Coverage Report

Created: 2026-04-01 11:49

next uncovered line (L), next uncovered region (R), next uncovered branch (B)
be/src/exec/pipeline/pipeline_fragment_context.cpp
Line
Count
Source
1
// Licensed to the Apache Software Foundation (ASF) under one
2
// or more contributor license agreements.  See the NOTICE file
3
// distributed with this work for additional information
4
// regarding copyright ownership.  The ASF licenses this file
5
// to you under the Apache License, Version 2.0 (the
6
// "License"); you may not use this file except in compliance
7
// with the License.  You may obtain a copy of the License at
8
//
9
//   http://www.apache.org/licenses/LICENSE-2.0
10
//
11
// Unless required by applicable law or agreed to in writing,
12
// software distributed under the License is distributed on an
13
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
14
// KIND, either express or implied.  See the License for the
15
// specific language governing permissions and limitations
16
// under the License.
17
18
#include "exec/pipeline/pipeline_fragment_context.h"
19
20
#include <gen_cpp/DataSinks_types.h>
21
#include <gen_cpp/PaloInternalService_types.h>
22
#include <gen_cpp/PlanNodes_types.h>
23
#include <pthread.h>
24
25
#include <algorithm>
26
#include <cstdlib>
27
// IWYU pragma: no_include <bits/chrono.h>
28
#include <fmt/format.h>
29
30
#include <chrono> // IWYU pragma: keep
31
#include <map>
32
#include <memory>
33
#include <ostream>
34
#include <utility>
35
36
#include "cloud/config.h"
37
#include "common/cast_set.h"
38
#include "common/config.h"
39
#include "common/exception.h"
40
#include "common/logging.h"
41
#include "common/status.h"
42
#include "exec/exchange/local_exchange_sink_operator.h"
43
#include "exec/exchange/local_exchange_source_operator.h"
44
#include "exec/exchange/local_exchanger.h"
45
#include "exec/exchange/vdata_stream_mgr.h"
46
#include "exec/operator/aggregation_sink_operator.h"
47
#include "exec/operator/aggregation_source_operator.h"
48
#include "exec/operator/analytic_sink_operator.h"
49
#include "exec/operator/analytic_source_operator.h"
50
#include "exec/operator/assert_num_rows_operator.h"
51
#include "exec/operator/blackhole_sink_operator.h"
52
#include "exec/operator/cache_sink_operator.h"
53
#include "exec/operator/cache_source_operator.h"
54
#include "exec/operator/datagen_operator.h"
55
#include "exec/operator/dict_sink_operator.h"
56
#include "exec/operator/distinct_streaming_aggregation_operator.h"
57
#include "exec/operator/empty_set_operator.h"
58
#include "exec/operator/es_scan_operator.h"
59
#include "exec/operator/exchange_sink_operator.h"
60
#include "exec/operator/exchange_source_operator.h"
61
#include "exec/operator/file_scan_operator.h"
62
#include "exec/operator/group_commit_block_sink_operator.h"
63
#include "exec/operator/group_commit_scan_operator.h"
64
#include "exec/operator/hashjoin_build_sink.h"
65
#include "exec/operator/hashjoin_probe_operator.h"
66
#include "exec/operator/hive_table_sink_operator.h"
67
#include "exec/operator/iceberg_delete_sink_operator.h"
68
#include "exec/operator/iceberg_merge_sink_operator.h"
69
#include "exec/operator/iceberg_table_sink_operator.h"
70
#include "exec/operator/jdbc_scan_operator.h"
71
#include "exec/operator/jdbc_table_sink_operator.h"
72
#include "exec/operator/local_merge_sort_source_operator.h"
73
#include "exec/operator/materialization_opertor.h"
74
#include "exec/operator/maxcompute_table_sink_operator.h"
75
#include "exec/operator/memory_scratch_sink_operator.h"
76
#include "exec/operator/meta_scan_operator.h"
77
#include "exec/operator/multi_cast_data_stream_sink.h"
78
#include "exec/operator/multi_cast_data_stream_source.h"
79
#include "exec/operator/nested_loop_join_build_operator.h"
80
#include "exec/operator/nested_loop_join_probe_operator.h"
81
#include "exec/operator/olap_scan_operator.h"
82
#include "exec/operator/olap_table_sink_operator.h"
83
#include "exec/operator/olap_table_sink_v2_operator.h"
84
#include "exec/operator/partition_sort_sink_operator.h"
85
#include "exec/operator/partition_sort_source_operator.h"
86
#include "exec/operator/partitioned_aggregation_sink_operator.h"
87
