be/src/load/channel/load_stream.cpp

Source
// 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.

#include "load/channel/load_stream.h"

#include <brpc/stream.h>
#include <bthread/bthread.h>
#include <bthread/condition_variable.h>
#include <bthread/mutex.h>

#include <memory>
#include <sstream>

#include "bvar/bvar.h"
#include "cloud/config.h"
#include "common/signal_handler.h"
#include "load/channel/load_channel.h"
#include "load/channel/load_stream_mgr.h"
#include "load/channel/load_stream_writer.h"
#include "load/delta_writer/delta_writer.h"
#include "runtime/exec_env.h"
#include "runtime/fragment_mgr.h"
#include "runtime/runtime_profile.h"
#include "runtime/workload_group/workload_group_manager.h"
#include "storage/rowset/rowset_factory.h"
#include "storage/rowset/rowset_meta.h"
#include "storage/storage_engine.h"
#include "storage/tablet/tablet.h"
#include "storage/tablet/tablet_fwd.h"
#include "storage/tablet/tablet_manager.h"
#include "storage/tablet/tablet_schema.h"
#include "storage/tablet_info.h"
#include "util/debug_points.h"
#include "util/thrift_util.h"
#include "util/uid_util.h"

#define UNKNOWN_ID_FOR_TEST 0x7c00

namespace doris {

bvar::Adder<int64_t> g_load_stream_cnt("load_stream_count");
bvar::LatencyRecorder g_load_stream_flush_wait_ms("load_stream_flush_wait_ms");
bvar::Adder<int> g_load_stream_flush_running_threads("load_stream_flush_wait_threads");

TabletStream::TabletStream(const PUniqueId& load_id, int64_t id, int64_t txn_id,
                           LoadStreamMgr* load_stream_mgr, RuntimeProfile* profile)
        : _id(id),
          _next_segid(0),
          _load_id(load_id),
          _txn_id(txn_id),
          _load_stream_mgr(load_stream_mgr) {
    load_stream_mgr->create_token(_flush_token);
    _profile = profile->create_child(fmt::format("TabletStream {}", id), true, true);
    _append_data_timer = ADD_TIMER(_profile, "AppendDataTime");
    _add_segment_timer = ADD_TIMER(_profile, "AddSegmentTime");
    _close_wait_timer = ADD_TIMER(_profile, "CloseWaitTime");
}

inline std::ostream& operator<<(std::ostream& ostr, const TabletStream& tablet_stream) {
    ostr << "load_id=" << print_id(tablet_stream._load_id) << ", txn_id=" << tablet_stream._txn_id
         << ", tablet_id=" << tablet_stream._id << ", status=" << tablet_stream._status.status();
    return ostr;
}

Status TabletStream::init(std::shared_ptr<OlapTableSchemaParam> schema, int64_t index_id,
                          int64_t partition_id) {
    WriteRequest req {
            .tablet_id = _id,
            .txn_id = _txn_id,
            .index_id = index_id,
            .partition_id = partition_id,
            .load_id = _load_id,
            .table_schema_param = schema,
            // TODO(plat1ko): write_file_cache
            .storage_vault_id {},
    };

    _load_stream_writer = std::make_shared<LoadStreamWriter>(&req, _profile);
    DBUG_EXECUTE_IF("TabletStream.init.uninited_writer", {
        _status.update(Status::Uninitialized("fault injection"));
        return _status.status();
    });
    _status.update(_load_stream_writer->init());
    if (!_status.ok()) {
        LOG(INFO) << "failed to init rowset builder due to " << *this;
    }
    return _status.status();
}

Status TabletStream::append_data(const PStreamHeader& header, butil::IOBuf* data) {
    if (!_status.ok()) {
        return _status.status();
    }

    // dispatch add_segment request
    if (header.opcode() == PStreamHeader::ADD_SEGMENT) {
        return add_segment(header, data);
    }

    SCOPED_TIMER(_append_data_timer);

    int64_t src_id = header.src_id();
    uint32_t segid = header.segment_id();
    // Ensure there are enough space and mapping are built.
    SegIdMapping* mapping = nullptr;
    {
        std::lock_guard lock_guard(_lock);
        if (!_segids_mapping.contains(src_id)) {
            _segids_mapping[src_id] = std::make_unique<SegIdMapping>();
        }
        mapping = _segids_mapping[src_id].get();
    }
    if (segid + 1 > mapping->size()) {
        // TODO: Each sender lock is enough.
        std::lock_guard lock_guard(_lock);
        ssize_t origin_size = mapping->size();
        if (segid + 1 > origin_size) {
            mapping->resize(segid + 1, std::numeric_limits<uint32_t>::max());
            for (size_t index = origin_size; index <= segid; index++) {
                mapping->at(index) = _next_segid;
                _next_segid++;
                VLOG_DEBUG << "src_id=" << src_id << ", segid=" << index << " to "
                           << " segid=" << _next_segid - 1 << ", " << *this;
            }
        }
    }

