be/src/runtime/query_context.h

// Licensed to the Apache Software Foundation (ASF) under one
// or more contributor license agreements.  See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership.  The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License.  You may obtain a copy of the License at
//
//   http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing,
// software distributed under the License is distributed on an
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, either express or implied.  See the License for the
// specific language governing permissions and limitations
// under the License.

#pragma once

#include <gen_cpp/PaloInternalService_types.h>
#include <gen_cpp/RuntimeProfile_types.h>
#include <gen_cpp/Types_types.h>
#include <glog/logging.h>

#include <atomic>
#include <cstdint>
#include <memory>
#include <mutex>
#include <string>
#include <unordered_map>

#include "common/config.h"
#include "common/factory_creator.h"
#include "common/object_pool.h"
#include "pipeline/dependency.h"
#include "runtime/exec_env.h"
#include "runtime/memory/mem_tracker_limiter.h"
#include "runtime/runtime_filter_mgr.h"
#include "runtime/runtime_predicate.h"
#include "runtime/workload_management/resource_context.h"
#include "util/hash_util.hpp"
#include "util/threadpool.h"
#include "vec/exec/scan/scanner_scheduler.h"
#include "vec/runtime/shared_hash_table_controller.h"
#include "workload_group/workload_group.h"

namespace doris {

namespace pipeline {
class PipelineFragmentContext;
class PipelineTask;
} // namespace pipeline

struct ReportStatusRequest {
    const Status status;
    std::vector<RuntimeState*> runtime_states;
    bool done;
    TNetworkAddress coord_addr;
    TUniqueId query_id;
    int fragment_id;
    TUniqueId fragment_instance_id;
    int backend_num;
    RuntimeState* runtime_state;
    std::function<void(const Status&)> cancel_fn;
};

enum class QuerySource {
    INTERNAL_FRONTEND,
    STREAM_LOAD,
    GROUP_COMMIT_LOAD,
    ROUTINE_LOAD,
    EXTERNAL_CONNECTOR
};

const std::string toString(QuerySource query_source);

// Save the common components of fragments in a query.
// Some components like DescriptorTbl may be very large
// that will slow down each execution of fragments when DeSer them every time.
class DescriptorTbl;
class QueryContext : public std::enable_shared_from_this<QueryContext> {
    ENABLE_FACTORY_CREATOR(QueryContext);

public:
    class QueryTaskController : public TaskController {
        ENABLE_FACTORY_CREATOR(QueryTaskController);

    public:
        static std::unique_ptr<TaskController> create(QueryContext* query_ctx);

        bool is_cancelled() const override;
        Status cancel(const Status& reason, int fragment_id);
        Status cancel(const Status& reason) override { return cancel(reason, -1); }

    private:
        QueryTaskController(const std::shared_ptr<QueryContext>& query_ctx)
                : query_ctx_(query_ctx) {}

        const std::weak_ptr<QueryContext> query_ctx_;
    };

    class QueryMemoryContext : public MemoryContext {
        ENABLE_FACTORY_CREATOR(QueryMemoryContext);

    public:
        static std::unique_ptr<MemoryContext> create();

        int64_t revokable_bytes() override {
            // TODO
            return 0;
        }

        bool ready_do_revoke() override {
            // TODO
            return true;
        }

        Status revoke(int64_t bytes) override {
            // TODO
            return Status::OK();
        }

        Status enter_arbitration(Status reason) override {
            // TODO, pause the pipeline
            return Status::OK();
        }

        Status leave_arbitration(Status reason) override {
            // TODO, start pipeline
            return Status::OK();
        }

    private:
        QueryMemoryContext() = default;
    };

    static std::shared_ptr<QueryContext> create(TUniqueId query_id, ExecEnv* exec_env,
                                                const TQueryOptions& query_options,
                                                TNetworkAddress coord_addr, bool is_nereids,
                                                TNetworkAddress current_connect_fe,
                                                QuerySource query_type);

    // use QueryContext::create, cannot be made private because of ENABLE_FACTORY_CREATOR::create_shared.
    QueryContext(TUniqueId query_id, ExecEnv* exec_env, const TQueryOptions& query_options,
                 TNetworkAddress coord_addr, bool is_nereids, TNetworkAddress current_connect_fe,
                 QuerySource query_type);

    ~QueryContext();

    void init_query_task_controller();