#include "exec/operator/partitioned_aggregation_source_operator.h"
88
#include "exec/operator/partitioned_hash_join_probe_operator.h"
89
#include "exec/operator/partitioned_hash_join_sink_operator.h"
90
#include "exec/operator/rec_cte_anchor_sink_operator.h"
91
#include "exec/operator/rec_cte_scan_operator.h"
92
#include "exec/operator/rec_cte_sink_operator.h"
93
#include "exec/operator/rec_cte_source_operator.h"
94
#include "exec/operator/repeat_operator.h"
95
#include "exec/operator/result_file_sink_operator.h"
96
#include "exec/operator/result_sink_operator.h"
97
#include "exec/operator/schema_scan_operator.h"
98
#include "exec/operator/select_operator.h"
99
#include "exec/operator/set_probe_sink_operator.h"
100
#include "exec/operator/set_sink_operator.h"
101
#include "exec/operator/set_source_operator.h"
102
#include "exec/operator/sort_sink_operator.h"
103
#include "exec/operator/sort_source_operator.h"
104
#include "exec/operator/spill_iceberg_table_sink_operator.h"
105
#include "exec/operator/spill_sort_sink_operator.h"
106
#include "exec/operator/spill_sort_source_operator.h"
107
#include "exec/operator/streaming_aggregation_operator.h"
108
#include "exec/operator/table_function_operator.h"
109
#include "exec/operator/tvf_table_sink_operator.h"
110
#include "exec/operator/union_sink_operator.h"
111
#include "exec/operator/union_source_operator.h"
112
#include "exec/pipeline/dependency.h"
113
#include "exec/pipeline/pipeline_task.h"
114
#include "exec/pipeline/task_scheduler.h"
115
#include "exec/runtime_filter/runtime_filter_mgr.h"
116
#include "exec/sort/topn_sorter.h"
117
#include "exec/spill/spill_file.h"
118
#include "io/fs/stream_load_pipe.h"
119
#include "load/stream_load/new_load_stream_mgr.h"
120
#include "runtime/exec_env.h"
121
#include "runtime/fragment_mgr.h"
122
#include "runtime/result_buffer_mgr.h"
123
#include "runtime/runtime_state.h"
124
#include "runtime/thread_context.h"
125
#include "util/countdown_latch.h"
126
#include "util/debug_util.h"
127
#include "util/uid_util.h"
128
129
namespace doris {
130
#include "common/compile_check_begin.h"
131
PipelineFragmentContext::PipelineFragmentContext(
132
        TUniqueId query_id, const TPipelineFragmentParams& request,
133
        std::shared_ptr<QueryContext> query_ctx, ExecEnv* exec_env,
134
        const std::function<void(RuntimeState*, Status*)>& call_back,
135
        report_status_callback report_status_cb)
136
321k
        : _query_id(std::move(query_id)),
137
321k
          _fragment_id(request.fragment_id),
138
321k
          _exec_env(exec_env),
139
321k
          _query_ctx(std::move(query_ctx)),
140
321k
          _call_back(call_back),
141
321k
          _is_report_on_cancel(true),
142
321k
          _report_status_cb(std::move(report_status_cb)),
143
321k
          _params(request),
144
321k
          _parallel_instances(_params.__isset.parallel_instances ? _params.parallel_instances : 0),
145
321k
          _need_notify_close(request.__isset.need_notify_close ? request.need_notify_close
146
321k
                                                               : false) {
147
321k
    _fragment_watcher.start();
148
321k
}
149
150
321k
PipelineFragmentContext::~PipelineFragmentContext() {
151
321k
    LOG_INFO("PipelineFragmentContext::~PipelineFragmentContext")
152
321k
            .tag("query_id", print_id(_query_id))
153
321k
            .tag("fragment_id", _fragment_id);
154
321k
    _release_resource();
155
321k
    {
156
        // The memory released by the query end is recorded in the query mem tracker.
157
321k
        SCOPED_SWITCH_THREAD_MEM_TRACKER_LIMITER(_query_ctx->query_mem_tracker());
158
321k
        _runtime_state.reset();
159
321k
        _query_ctx.reset();
160
321k
    }
161
321k
}
162
163
66
bool PipelineFragmentContext::is_timeout(timespec now) const {
164
66
    if (_timeout <= 0) {
165
0
        return false;
166
0
    }
167
66
    return _fragment_watcher.elapsed_time_seconds(now) > _timeout;
168
66
}
169
170
// notify_close() transitions the PFC from "waiting for external close notification" to
171
// "self-managed close". For recursive CTE fragments, the old PFC is kept alive until
172
// the rerun_fragment(wait_for_destroy) RPC calls this to trigger shutdown.