    // Each sender sends data in one segment sequential, so we also do not
    // need a lock here.
    bool eos = header.segment_eos();
    FileType file_type = header.file_type();
    uint32_t new_segid = mapping->at(segid);
    DCHECK(new_segid != std::numeric_limits<uint32_t>::max());
    butil::IOBuf buf = data->movable();
    auto flush_func = [this, new_segid, eos, buf, header, file_type]() mutable {
        signal::set_signal_task_id(_load_id);
        g_load_stream_flush_running_threads << -1;
        auto st = _load_stream_writer->append_data(new_segid, header.offset(), buf, file_type);
        if (!st.ok() && !config::is_cloud_mode()) {
            auto res = ExecEnv::get_tablet(_id);
            TabletSharedPtr tablet =
                    res.has_value() ? std::dynamic_pointer_cast<Tablet>(res.value()) : nullptr;
            if (tablet) {
                tablet->report_error(st);
            }
        }
        if (eos && st.ok()) {
            DBUG_EXECUTE_IF("TabletStream.append_data.unknown_file_type",
                            { file_type = static_cast<FileType>(-1); });
            if (file_type == FileType::SEGMENT_FILE || file_type == FileType::INVERTED_INDEX_FILE) {
                st = _load_stream_writer->close_writer(new_segid, file_type);
            } else {
                st = Status::InternalError(
                        "appent data failed, file type error, file type = {}, "
                        "segment_id={}",
                        file_type, new_segid);
            }
        }
        DBUG_EXECUTE_IF("TabletStream.append_data.append_failed",
                        { st = Status::InternalError("fault injection"); });
        if (!st.ok()) {
            _status.update(st);
            LOG(WARNING) << "write data failed " << st << ", " << *this;
        }
    };
    auto load_stream_flush_token_max_tasks = config::load_stream_flush_token_max_tasks;
    auto load_stream_max_wait_flush_token_time_ms =
            config::load_stream_max_wait_flush_token_time_ms;
    DBUG_EXECUTE_IF("TabletStream.append_data.long_wait", {
        load_stream_flush_token_max_tasks = 0;
        load_stream_max_wait_flush_token_time_ms = 1000;
    });
    MonotonicStopWatch timer;
    timer.start();
    while (_flush_token->num_tasks() >= load_stream_flush_token_max_tasks) {
        if (timer.elapsed_time() / 1000 / 1000 >= load_stream_max_wait_flush_token_time_ms) {
            _status.update(
                    Status::Error<true>("wait flush token back pressure time is more than "
                                        "load_stream_max_wait_flush_token_time {}",
                                        load_stream_max_wait_flush_token_time_ms));
            return _status.status();
        }
        bthread_usleep(2 * 1000); // 2ms
    }
    timer.stop();
    int64_t time_ms = timer.elapsed_time() / 1000 / 1000;
    g_load_stream_flush_wait_ms << time_ms;
    g_load_stream_flush_running_threads << 1;
    Status st = Status::OK();
    DBUG_EXECUTE_IF("TabletStream.append_data.submit_func_failed",
                    { st = Status::InternalError("fault injection"); });
    if (st.ok()) {
        st = _flush_token->submit_func(flush_func);
    }
    if (!st.ok()) {
        _status.update(st);
    }
    return _status.status();
}

Status TabletStream::add_segment(const PStreamHeader& header, butil::IOBuf* data) {
    if (!_status.ok()) {
        return _status.status();
    }