    ExecEnv* exec_env() { return _exec_env; }

    bool is_timeout(timespec now) const {
        if (_timeout_second <= 0) {
            return false;
        }
        return _query_watcher.elapsed_time_seconds(now) > _timeout_second;
    }

    void set_thread_token(int concurrency, bool is_serial) {
        _thread_token = _exec_env->scanner_scheduler()->new_limited_scan_pool_token(
                is_serial ? ThreadPool::ExecutionMode::SERIAL
                          : ThreadPool::ExecutionMode::CONCURRENT,
                concurrency);
    }

    ThreadPoolToken* get_token() { return _thread_token.get(); }

    void set_ready_to_execute(Status reason);

    [[nodiscard]] bool is_cancelled() const { return !_exec_status.ok(); }

    void cancel_all_pipeline_context(const Status& reason, int fragment_id = -1);
    std::string print_all_pipeline_context();
    void set_pipeline_context(const int fragment_id,
                              std::shared_ptr<pipeline::PipelineFragmentContext> pip_ctx);
    void cancel(Status new_status, int fragment_id = -1);

    [[nodiscard]] Status exec_status() { return _exec_status.status(); }

    void set_execution_dependency_ready();

    void set_memory_sufficient(bool sufficient);

    void set_ready_to_execute_only();

    std::shared_ptr<vectorized::SharedHashTableController> get_shared_hash_table_controller() {
        return _shared_hash_table_controller;
    }

    bool has_runtime_predicate(int source_node_id) {
        return _runtime_predicates.contains(source_node_id);
    }

    vectorized::RuntimePredicate& get_runtime_predicate(int source_node_id) {
        DCHECK(has_runtime_predicate(source_node_id));
        return _runtime_predicates.find(source_node_id)->second;
    }

    void init_runtime_predicates(const std::vector<TTopnFilterDesc>& topn_filter_descs) {
        for (auto desc : topn_filter_descs) {
            _runtime_predicates.try_emplace(desc.source_node_id, desc);
        }
    }

    void set_workload_group(WorkloadGroupPtr& wg);

    int execution_timeout() const {
        return _query_options.__isset.execution_timeout ? _query_options.execution_timeout
                                                        : _query_options.query_timeout;
    }

    int32_t runtime_filter_wait_time_ms() const {
        return _query_options.runtime_filter_wait_time_ms;
    }

    bool runtime_filter_wait_infinitely() const {
        return _query_options.__isset.runtime_filter_wait_infinitely &&
               _query_options.runtime_filter_wait_infinitely;
    }

    int be_exec_version() const {
        if (!_query_options.__isset.be_exec_version) {
            return 0;
        }
        return _query_options.be_exec_version;
    }

    [[nodiscard]] int64_t get_fe_process_uuid() const {
        return _query_options.__isset.fe_process_uuid ? _query_options.fe_process_uuid : 0;
    }

    bool ignore_runtime_filter_error() const {
        return _query_options.__isset.ignore_runtime_filter_error
                       ? _query_options.ignore_runtime_filter_error
                       : false;
    }

    bool enable_force_spill() const {
        return _query_options.__isset.enable_force_spill && _query_options.enable_force_spill;
    }

    // global runtime filter mgr, the runtime filter have remote target or
    // need local merge should regist here. before publish() or push_to_remote()
    // the runtime filter should do the local merge work
    RuntimeFilterMgr* runtime_filter_mgr() { return _runtime_filter_mgr.get(); }

    TUniqueId query_id() const { return _query_id; }

    vectorized::SimplifiedScanScheduler* get_scan_scheduler() { return _scan_task_scheduler; }

    vectorized::SimplifiedScanScheduler* get_remote_scan_scheduler() {
        return _remote_scan_task_scheduler;
    }

    pipeline::Dependency* get_execution_dependency() { return _execution_dependency.get(); }
    pipeline::Dependency* get_memory_sufficient_dependency() {
        return _memory_sufficient_dependency.get();
    }

    std::vector<pipeline::PipelineTask*> get_revocable_tasks() const;

    Status revoke_memory();

    doris::pipeline::TaskScheduler* get_pipe_exec_scheduler();