173
// Returns true if all tasks have already closed (i.e., the PFC can be safely destroyed).
174
8.93k
bool PipelineFragmentContext::notify_close() {
175
8.93k
    bool all_closed = false;
176
8.93k
    bool need_remove = false;
177
8.93k
    {
178
8.93k
        std::lock_guard<std::mutex> l(_task_mutex);
179
8.93k
        if (_closed_tasks >= _total_tasks) {
180
3.44k
            if (_need_notify_close) {
181
                // Fragment was cancelled and waiting for notify to close.
182
                // Record that we need to remove from fragment mgr, but do it
183
                // after releasing _task_mutex to avoid ABBA deadlock with
184
                // dump_pipeline_tasks() (which acquires _pipeline_map lock
185
                // first, then _task_mutex via debug_string()).
186
3.41k
                need_remove = true;
187
3.41k
            }
188
3.44k
            all_closed = true;
189
3.44k
        }
190
        // make fragment release by self after cancel
191
8.93k
        _need_notify_close = false;
192
8.93k
    }
193
8.93k
    if (need_remove) {
194
3.41k
        _exec_env->fragment_mgr()->remove_pipeline_context({_query_id, _fragment_id});
195
3.41k
    }
196
8.93k
    return all_closed;
197
8.93k
}
198
199
// Must not add lock in this method. Because it will call query ctx cancel. And
200
// QueryCtx cancel will call fragment ctx cancel. And Also Fragment ctx's running
201
// Method like exchange sink buffer will call query ctx cancel. If we add lock here
202
// There maybe dead lock.
203
5.46k
void PipelineFragmentContext::cancel(const Status reason) {
204
5.46k
    LOG_INFO("PipelineFragmentContext::cancel")
205
5.46k
            .tag("query_id", print_id(_query_id))
206
5.46k
            .tag("fragment_id", _fragment_id)
207
5.46k
            .tag("reason", reason.to_string());
208
5.46k
    if (notify_close()) {
209
54
        return;
210
54
    }
211
    // Timeout is a special error code, we need print current stack to debug timeout issue.
212
5.40k
    if (reason.is<ErrorCode::TIMEOUT>()) {
213
8
        auto dbg_str = fmt::format("PipelineFragmentContext is cancelled due to timeout:\n{}",
214
8
                                   debug_string());
215
8
        LOG_LONG_STRING(WARNING, dbg_str);
216
8
    }
217
218
    // `ILLEGAL_STATE` means queries this fragment belongs to was not found in FE (maybe finished)
219
5.40k
    if (reason.is<ErrorCode::ILLEGAL_STATE>()) {
220
0
        LOG_WARNING("PipelineFragmentContext is cancelled due to illegal state : {}",
221
0
                    debug_string());
222
0
    }
223
224
5.40k
    if (reason.is<ErrorCode::MEM_LIMIT_EXCEEDED>() || reason.is<ErrorCode::MEM_ALLOC_FAILED>()) {
225
0
        print_profile("cancel pipeline, reason: " + reason.to_string());
226
0
    }
227
228
5.40k
    if (auto error_url = get_load_error_url(); !error_url.empty()) {
229
22
        _query_ctx->set_load_error_url(error_url);
230
22
    }
231
232
5.40k
    if (auto first_error_msg = get_first_error_msg(); !first_error_msg.empty()) {
233
22
        _query_ctx->set_first_error_msg(first_error_msg);
234
22
    }
235
236
5.40k
    _query_ctx->cancel(reason, _fragment_id);
237
5.40k
    if (reason.is<ErrorCode::LIMIT_REACH>()) {
238
199
        _is_report_on_cancel = false;
239
5.20k
    } else {
240
27.5k
        for (auto& id : _fragment_instance_ids) {
241
27.5k
            LOG(WARNING) << "PipelineFragmentContext cancel instance: " << print_id(id);
242
27.5k
        }
243
5.20k
    }
244
    // Get pipe from new load stream manager and send cancel to it or the fragment may hang to wait read from pipe
245
    // For stream load the fragment's query_id == load id, it is set in FE.
246
5.40k
    auto stream_load_ctx = _exec_env->new_load_stream_mgr()->get(_query_id);
247
5.40k
    if (stream_load_ctx != nullptr) {
248
30
        stream_load_ctx->pipe->cancel(reason.to_string());
249
        // Set error URL here because after pipe is cancelled, stream load execution may return early.