    SCOPED_TIMER(_add_segment_timer);
    DCHECK(header.has_segment_statistics());
    SegmentStatistics stat(header.segment_statistics());

    int64_t src_id = header.src_id();
    uint32_t segid = header.segment_id();
    uint32_t new_segid;
    DBUG_EXECUTE_IF("TabletStream.add_segment.unknown_segid", { segid = UNKNOWN_ID_FOR_TEST; });
    {
        std::lock_guard lock_guard(_lock);
        if (!_segids_mapping.contains(src_id)) {
            _status.update(Status::InternalError(
                    "add segment failed, no segment written by this src be yet, src_id={}, "
                    "segment_id={}",
                    src_id, segid));
            return _status.status();
        }
        DBUG_EXECUTE_IF("TabletStream.add_segment.segid_never_written",
                        { segid = static_cast<uint32_t>(_segids_mapping[src_id]->size()); });
        if (segid >= _segids_mapping[src_id]->size()) {
            _status.update(Status::InternalError(
                    "add segment failed, segment is never written, src_id={}, segment_id={}",
                    src_id, segid));
            return _status.status();
        }
        new_segid = _segids_mapping[src_id]->at(segid);
    }
    DCHECK(new_segid != std::numeric_limits<uint32_t>::max());

    auto add_segment_func = [this, new_segid, stat]() {
        signal::set_signal_task_id(_load_id);
        auto st = _load_stream_writer->add_segment(new_segid, stat);
        DBUG_EXECUTE_IF("TabletStream.add_segment.add_segment_failed",
                        { st = Status::InternalError("fault injection"); });
        if (!st.ok()) {
            _status.update(st);
            LOG(INFO) << "add segment failed " << *this;
        }
    };
    Status st = Status::OK();
    DBUG_EXECUTE_IF("TabletStream.add_segment.submit_func_failed",
                    { st = Status::InternalError("fault injection"); });
    if (st.ok()) {
        st = _flush_token->submit_func(add_segment_func);
    }
    if (!st.ok()) {
        _status.update(st);
    }
    return _status.status();
}

Status TabletStream::_run_in_heavy_work_pool(std::function<Status()> fn) {
    bthread::Mutex mu;
    std::unique_lock<bthread::Mutex> lock(mu);
    bthread::ConditionVariable cv;
    auto st = Status::OK();
    auto func = [this, &mu, &cv, &st, &fn] {
        signal::set_signal_task_id(_load_id);
        st = fn();
        std::lock_guard<bthread::Mutex> lock(mu);
        cv.notify_one();
    };
    bool ret = _load_stream_mgr->heavy_work_pool()->try_offer(func);
    if (!ret) {
        return Status::Error<ErrorCode::INTERNAL_ERROR>(
                "there is not enough thread resource for close load");
    }
    cv.wait(lock);
    return st;
}

void TabletStream::wait_for_flush_tasks() {
    {
        std::lock_guard lock_guard(_lock);
        if (_flush_tasks_done) {
            return;
        }
        _flush_tasks_done = true;
    }

    if (!_status.ok()) {
        _flush_token->shutdown();
        return;
    }

    // Use heavy_work_pool to avoid blocking bthread
    auto st = _run_in_heavy_work_pool([this]() {
        _flush_token->wait();
        return Status::OK();
    });
    if (!st.ok()) {
        // If heavy_work_pool is unavailable, fall back to shutdown
        // which will cancel pending tasks and wait for running tasks
        _flush_token->shutdown();
        _status.update(st);
    }
}

void TabletStream::pre_close() {
    SCOPED_TIMER(_close_wait_timer);
    wait_for_flush_tasks();

    if (!_status.ok()) {
        return;
    }

    DBUG_EXECUTE_IF("TabletStream.close.segment_num_mismatch", { _num_segments++; });
    if (_check_num_segments && (_next_segid.load() != _num_segments)) {
        _status.update(Status::Corruption(
                "segment num mismatch in tablet {}, expected: {}, actual: {}, load_id: {}", _id,
                _num_segments, _next_segid.load(), print_id(_load_id)));
        return;
    }

    _status.update(_run_in_heavy_work_pool([this]() { return _load_stream_writer->pre_close(); }));
}

Status TabletStream::close() {
    if (!_status.ok()) {
        return _status.status();
    }

    SCOPED_TIMER(_close_wait_timer);
    _status.update(_run_in_heavy_work_pool([this]() { return _load_stream_writer->close(); }));
    return _status.status();
}

IndexStream::IndexStream(const PUniqueId& load_id, int64_t id, int64_t txn_id,
                         std::shared_ptr<OlapTableSchemaParam> schema,
                         LoadStreamMgr* load_stream_mgr, RuntimeProfile* profile)
        : _id(id),
          _load_id(load_id),
          _txn_id(txn_id),
          _schema(schema),
          _load_stream_mgr(load_stream_mgr) {
    _profile = profile->create_child(fmt::format("IndexStream {}", id), true, true);
    _append_data_timer = ADD_TIMER(_profile, "AppendDataTime");
    _close_wait_timer = ADD_TIMER(_profile, "CloseWaitTime");
}