    ThreadPool* get_memtable_flush_pool();

    void set_merge_controller_handler(
            std::shared_ptr<RuntimeFilterMergeControllerEntity>& handler) {
        _merge_controller_handler = handler;
    }

    bool is_nereids() const { return _is_nereids; }

    WorkloadGroupPtr workload_group() const { return _resource_ctx->workload_group(); }
    std::shared_ptr<MemTrackerLimiter> query_mem_tracker() const {
        return _resource_ctx->memory_context()->mem_tracker();
    }

    void increase_revoking_tasks_count() { _revoking_tasks_count.fetch_add(1); }

    void decrease_revoking_tasks_count();

    int get_revoking_tasks_count() const { return _revoking_tasks_count.load(); }

    void get_revocable_info(size_t* revocable_size, size_t* memory_usage,
                            bool* has_running_task) const;
    size_t get_revocable_size() const;

    // This method is called by workload group manager to set query's memlimit using slot
    // If user set query limit explicitly, then should use less one
    void set_mem_limit(int64_t new_mem_limit) {
        _resource_ctx->memory_context()->mem_tracker()->set_limit(new_mem_limit);
    }

    int64_t get_mem_limit() const {
        return _resource_ctx->memory_context()->mem_tracker()->limit();
    }

    // The new memlimit should be less than user set memlimit.
    void set_adjusted_mem_limit(int64_t new_mem_limit) {
        _adjusted_mem_limit = std::min<int64_t>(new_mem_limit, _user_set_mem_limit);
    }

    // Expected mem limit is the limit when workload group reached limit.
    int64_t adjusted_mem_limit() { return _adjusted_mem_limit; }

    MemTrackerLimiter* get_mem_tracker() {
        return _resource_ctx->memory_context()->mem_tracker().get();
    }

    int32_t get_slot_count() const {
        return _query_options.__isset.query_slot_count ? _query_options.query_slot_count : 1;
    }

    DescriptorTbl* desc_tbl = nullptr;
    bool set_rsc_info = false;
    std::string user;
    std::string group;
    TNetworkAddress coord_addr;
    TNetworkAddress current_connect_fe;
    TQueryGlobals query_globals;

    ObjectPool obj_pool;

    std::shared_ptr<ResourceContext> resource_ctx() { return _resource_ctx; }

    std::vector<TUniqueId> fragment_instance_ids;

    // plan node id -> TFileScanRangeParams
    // only for file scan node
    std::map<int, TFileScanRangeParams> file_scan_range_params_map;

    void add_using_brpc_stub(const TNetworkAddress& network_address,
                             std::shared_ptr<PBackendService_Stub> brpc_stub) {
        if (network_address.port == 0) {
            return;
        }
        std::lock_guard<std::mutex> lock(_brpc_stubs_mutex);
        if (!_using_brpc_stubs.contains(network_address)) {
            _using_brpc_stubs.emplace(network_address, brpc_stub);
        }

        DCHECK_EQ(_using_brpc_stubs[network_address].get(), brpc_stub.get());
    }

    std::unordered_map<TNetworkAddress, std::shared_ptr<PBackendService_Stub>>
    get_using_brpc_stubs() {
        std::lock_guard<std::mutex> lock(_brpc_stubs_mutex);
        return _using_brpc_stubs;
    }

    void set_low_memory_mode() { _low_memory_mode = true; }

    bool low_memory_mode() { return _low_memory_mode; }

    void disable_reserve_memory() { _enable_reserve_memory = false; }

    bool enable_reserve_memory() const {
        return _query_options.__isset.enable_reserve_memory &&
               _query_options.enable_reserve_memory && _enable_reserve_memory;
    }

    void update_paused_reason(const Status& st) {
        std::lock_guard l(_paused_mutex);
        if (_paused_reason.is<ErrorCode::QUERY_MEMORY_EXCEEDED>()) {
            return;
        } else if (_paused_reason.is<ErrorCode::WORKLOAD_GROUP_MEMORY_EXCEEDED>()) {
            if (st.is<ErrorCode::QUERY_MEMORY_EXCEEDED>()) {
                _paused_reason = st;
                return;
            } else {
                return;
            }
        } else {
            _paused_reason = st;
        }
    }

    Status paused_reason() {
        std::lock_guard l(_paused_mutex);
        return _paused_reason;
    }

    bool is_pure_load_task() {
        return _query_source == QuerySource::STREAM_LOAD ||
               _query_source == QuerySource::ROUTINE_LOAD ||
               _query_source == QuerySource::GROUP_COMMIT_LOAD;
    }

    std::string debug_string();

private:
    int _timeout_second;
    TUniqueId _query_id;
    ExecEnv* _exec_env = nullptr;
    MonotonicStopWatch _query_watcher;
    bool _is_nereids = false;

    std::shared_ptr<ResourceContext> _resource_ctx;

    std::mutex _revoking_tasks_mutex;
    std::atomic<int> _revoking_tasks_count = 0;