250
        // We need to set the error URL at this point to ensure error information is properly
251
        // propagated to the client.
252
30
        stream_load_ctx->error_url = get_load_error_url();
253
30
        stream_load_ctx->first_error_msg = get_first_error_msg();
254
30
    }
255
256
28.6k
    for (auto& tasks : _tasks) {
257
63.3k
        for (auto& task : tasks) {
258
63.3k
            task.first->unblock_all_dependencies();
259
63.3k
        }
260
28.6k
    }
261
5.40k
}
262
263
508k
PipelinePtr PipelineFragmentContext::add_pipeline(PipelinePtr parent, int idx) {
264
508k
    PipelineId id = _next_pipeline_id++;
265
508k
    auto pipeline = std::make_shared<Pipeline>(
266
508k
            id, parent ? std::min(parent->num_tasks(), _num_instances) : _num_instances,
267
508k
            parent ? parent->num_tasks() : _num_instances);
268
508k
    if (idx >= 0) {
269
101k
        _pipelines.insert(_pipelines.begin() + idx, pipeline);
270
407k
    } else {
271
407k
        _pipelines.emplace_back(pipeline);
272
407k
    }
273
508k
    if (parent) {
274
186k
        parent->set_children(pipeline);
275
186k
    }
276
508k
    return pipeline;
277
508k
}
278
279
320k
Status PipelineFragmentContext::_build_and_prepare_full_pipeline(ThreadPool* thread_pool) {
280
320k
    {
281
320k
        SCOPED_TIMER(_build_pipelines_timer);
282
        // 2. Build pipelines with operators in this fragment.
283
320k
        auto root_pipeline = add_pipeline();
284
320k
        RETURN_IF_ERROR(_build_pipelines(_runtime_state->obj_pool(), *_query_ctx->desc_tbl,
285
320k
                                         &_root_op, root_pipeline));
286
287
        // 3. Create sink operator
288
320k
        if (!_params.fragment.__isset.output_sink) {
289
0
            return Status::InternalError("No output sink in this fragment!");
290
0
        }
291
320k
        RETURN_IF_ERROR(_create_data_sink(_runtime_state->obj_pool(), _params.fragment.output_sink,
292
320k
                                          _params.fragment.output_exprs, _params,
293
320k
                                          root_pipeline->output_row_desc(), _runtime_state.get(),
294
320k
                                          *_desc_tbl, root_pipeline->id()));
295
320k
        RETURN_IF_ERROR(_sink->init(_params.fragment.output_sink));
296
320k
        RETURN_IF_ERROR(root_pipeline->set_sink(_sink));
297
298
407k
        for (PipelinePtr& pipeline : _pipelines) {
299
18.4E
            DCHECK(pipeline->sink() != nullptr) << pipeline->operators().size();
300
407k
            RETURN_IF_ERROR(pipeline->sink()->set_child(pipeline->operators().back()));
301
407k
        }
302
320k
    }
303
    // 4. Build local exchanger
304
320k
    if (_runtime_state->enable_local_shuffle()) {
305
318k
        SCOPED_TIMER(_plan_local_exchanger_timer);
306
318k
        RETURN_IF_ERROR(_plan_local_exchange(_params.num_buckets,
307
318k
                                             _params.bucket_seq_to_instance_idx,
308
318k
                                             _params.shuffle_idx_to_instance_idx));
309
318k
    }
310
311
    // 5. Initialize global states in pipelines.
312
509k
    for (PipelinePtr& pipeline : _pipelines) {
313
509k
        SCOPED_TIMER(_prepare_all_pipelines_timer);
314
509k
        pipeline->children().clear();
315
509k
        RETURN_IF_ERROR(pipeline->prepare(_runtime_state.get()));
316
509k
    }
317
318
319k
    {
319
319k
        SCOPED_TIMER(_build_tasks_timer);
320
        // 6. Build pipeline tasks and initialize local state.