IndexStream::~IndexStream() {
    // Ensure all TabletStreams have their flush tokens properly handled before destruction.
    // In normal flow, close() should have called pre_close() on all tablet streams.
    // But if IndexStream is destroyed without close() being called (e.g., on_idle_timeout),
    // we need to wait for flush tasks here to ensure flush tokens are properly shut down.
    for (auto& [_, tablet_stream] : _tablet_streams_map) {
        tablet_stream->wait_for_flush_tasks();
    }
}

Status IndexStream::append_data(const PStreamHeader& header, butil::IOBuf* data) {
    SCOPED_TIMER(_append_data_timer);
    int64_t tablet_id = header.tablet_id();
    TabletStreamSharedPtr tablet_stream;
    {
        std::lock_guard lock_guard(_lock);
        auto it = _tablet_streams_map.find(tablet_id);
        if (it == _tablet_streams_map.end()) {
            _init_tablet_stream(tablet_stream, tablet_id, header.partition_id());
        } else {
            tablet_stream = it->second;
        }
    }

    return tablet_stream->append_data(header, data);
}

void IndexStream::_init_tablet_stream(TabletStreamSharedPtr& tablet_stream, int64_t tablet_id,
                                      int64_t partition_id) {
    tablet_stream = std::make_shared<TabletStream>(_load_id, tablet_id, _txn_id, _load_stream_mgr,
                                                   _profile);
    _tablet_streams_map[tablet_id] = tablet_stream;
    auto st = tablet_stream->init(_schema, _id, partition_id);
    if (!st.ok()) {
        LOG(WARNING) << "tablet stream init failed " << *tablet_stream;
    }
}

void IndexStream::get_all_write_tablet_ids(std::vector<int64_t>* tablet_ids) {
    std::lock_guard lock_guard(_lock);
    for (const auto& [tablet_id, _] : _tablet_streams_map) {
        tablet_ids->push_back(tablet_id);
    }
}

void IndexStream::close(const std::vector<PTabletID>& tablets_to_commit,
                        std::vector<int64_t>* success_tablet_ids, FailedTablets* failed_tablets) {
    std::lock_guard lock_guard(_lock);
    SCOPED_TIMER(_close_wait_timer);
    // open all need commit tablets
    for (const auto& tablet : tablets_to_commit) {
        if (_id != tablet.index_id()) {
            continue;
        }
        TabletStreamSharedPtr tablet_stream;
        auto it = _tablet_streams_map.find(tablet.tablet_id());
        if (it == _tablet_streams_map.end()) {
            _init_tablet_stream(tablet_stream, tablet.tablet_id(), tablet.partition_id());
        } else {
            tablet_stream = it->second;
        }
        if (tablet.has_num_segments()) {
            tablet_stream->add_num_segments(tablet.num_segments());
        } else {
            // for compatibility reasons (sink from old version BE)
            tablet_stream->disable_num_segments_check();
        }
    }

    for (auto& [_, tablet_stream] : _tablet_streams_map) {
        tablet_stream->pre_close();
    }

    for (auto& [_, tablet_stream] : _tablet_streams_map) {
        auto st = tablet_stream->close();
        if (st.ok()) {
            success_tablet_ids->push_back(tablet_stream->id());
        } else {
            LOG(INFO) << "close tablet stream " << *tablet_stream << ", status=" << st;
            failed_tablets->emplace_back(tablet_stream->id(), st);
        }
    }
}

