16_t, std::string_view, const tracing::trace_state_ptr& tr_state) const;
std::unique_ptr<cql_server::response> make_auth_success(int16_t, bytes, const tracing::trace_state_ptr& tr_state) const;
std::unique_ptr<cql_server::response> make_auth_challenge(int16_t, bytes, const tracing::trace_state_ptr& tr_state) const;
cql3::dialect get_dialect() const;
void write_response(foreign_ptr<std::unique_ptr<cql_server::response>>&& response, cql_compression compression = cql_compression::none);
void update_user_scheduling_group(const std::optional<auth::authenticated_user>& usr);
void update_control_connection_scheduling_group();
friend event_notifier;
};
using handling_node_bounce = bool_class<class handling_node_bounce_class>;
// Helper functions to encapsulate bounce processing for query, execute and batch verbs
future<process_fn_return_type>
process(uint16_t stream, request_reader in, service::client_state& client_state, service_permit permit, tracing::trace_state_ptr trace_state,
cql_binary_opcode opcode, cql_protocol_version_type version, cql3::dialect dialect, cql3::computed_function_values cached_fn_calls = {}, handling_node_bounce bounced = handling_node_bounce::no);
friend class type_codec;
private:
void init_messaging_service();
future<> uninit_messaging_service();
future<forward_cql_execute_response> handle_forward_execute(service::query_state& qs, forward_cql_execute_request& req);
struct forward_cql_result {
foreign_ptr<std::unique_ptr<cql_transport::response>> response;
locator::host_id final_host;
};
future<forward_cql_result> forward_cql(locator::host_id target_host, unsigned target_shard, seastar::lowres_clock::time_point timeout,
bool is_write, uint16_t stream, tracing::trace_state_ptr trace_state, forward_cql_execute_request req);
virtual shared_ptr<generic_server::connection> make_connection(socket_address server_addr, connected_socket&& fd, socket_address addr, named_semaphore& sem, semaphore_units<named_semaphore_exception_factory> initial_sem_units) override;
scheduling_group get_scheduling_group_for_new_connection() const override {
if (_sl_controller.has_service_level(qos::service_level_controller::driver_service_level_name)) {
return _sl_controller.get_scheduling_group(qos::service_level_controller::driver_service_level_name);
}
return default_scheduling_group();
}
::timeout_config timeout_config() const { return _config.timeout_config.current_values(); }
};
class cql_server::event_notifier : public service::migration_listener,
public service::endpoint_lifecycle_subscriber,
public qos::qos_configuration_change_subscriber
{
cql_server& _server;
std::set<cql_server::connection*> _topology_change_listeners;
std::set<cql_server::connection*> _status_change_listeners;
std::set<cql_server::connection*> _schema_change_listeners;
std::set<cql_server::connection*> _client_routes_change_listeners;
std::unordered_map<gms::inet_address, event::status_change::status_type> _last_status_change;
// We want to delay sending NEW_NODE CQL event to clients until the new node
// has started listening for CQL requests.
std::unordered_set<gms::inet_address> _endpoints_pending_joined_notification;
void send_join_cluster(const gms::inet_address& endpoint);
public:
explicit event_notifier(cql_server& s) noexcept : _server(s) {}
void register_event(cql_transport::event::event_type et, cql_server::connection* conn);
void unregister_connection(cql_server::connection* conn);
virtual void on_create_keyspace(const sstring& ks_name) override;
virtual void on_create_column_family(const sstring& ks_name, const sstring& cf_name) override;
virtual void on_create_user_type(const sstring& ks_name, const sstring& type_name) override;
virtual void on_create_view(const sstring& ks_name, const sstring& view_name) override;
virtual void on_create_function(const sstring& ks_name, const sstring& function_name) override;
virtual void on_create_aggregate(const sstring& ks_name, const sstring& aggregate_name) override;
virtual void on_update_keyspace(const sstring& ks_name) override;
virtual void on_update_column_family(const sstring& ks_name, const sstring& cf_name, bool columns_changed) override;
virtual void on_update_user_type(const sstring& ks_name, const sstring& type_name) override;
virtual void on_update_view(const sstring& ks_name, const sstring& view_name, bool columns_changed) override;
virtual void on_update_function(const sstring& ks_name, const sstring& function_name) override;
virtual void on_update_aggregate(const sstring& ks_name, const sstring& aggregate_name) override;
virtual void on_drop_keyspace(const sstring& ks_name) override;
virtual void on_drop_column_family(const sstring& ks_name, const sstring& cf_name) override;
virtual void on_drop_user_type(const sstring& ks_name, const sstring& type_name) override;
virtual void on_drop_view(const sstring& ks_name, const sstring& view_name) override;
virtual void on_drop_function(const sstring& ks_name, const sstring& function_name) override;
virtual void on_drop_aggregate(const sstring& ks_name, const sstring& aggregate_name) override;
virtual future<> on_before_service_level_add(qos::service_level_options, qos::service_level_info sl_info) override;
virtual future<> on_after_service_level_remove(qos::service_level_info sl_info) override;
virtual future<> on_before_service_level_change(qos::service_level_options slo_before, qos::service_level_options slo_after, qos::service_level_info sl_info) override;
virtual future<> on_effective_service_levels_cache_reloaded() override;
virtual void on_join_cluster(const gms::inet_address& endpoint, locator::host_id hid) override;
virtual void on_leave_cluster(const gms::inet_address& endpoint, const locator::host_id& hid) override;
virtual void on_up(const gms::inet_address& endpoint, locator::host_id hid) override;
virtual void on_down(const gms::inet_address& endpoint, locator::host_id hid) override;
virtual void on_client_routes_change(const service::client_routes_service::client_route_keys& client_route_keys) override;
};
inline service::endpoint_lifecycle_subscriber* cql_server::get_lifecycle_listener() const noexcept { return _notifier.get(); }
inline service::migration_listener* cql_server::get_migration_listener() const noexcept { return _notifier.get(); }
inline qos::qos_configuration_change_subscriber* cql_server::get_qos_configuration_listener() const noexcept { return _notifier.get(); }
}
Reference in New Issue View Git Blame Copy Permalink