321
319k
        RETURN_IF_ERROR(_build_pipeline_tasks(thread_pool));
322
319k
    }
323
324
319k
    return Status::OK();
325
319k
}
326
327
321k
Status PipelineFragmentContext::prepare(ThreadPool* thread_pool) {
328
321k
    if (_prepared) {
329
0
        return Status::InternalError("Already prepared");
330
0
    }
331
321k
    if (_params.__isset.query_options && _params.query_options.__isset.execution_timeout) {
332
321k
        _timeout = _params.query_options.execution_timeout;
333
321k
    }
334
335
321k
    _fragment_level_profile = std::make_unique<RuntimeProfile>("PipelineContext");
336
321k
    _prepare_timer = ADD_TIMER(_fragment_level_profile, "PrepareTime");
337
321k
    SCOPED_TIMER(_prepare_timer);
338
321k
    _build_pipelines_timer = ADD_TIMER(_fragment_level_profile, "BuildPipelinesTime");
339
321k
    _init_context_timer = ADD_TIMER(_fragment_level_profile, "InitContextTime");
340
321k
    _plan_local_exchanger_timer = ADD_TIMER(_fragment_level_profile, "PlanLocalLocalExchangerTime");
341
321k
    _build_tasks_timer = ADD_TIMER(_fragment_level_profile, "BuildTasksTime");
342
321k
    _prepare_all_pipelines_timer = ADD_TIMER(_fragment_level_profile, "PrepareAllPipelinesTime");
343
321k
    {
344
321k
        SCOPED_TIMER(_init_context_timer);
345
321k
        cast_set(_num_instances, _params.local_params.size());
346
321k
        _total_instances =
347
321k
                _params.__isset.total_instances ? _params.total_instances : _num_instances;
348
349
321k
        auto* fragment_context = this;
350
351
321k
        if (_params.query_options.__isset.is_report_success) {
352
319k
            fragment_context->set_is_report_success(_params.query_options.is_report_success);
353
319k
        }
354
355
        // 1. Set up the global runtime state.
356
321k
        _runtime_state = RuntimeState::create_unique(
357
321k
                _params.query_id, _params.fragment_id, _params.query_options,
358
321k
                _query_ctx->query_globals, _exec_env, _query_ctx.get());
359
321k
        _runtime_state->set_task_execution_context(shared_from_this());
360
321k
        SCOPED_SWITCH_THREAD_MEM_TRACKER_LIMITER(_runtime_state->query_mem_tracker());
361
321k
        if (_params.__isset.backend_id) {
362
319k
            _runtime_state->set_backend_id(_params.backend_id);
363
319k
        }
364
321k
        if (_params.__isset.import_label) {
365
239
            _runtime_state->set_import_label(_params.import_label);
366
239
        }
367
321k
        if (_params.__isset.db_name) {
368
191
            _runtime_state->set_db_name(_params.db_name);
369
191
        }
370
321k
        if (_params.__isset.load_job_id) {
371
0
            _runtime_state->set_load_job_id(_params.load_job_id);
372
0
        }
373
374
321k
        if (_params.is_simplified_param) {
375
110k
            _desc_tbl = _query_ctx->desc_tbl;
376
210k
        } else {
377
210k
            DCHECK(_params.__isset.desc_tbl);
378
210k
            RETURN_IF_ERROR(DescriptorTbl::create(_runtime_state->obj_pool(), _params.desc_tbl,
379
210k
                                                  &_desc_tbl));
380
210k
        }
381
321k
        _runtime_state->set_desc_tbl(_desc_tbl);
382
321k
        _runtime_state->set_num_per_fragment_instances(_params.num_senders);
383
321k
        _runtime_state->set_load_stream_per_node(_params.load_stream_per_node);
384
321k
        _runtime_state->set_total_load_streams(_params.total_load_streams);
385
321k
        _runtime_state->set_num_local_sink(_params.num_local_sink);
386
387
        // init fragment_instance_ids
388
321k
        const auto target_size = _params.local_params.size();
389
321k
        _fragment_instance_ids.resize(target_size);
390
1.31M
        for (size_t i = 0; i < _params.local_params.size(); i++) {
391
991k
            auto fragment_instance_id = _params.local_params[i].fragment_instance_id;
392
991k
            _fragment_instance_ids[i] = fragment_instance_id;
393
991k
        }
394
321k
    }
395
396
321k
    RETURN_IF_ERROR(_build_and_prepare_full_pipeline(thread_pool));
397
398
320k
    _init_next_report_time();
399
400
320k
    _prepared = true;
401
320k
    return Status::OK();
402
321k
}
403
404
Status PipelineFragmentContext::_build_pipeline_tasks_for_instance(
405
        int instance_idx,
406
990k
        const std::vector<std::shared_ptr<RuntimeProfile>>& pipeline_id_to_profile) {
407
990k
    const auto& local_params = _params.local_params[instance_idx];
408
990k
    auto fragment_instance_id = local_params.fragment_instance_id;
409
990k
    auto runtime_filter_mgr = std::make_unique<RuntimeFilterMgr>(false);
410
990k
    std::map<PipelineId, PipelineTask*> pipeline_id_to_task;
411
990k
    auto get_shared_state = [&](PipelinePtr pipeline)
412
990k
            -> std::map<int, std::pair<std::shared_ptr<BasicSharedState>,
413
1.71M