// TODO: Profile is temporary disabled, because:
// 1. It's not being processed by the upstream for now
// 2. There are some problems in _profile->to_thrift()
LoadStream::LoadStream(const PUniqueId& load_id, LoadStreamMgr* load_stream_mgr,
                       bool enable_profile)
        : _load_id(load_id), _enable_profile(false), _load_stream_mgr(load_stream_mgr) {
    g_load_stream_cnt << 1;
    _profile = std::make_unique<RuntimeProfile>("LoadStream");
    _append_data_timer = ADD_TIMER(_profile, "AppendDataTime");
    _close_wait_timer = ADD_TIMER(_profile, "CloseWaitTime");
    TUniqueId load_tid = ((UniqueId)load_id).to_thrift();
#ifndef BE_TEST
    std::shared_ptr<QueryContext> query_context =
            ExecEnv::GetInstance()->fragment_mgr()->get_query_ctx(load_tid);
    if (query_context != nullptr) {
        _resource_ctx = query_context->resource_ctx();
    } else {
        _resource_ctx = ResourceContext::create_shared();
        _resource_ctx->task_controller()->set_task_id(load_tid);
        std::shared_ptr<MemTrackerLimiter> mem_tracker = MemTrackerLimiter::create_shared(
                MemTrackerLimiter::Type::LOAD,
                fmt::format("(FromLoadStream)Load#Id={}", ((UniqueId)load_id).to_string()));
        _resource_ctx->memory_context()->set_mem_tracker(mem_tracker);
    }
#else
    _resource_ctx = ResourceContext::create_shared();
    _resource_ctx->task_controller()->set_task_id(load_tid);
    std::shared_ptr<MemTrackerLimiter> mem_tracker = MemTrackerLimiter::create_shared(
            MemTrackerLimiter::Type::LOAD,
            fmt::format("(FromLoadStream)Load#Id={}", ((UniqueId)load_id).to_string()));
    _resource_ctx->memory_context()->set_mem_tracker(mem_tracker);
#endif
}

LoadStream::~LoadStream() {
    g_load_stream_cnt << -1;
    LOG(INFO) << "load stream is deconstructed " << *this;
}

Status LoadStream::init(const POpenLoadStreamRequest* request) {
    _txn_id = request->txn_id();
    _total_streams = static_cast<int32_t>(request->total_streams());
    _is_incremental = (_total_streams == 0);

    _schema = std::make_shared<OlapTableSchemaParam>();
    RETURN_IF_ERROR(_schema->init(request->schema()));
    for (auto& index : request->schema().indexes()) {
        _index_streams_map[index.id()] = std::make_shared<IndexStream>(
                _load_id, index.id(), _txn_id, _schema, _load_stream_mgr, _profile.get());
    }
    LOG(INFO) << "succeed to init load stream " << *this;
    return Status::OK();
}

bool LoadStream::close(int64_t src_id, const std::vector<PTabletID>& tablets_to_commit,
                       std::vector<int64_t>* success_tablet_ids, FailedTablets* failed_tablets) {
    std::lock_guard<bthread::Mutex> lock_guard(_lock);
    SCOPED_TIMER(_close_wait_timer);

    // we do nothing until recv CLOSE_LOAD from all stream to ensure all data are handled before ack
    _open_streams[src_id]--;
    if (_open_streams[src_id] == 0) {
        _open_streams.erase(src_id);
    }
    _close_load_cnt++;
    LOG(INFO) << "received CLOSE_LOAD from sender " << src_id << ", remaining "
              << _total_streams - _close_load_cnt << " senders, " << *this;

    _tablets_to_commit.insert(_tablets_to_commit.end(), tablets_to_commit.begin(),
                              tablets_to_commit.end());

    if (_close_load_cnt < _total_streams) {
        // do not return commit info if there is remaining streams.
        return false;
    }

    for (auto& [_, index_stream] : _index_streams_map) {
        index_stream->close(_tablets_to_commit, success_tablet_ids, failed_tablets);
    }
    LOG(INFO) << "close load " << *this << ", success_tablet_num=" << success_tablet_ids->size()
              << ", failed_tablet_num=" << failed_tablets->size();
    return true;
}

std::vector<int64_t> LoadStream::mark_eos_sent_and_collect(int64_t stream_id, bool is_incremental) {
    std::lock_guard<bthread::Mutex> lock_guard(_lock);
    std::vector<int64_t> to_close;
    // A non-incremental stream is closed as soon as its own CLOSE_LOAD (and EOS)
    // is handled -- this is the first batch of streams, known up front, not subject
    // to fencing. Closing it promptly also means a duplicate/late CLOSE_LOAD lands
    // on an already-closed stream and is dropped, instead of being counted again.
    // An incremental stream must be deferred (fencing #56120: it may only close once
    // every non-incremental stream is closed), so it is parked in _eos_sent_stream_ids
    // until all CLOSE_LOADs have been received.
    if (is_incremental) {
        // Parked only after the caller sent this stream's EOS via _report_result,
        // so every parked id is safe to close.
        _eos_sent_stream_ids.push_back(stream_id);
    } else {
        to_close.push_back(stream_id);
    }
    // `_close_load_cnt == _total_streams` means every CLOSE_LOAD has been counted by
    // close(). Latch it so that any thread reaching here afterwards also drains the
    // parked incremental streams, guaranteeing none is left un-closed regardless of
    // thread interleaving (fixes the split-lock leak race).
    if (_close_load_cnt >= _total_streams) {
        _all_close_load_received = true;
    }
    if (_all_close_load_received) {
        for (const auto& parked_id : _eos_sent_stream_ids) {
            to_close.push_back(parked_id);
        }
        _eos_sent_stream_ids.clear();
    }
    return to_close;
}