    // A token used to submit olap scanner to the "_limited_scan_thread_pool",
    // This thread pool token is created from "_limited_scan_thread_pool" from exec env.
    // And will be shared by all instances of this query.
    // So that we can control the max thread that a query can be used to execute.
    // If this token is not set, the scanner will be executed in "_scan_thread_pool" in exec env.
    std::unique_ptr<ThreadPoolToken> _thread_token {nullptr};

    void _init_resource_context();
    void _init_query_mem_tracker();

    std::shared_ptr<vectorized::SharedHashTableController> _shared_hash_table_controller;
    std::unordered_map<int, vectorized::RuntimePredicate> _runtime_predicates;

    std::unique_ptr<RuntimeFilterMgr> _runtime_filter_mgr;
    const TQueryOptions _query_options;

    // All pipeline tasks use the same query context to report status. So we need a `_exec_status`
    // to report the real message if failed.
    AtomicStatus _exec_status;

    doris::pipeline::TaskScheduler* _task_scheduler = nullptr;
    vectorized::SimplifiedScanScheduler* _scan_task_scheduler = nullptr;
    ThreadPool* _memtable_flush_pool = nullptr;
    vectorized::SimplifiedScanScheduler* _remote_scan_task_scheduler = nullptr;
    // This dependency indicates if the 2nd phase RPC received from FE.
    std::unique_ptr<pipeline::Dependency> _execution_dependency;
    // This dependency indicates if memory is sufficient to execute.
    std::unique_ptr<pipeline::Dependency> _memory_sufficient_dependency;

    // This shared ptr is never used. It is just a reference to hold the object.
    // There is a weak ptr in runtime filter manager to reference this object.
    std::shared_ptr<RuntimeFilterMergeControllerEntity> _merge_controller_handler;

    std::map<int, std::weak_ptr<pipeline::PipelineFragmentContext>> _fragment_id_to_pipeline_ctx;
    std::mutex _pipeline_map_write_lock;

    std::mutex _paused_mutex;
    Status _paused_reason;
    std::atomic<int64_t> _paused_count = 0;
    std::atomic<bool> _low_memory_mode = false;
    std::atomic<bool> _enable_reserve_memory = true;
    int64_t _user_set_mem_limit = 0;
    std::atomic<int64_t> _adjusted_mem_limit = 0;

    std::mutex _profile_mutex;
    timespec _query_arrival_timestamp;
    // Distinguish the query source, for query that comes from fe, we will have some memory structure on FE to
    // help us manage the query.
    QuerySource _query_source;

    std::mutex _brpc_stubs_mutex;
    std::unordered_map<TNetworkAddress, std::shared_ptr<PBackendService_Stub>> _using_brpc_stubs;

    // when fragment of pipeline is closed, it will register its profile to this map by using add_fragment_profile
    // flatten profile of one fragment:
    // Pipeline 0
    //      PipelineTask 0
    //              Operator 1
    //              Operator 2
    //              Scanner
    //      PipelineTask 1
    //              Operator 1
    //              Operator 2
    //              Scanner
    // Pipeline 1
    //      PipelineTask 2
    //              Operator 3
    //      PipelineTask 3
    //              Operator 3
    // fragment_id -> list<profile>
    std::unordered_map<int, std::vector<std::shared_ptr<TRuntimeProfileTree>>> _profile_map;
    std::unordered_map<int, std::shared_ptr<TRuntimeProfileTree>> _load_channel_profile_map;

    void _report_query_profile();

    std::unordered_map<int, std::vector<std::shared_ptr<TRuntimeProfileTree>>>
    _collect_realtime_query_profile() const;

public:
    // when fragment of pipeline is closed, it will register its profile to this map by using add_fragment_profile
    void add_fragment_profile(
            int fragment_id,
            const std::vector<std::shared_ptr<TRuntimeProfileTree>>& pipeline_profile,
            std::shared_ptr<TRuntimeProfileTree> load_channel_profile);

    TReportExecStatusParams get_realtime_exec_status() const;

    bool enable_profile() const {
        return _query_options.__isset.enable_profile && _query_options.enable_profile;
    }

    timespec get_query_arrival_timestamp() const { return this->_query_arrival_timestamp; }
    QuerySource get_query_source() const { return this->_query_source; }
};

} // namespace doris