                                       std::vector<std::shared_ptr<Dependency>>>> {
414
1.71M
        std::map<int, std::pair<std::shared_ptr<BasicSharedState>,
415
1.71M
                                std::vector<std::shared_ptr<Dependency>>>>
416
1.71M
                shared_state_map;
417
2.29M
        for (auto& op : pipeline->operators()) {
418
2.29M
            auto source_id = op->operator_id();
419
2.29M
            if (auto iter = _op_id_to_shared_state.find(source_id);
420
2.29M
                iter != _op_id_to_shared_state.end()) {
421
700k
                shared_state_map.insert({source_id, iter->second});
422
700k
            }
423
2.29M
        }
424
1.71M
        for (auto sink_to_source_id : pipeline->sink()->dests_id()) {
425
1.71M
            if (auto iter = _op_id_to_shared_state.find(sink_to_source_id);
426
1.71M
                iter != _op_id_to_shared_state.end()) {
427
336k
                shared_state_map.insert({sink_to_source_id, iter->second});
428
336k
            }
429
1.71M
        }
430
1.71M
        return shared_state_map;
431
1.71M
    };
432
433
3.06M
    for (size_t pip_idx = 0; pip_idx < _pipelines.size(); pip_idx++) {
434
2.07M
        auto& pipeline = _pipelines[pip_idx];
435
2.07M
        if (pipeline->num_tasks() > 1 || instance_idx == 0) {
436
1.71M
            auto task_runtime_state = RuntimeState::create_unique(
437
1.71M
                    local_params.fragment_instance_id, _params.query_id, _params.fragment_id,
438
1.71M
                    _params.query_options, _query_ctx->query_globals, _exec_env, _query_ctx.get());
439
1.71M
            {
440
                // Initialize runtime state for this task
441
1.71M
                task_runtime_state->set_query_mem_tracker(_query_ctx->query_mem_tracker());
442
443
1.71M
                task_runtime_state->set_task_execution_context(shared_from_this());
444
1.71M
                task_runtime_state->set_be_number(local_params.backend_num);
445
446
1.71M
                if (_params.__isset.backend_id) {
447
1.71M
                    task_runtime_state->set_backend_id(_params.backend_id);
448
1.71M
                }
449
1.71M
                if (_params.__isset.import_label) {
450
240
                    task_runtime_state->set_import_label(_params.import_label);
451
240
                }
452
1.71M
                if (_params.__isset.db_name) {
453
192
                    task_runtime_state->set_db_name(_params.db_name);
454
192
                }
455
1.71M
                if (_params.__isset.load_job_id) {
456
0
                    task_runtime_state->set_load_job_id(_params.load_job_id);
457
0
                }
458
1.71M
                if (_params.__isset.wal_id) {
459
114
                    task_runtime_state->set_wal_id(_params.wal_id);
460
114
                }
461
1.71M
                if (_params.__isset.content_length) {
462
31
                    task_runtime_state->set_content_length(_params.content_length);
463
31
                }
464
465
1.71M
                task_runtime_state->set_desc_tbl(_desc_tbl);
466
1.71M
                task_runtime_state->set_per_fragment_instance_idx(local_params.sender_id);
467
1.71M
                task_runtime_state->set_num_per_fragment_instances(_params.num_senders);
468
1.71M
                task_runtime_state->resize_op_id_to_local_state(max_operator_id());
469
1.71M
                task_runtime_state->set_max_operator_id(max_operator_id());
470
1.71M
                task_runtime_state->set_load_stream_per_node(_params.load_stream_per_node);
471
1.71M
                task_runtime_state->set_total_load_streams(_params.total_load_streams);
472
1.71M
                task_runtime_state->set_num_local_sink(_params.num_local_sink);
473
474
1.71M
                task_runtime_state->set_runtime_filter_mgr(runtime_filter_mgr.get());
475
1.71M
            }
476
1.71M
            auto cur_task_id = _total_tasks++;
477
1.71M
            task_runtime_state->set_task_id(cur_task_id);
478
1.71M
            task_runtime_state->set_task_num(pipeline->num_tasks());
479
1.71M
            auto task = std::make_shared<PipelineTask>(
480
1.71M
                    pipeline, cur_task_id, task_runtime_state.get(),
481
1.71M
                    std::dynamic_pointer_cast<PipelineFragmentContext>(shared_from_this()),
482
1.71M
                    pipeline_id_to_profile[pip_idx].get(), get_shared_state(pipeline),
483
1.71M
                    instance_idx);
484
1.71M
            pipeline->incr_created_tasks(instance_idx, task.get());
485
1.71M
            pipeline_id_to_task.insert({pipeline->id(), task.get()});
486
1.71M
            _tasks[instance_idx].emplace_back(
487
1.71M
                    std::pair<std::shared_ptr<PipelineTask>, std::unique_ptr<RuntimeState>> {
488
1.71M
                            std::move(task), std::move(task_runtime_state)});
489
1.71M
        }
490
2.07M
    }
491
492
    /**
493
         * Build DAG for pipeline tasks.