void LoadStream::_report_result(StreamId stream, const Status& status,
                                const std::vector<int64_t>& success_tablet_ids,
                                const FailedTablets& failed_tablets, bool eos) {
    LOG(INFO) << "report result " << *this << ", success tablet num " << success_tablet_ids.size()
              << ", failed tablet num " << failed_tablets.size();
    butil::IOBuf buf;
    PLoadStreamResponse response;
    response.set_eos(eos);
    status.to_protobuf(response.mutable_status());
    for (auto& id : success_tablet_ids) {
        response.add_success_tablet_ids(id);
    }
    for (auto& [id, st] : failed_tablets) {
        auto pb = response.add_failed_tablets();
        pb->set_id(id);
        st.to_protobuf(pb->mutable_status());
    }

    if (_enable_profile && _close_load_cnt == _total_streams) {
        TRuntimeProfileTree tprofile;
        ThriftSerializer ser(false, 4096);
        uint8_t* profile_buf = nullptr;
        uint32_t len = 0;
        std::unique_lock<bthread::Mutex> l(_lock);

        _profile->to_thrift(&tprofile);
        auto st = ser.serialize(&tprofile, &len, &profile_buf);
        if (st.ok()) {
            response.set_load_stream_profile(profile_buf, len);
        } else {
            LOG(WARNING) << "TRuntimeProfileTree serialize failed, errmsg=" << st << ", " << *this;
        }
    }

    buf.append(response.SerializeAsString());
    auto wst = _write_stream(stream, buf);
    if (!wst.ok()) {
        LOG(WARNING) << " report result failed with " << wst << ", " << *this;
    }
}

void LoadStream::_report_schema(StreamId stream, const PStreamHeader& hdr) {
    butil::IOBuf buf;
    PLoadStreamResponse response;
    Status st = Status::OK();
    for (const auto& req : hdr.tablets()) {
        BaseTabletSPtr tablet;
        if (auto res = ExecEnv::get_tablet(req.tablet_id()); res.has_value()) {
            tablet = std::move(res).value();
        } else {
            st = std::move(res).error();
            break;
        }
        auto* resp = response.add_tablet_schemas();
        resp->set_index_id(req.index_id());
        resp->set_enable_unique_key_merge_on_write(tablet->enable_unique_key_merge_on_write());
        tablet->tablet_schema()->to_schema_pb(resp->mutable_tablet_schema());
    }
    st.to_protobuf(response.mutable_status());

    buf.append(response.SerializeAsString());
    auto wst = _write_stream(stream, buf);
    if (!wst.ok()) {
        LOG(WARNING) << " report result failed with " << wst << ", " << *this;
    }
}

void LoadStream::_report_tablet_load_info(StreamId stream, int64_t index_id) {
    std::vector<int64_t> write_tablet_ids;
    auto it = _index_streams_map.find(index_id);
    if (it != _index_streams_map.end()) {
        it->second->get_all_write_tablet_ids(&write_tablet_ids);
    }

    if (!write_tablet_ids.empty()) {
        butil::IOBuf buf;
        PLoadStreamResponse response;
        auto* tablet_load_infos = response.mutable_tablet_load_rowset_num_infos();
        _collect_tablet_load_info_from_tablets(write_tablet_ids, tablet_load_infos);
        if (tablet_load_infos->empty()) {
            return;
        }
        buf.append(response.SerializeAsString());
        auto wst = _write_stream(stream, buf);
        if (!wst.ok()) {
            LOG(WARNING) << "report tablet load info failed with " << wst << ", " << *this;
        }
    }
}

void LoadStream::_collect_tablet_load_info_from_tablets(
        const std::vector<int64_t>& tablet_ids,
        google::protobuf::RepeatedPtrField<PTabletLoadRowsetInfo>* tablet_load_infos) {
    for (auto tablet_id : tablet_ids) {
        BaseTabletSPtr tablet;
        if (auto res = ExecEnv::get_tablet(tablet_id); res.has_value()) {
            tablet = std::move(res).value();
        } else {
            continue;
        }
        BaseDeltaWriter::collect_tablet_load_rowset_num_info(tablet.get(), tablet_load_infos);
    }
}

Status LoadStream::_write_stream(StreamId stream, butil::IOBuf& buf) {
    for (;;) {
        int ret = 0;
        DBUG_EXECUTE_IF("LoadStream._write_stream.EAGAIN", { ret = EAGAIN; });
        if (ret == 0) {
            ret = brpc::StreamWrite(stream, buf);
        }
        switch (ret) {
        case 0:
            return Status::OK();
        case EAGAIN: {
            const timespec time = butil::seconds_from_now(config::load_stream_eagain_wait_seconds);
            int wait_ret = brpc::StreamWait(stream, &time);
            if (wait_ret != 0) {
                return Status::InternalError("StreamWait failed, err={}", wait_ret);
            }
            break;
        }
        default:
            return Status::InternalError("StreamWrite failed, err={}", ret);
        }
    }
    return Status::OK();
}

void LoadStream::_parse_header(butil::IOBuf* const message, PStreamHeader& hdr) {
    butil::IOBufAsZeroCopyInputStream wrapper(*message);
    hdr.ParseFromZeroCopyStream(&wrapper);
    VLOG_DEBUG << "header parse result: " << hdr.DebugString();
}

Status LoadStream::_append_data(const PStreamHeader& header, butil::IOBuf* data) {
    SCOPED_TIMER(_append_data_timer);
    IndexStreamSharedPtr index_stream;

    int64_t index_id = header.index_id();
    DBUG_EXECUTE_IF("TabletStream._append_data.unknown_indexid",
                    { index_id = UNKNOWN_ID_FOR_TEST; });
    auto it = _index_streams_map.find(index_id);
    if (it == _index_streams_map.end()) {
        return Status::Error<ErrorCode::INVALID_ARGUMENT>("unknown index_id {}", index_id);
    } else {
        index_stream = it->second;
    }

    return index_stream->append_data(header, data);
}

int LoadStream::on_received_messages(StreamId id, butil::IOBuf* const messages[], size_t size) {
    VLOG_DEBUG << "on_received_messages " << id << " " << size;
    for (size_t i = 0; i < size; ++i) {
        while (messages[i]->size() > 0) {
            // step 1: parse header
            size_t hdr_len = 0;
            messages[i]->cutn((void*)&hdr_len, sizeof(size_t));
            butil::IOBuf hdr_buf;
            PStreamHeader hdr;
            messages[i]->cutn(&hdr_buf, hdr_len);
            _parse_header(&hdr_buf, hdr);

            // step 2: cut data
            size_t data_len = 0;
            messages[i]->cutn((void*)&data_len, sizeof(size_t));
            butil::IOBuf data_buf;
            PStreamHeader data;
            messages[i]->cutn(&data_buf, data_len);

            // step 3: dispatch
            _dispatch(id, hdr, &data_buf);
        }
    }
    return 0;
}

void LoadStream::_dispatch(StreamId id, const PStreamHeader& hdr, butil::IOBuf* data) {
    VLOG_DEBUG << PStreamHeader_Opcode_Name(hdr.opcode()) << " from " << hdr.src_id()
               << " with tablet " << hdr.tablet_id();
    SCOPED_ATTACH_TASK(_resource_ctx);
    // CLOSE_LOAD message should not be fault injected,
    // otherwise the message will be ignored and causing close wait timeout
    if (hdr.opcode() != PStreamHeader::CLOSE_LOAD) {
        DBUG_EXECUTE_IF("LoadStream._dispatch.unknown_loadid", {
            PStreamHeader& t_hdr = const_cast<PStreamHeader&>(hdr);
            PUniqueId* load_id = t_hdr.mutable_load_id();
            load_id->set_hi(UNKNOWN_ID_FOR_TEST);
            load_id->set_lo(UNKNOWN_ID_FOR_TEST);
        });
        DBUG_EXECUTE_IF("LoadStream._dispatch.unknown_srcid", {
            PStreamHeader& t_hdr = const_cast<PStreamHeader&>(hdr);
            t_hdr.set_src_id(UNKNOWN_ID_FOR_TEST);
        });
    }
    if (UniqueId(hdr.load_id()) != UniqueId(_load_id)) {
        Status st = Status::Error<ErrorCode::INVALID_ARGUMENT>(
                "invalid load id {}, expected {}", print_id(hdr.load_id()), print_id(_load_id));
        _report_failure(id, st, hdr);
        return;
    }