494
         * For example, we have
495
         *
496
         *   ExchangeSink (Pipeline1)     JoinBuildSink (Pipeline2)
497
         *            \                      /
498
         *          JoinProbeOperator1 (Pipeline1)    JoinBuildSink (Pipeline3)
499
         *                 \                          /
500
         *               JoinProbeOperator2 (Pipeline1)
501
         *
502
         * In this fragment, we have three pipelines and pipeline 1 depends on pipeline 2 and pipeline 3.
503
         * To build this DAG, `_dag` manage dependencies between pipelines by pipeline ID and
504
         * `pipeline_id_to_task` is used to find the task by a unique pipeline ID.
505
         *
506
         * Finally, we have two upstream dependencies in Pipeline1 corresponding to JoinProbeOperator1
507
         * and JoinProbeOperator2.
508
         */
509
2.07M
    for (auto& _pipeline : _pipelines) {
510
2.07M
        if (pipeline_id_to_task.contains(_pipeline->id())) {
511
1.71M
            auto* task = pipeline_id_to_task[_pipeline->id()];
512
1.71M
            DCHECK(task != nullptr);
513
514
            // If this task has upstream dependency, then inject it into this task.
515
1.71M
            if (_dag.contains(_pipeline->id())) {
516
1.09M
                auto& deps = _dag[_pipeline->id()];
517
1.77M
                for (auto& dep : deps) {
518
1.77M
                    if (pipeline_id_to_task.contains(dep)) {
519
1.04M
                        auto ss = pipeline_id_to_task[dep]->get_sink_shared_state();
520
1.04M
                        if (ss) {
521
377k
                            task->inject_shared_state(ss);
522
667k
                        } else {
523
667k
                            pipeline_id_to_task[dep]->inject_shared_state(
524
667k
                                    task->get_source_shared_state());
525
667k
                        }
526
1.04M
                    }
527
1.77M
                }
528
1.09M
            }
529
1.71M
        }
530
2.07M
    }
531
3.06M
    for (size_t pip_idx = 0; pip_idx < _pipelines.size(); pip_idx++) {
532
2.07M
        if (pipeline_id_to_task.contains(_pipelines[pip_idx]->id())) {
533
1.71M
            auto* task = pipeline_id_to_task[_pipelines[pip_idx]->id()];
534
1.71M
            DCHECK(pipeline_id_to_profile[pip_idx]);
535
1.71M
            std::vector<TScanRangeParams> scan_ranges;
536
1.71M
            auto node_id = _pipelines[pip_idx]->operators().front()->node_id();
537
1.71M
            if (local_params.per_node_scan_ranges.contains(node_id)) {
538
271k
                scan_ranges = local_params.per_node_scan_ranges.find(node_id)->second;
539
271k
            }
540
1.71M
            RETURN_IF_ERROR_OR_CATCH_EXCEPTION(task->prepare(scan_ranges, local_params.sender_id,
541
1.71M
                                                             _params.fragment.output_sink));
542
1.71M
        }
543
2.07M
    }
544
992k
    {
545
992k
        std::lock_guard<std::mutex> l(_state_map_lock);
546
992k
        _runtime_filter_mgr_map[instance_idx] = std::move(runtime_filter_mgr);
547
992k
    }
548
992k
    return Status::OK();
549
990k
}
550
551
320k
Status PipelineFragmentContext::_build_pipeline_tasks(ThreadPool* thread_pool) {
552
320k
    _total_tasks = 0;
553
320k
    _closed_tasks = 0;
554
320k
    const auto target_size = _params.local_params.size();
555
320k
    _tasks.resize(target_size);
556
320k
    _runtime_filter_mgr_map.resize(target_size);
557
828k
    for (size_t pip_idx = 0; pip_idx < _pipelines.size(); pip_idx++) {
558
508k
        _pip_id_to_pipeline[_pipelines[pip_idx]->id()] = _pipelines[pip_idx].get();
559
508k
    }
560
320k
    auto pipeline_id_to_profile = _runtime_state->build_pipeline_profile(_pipelines.size());
561
562
320k
    if (target_size > 1 &&
563
320k
        (_runtime_state->query_options().__isset.parallel_prepare_threshold &&
564
122k
         target_size > _runtime_state->query_options().parallel_prepare_threshold)) {