    {
        std::lock_guard lock_guard(_lock);
        if (!_open_streams.contains(hdr.src_id())) {
            Status st = Status::Error<ErrorCode::INVALID_ARGUMENT>("no open stream from source {}",
                                                                   hdr.src_id());
            _report_failure(id, st, hdr);
            return;
        }
    }

    switch (hdr.opcode()) {
    case PStreamHeader::ADD_SEGMENT: {
        auto st = _append_data(hdr, data);
        if (!st.ok()) {
            _report_failure(id, st, hdr);
        } else {
            // Report tablet load info only on ADD_SEGMENT to reduce frequency.
            // ADD_SEGMENT is sent once per segment, while APPEND_DATA is sent
            // for every data batch. This reduces unnecessary writes and avoids
            // potential stream write failures when the sender is closing.
            _report_tablet_load_info(id, hdr.index_id());
        }
    } break;
    case PStreamHeader::APPEND_DATA: {
        auto st = _append_data(hdr, data);
        if (!st.ok()) {
            _report_failure(id, st, hdr);
        }
    } break;
    case PStreamHeader::CLOSE_LOAD: {
        DBUG_EXECUTE_IF("LoadStream.close_load.block", DBUG_BLOCK);
        std::vector<int64_t> success_tablet_ids;
        FailedTablets failed_tablets;
        std::vector<PTabletID> tablets_to_commit(hdr.tablets().begin(), hdr.tablets().end());
        // Step 1: count this CLOSE_LOAD and, if this is the last one, commit. Under _lock.
        bool all_received =
                close(hdr.src_id(), tablets_to_commit, &success_tablet_ids, &failed_tablets);
        // Step 2: send THIS stream's EOS (network IO, must be outside _lock). A stream
        // must not be StreamClose'd before its own EOS is delivered, otherwise the
        // sender sees on_closed without EOS and reports "Stream closed without EOS".
        _report_result(id, Status::OK(), success_tablet_ids, failed_tablets, true);
        bool is_incremental =
                hdr.has_num_incremental_streams() && hdr.num_incremental_streams() > 0;
        // Test-only: delay every incremental stream except the one that made
        // all_received, so a non-last incremental stream parks after the last
        // stream drained the list. On the buggy code this orphans it and the
        // load hangs; on the fix the latch drains the late registration under
        // the same lock. Inert unless enable_debug_points=true.
        if (is_incremental && !all_received) {
            DBUG_EXECUTE_IF("LoadStream.close_load.delay_incremental_register",
                            { bthread_usleep(3000000); });
        }
        // Step 3: close this stream (non-incremental) or park it for deferred close
        // (incremental, fencing), then collect everything that is now safe to close.
        // Registration happens only after step 2, so a collected stream already had
        // its EOS delivered (fixes the close-before-EOS race); the all-received latch
        // inside makes any late thread drain the parked streams (fixes the leak race).
        auto streams_to_close = mark_eos_sent_and_collect(id, is_incremental);
        for (auto& closing_id : streams_to_close) {
            brpc::StreamClose(closing_id);
        }
    } break;
    case PStreamHeader::GET_SCHEMA: {
        _report_schema(id, hdr);
    } break;
    default:
        LOG(WARNING) << "unexpected stream message " << hdr.opcode() << ", " << *this;
        DCHECK(false);
    }
}

void LoadStream::on_idle_timeout(StreamId id) {
    LOG(WARNING) << "closing load stream on idle timeout, " << *this;
    brpc::StreamClose(id);
}

void LoadStream::on_closed(StreamId id) {
    // `this` may be freed by other threads after increasing `_close_rpc_cnt`,
    // format string first to prevent use-after-free
    std::stringstream ss;
    ss << *this;
    auto remaining_streams = _total_streams - _close_rpc_cnt.fetch_add(1) - 1;
    LOG(INFO) << "stream " << id << " on_closed, remaining streams = " << remaining_streams << ", "
              << ss.str();
    if (remaining_streams == 0) {
        _load_stream_mgr->clear_load(_load_id);
    }
}

inline std::ostream& operator<<(std::ostream& ostr, const LoadStream& load_stream) {
    ostr << "load_id=" << print_id(load_stream._load_id) << ", txn_id=" << load_stream._txn_id;
    return ostr;
}

} // namespace doris

Coverage Report

Created: 2026-08-26 16:25