565
        // If instances parallelism is big enough ( > parallel_prepare_threshold), we will prepare all tasks by multi-threads
566
17.1k
        std::vector<Status> prepare_status(target_size);
567
17.1k
        int submitted_tasks = 0;
568
17.1k
        Status submit_status;
569
17.1k
        CountDownLatch latch((int)target_size);
570
235k
        for (int i = 0; i < target_size; i++) {
571
218k
            submit_status = thread_pool->submit_func([&, i]() {
572
218k
                SCOPED_ATTACH_TASK(_query_ctx.get());
573
218k
                prepare_status[i] = _build_pipeline_tasks_for_instance(i, pipeline_id_to_profile);
574
218k
                latch.count_down();
575
218k
            });
576
218k
            if (LIKELY(submit_status.ok())) {
577
218k
                submitted_tasks++;
578
18.4E
            } else {
579
18.4E
                break;
580
18.4E
            }
581
218k
        }
582
17.1k
        latch.arrive_and_wait(target_size - submitted_tasks);
583
17.1k
        if (UNLIKELY(!submit_status.ok())) {
584
0
            return submit_status;
585
0
        }
586
235k
        for (int i = 0; i < submitted_tasks; i++) {
587
218k
            if (!prepare_status[i].ok()) {
588
0
                return prepare_status[i];
589
0
            }
590
218k
        }
591
303k
    } else {
592
1.07M
        for (int i = 0; i < target_size; i++) {
593
773k
            RETURN_IF_ERROR(_build_pipeline_tasks_for_instance(i, pipeline_id_to_profile));
594
773k
        }
595
303k
    }
596
320k
    _pipeline_parent_map.clear();
597
320k
    _op_id_to_shared_state.clear();
598
599
320k
    return Status::OK();
600
320k
}
601
602
319k
void PipelineFragmentContext::_init_next_report_time() {
603
319k
    auto interval_s = config::pipeline_status_report_interval;
604
319k
    if (_is_report_success && interval_s > 0 && _timeout > interval_s) {
605
31.7k
        VLOG_FILE << "enable period report: fragment id=" << _fragment_id;
606
31.7k
        uint64_t report_fragment_offset = (uint64_t)(rand() % interval_s) * NANOS_PER_SEC;
607
        // We don't want to wait longer than it takes to run the entire fragment.
608
31.7k
        _previous_report_time =
609
31.7k
                MonotonicNanos() + report_fragment_offset - (uint64_t)(interval_s)*NANOS_PER_SEC;
610
31.7k
        _disable_period_report = false;
611
31.7k
    }
612
319k
}
613
614
3.86k
void PipelineFragmentContext::refresh_next_report_time() {
615
3.86k
    auto disable = _disable_period_report.load(std::memory_order_acquire);
616
3.86k
    DCHECK(disable == true);
617
3.86k
    _previous_report_time.store(MonotonicNanos(), std::memory_order_release);
618
3.86k
    _disable_period_report.compare_exchange_strong(disable, false);
619
3.86k
}
620
621
5.82M
void PipelineFragmentContext::trigger_report_if_necessary() {
622
5.82M
    if (!_is_report_success) {
623
5.40M
        return;
624
5.40M
    }
625
414k
    auto disable = _disable_period_report.load(std::memory_order_acquire);
626
414k
    if (disable) {
627
8.72k
        return;
628
8.72k
    }
629
406k
    int32_t interval_s = config::pipeline_status_report_interval;
630
406k
    if (interval_s <= 0) {
631
0
        LOG(WARNING) << "config::status_report_interval is equal to or less than zero, do not "
632
0
                        "trigger "
633
0
                        "report.";
634
0
    }
635
406k
    uint64_t next_report_time = _previous_report_time.load(std::memory_order_acquire) +
636
406k
                                (uint64_t)(interval_s)*NANOS_PER_SEC;
637
406k
    if (MonotonicNanos() > next_report_time) {
638
3.86k
        if (!_disable_period_report.compare_exchange_strong(disable, true,
639
3.86k
                                                            std::memory_order_acq_rel)) {
640
3
            return;
641
3
        }
642
3.86k
        if (VLOG_FILE_IS_ON) {
643
0
            VLOG_FILE << "Reporting "
644
0
                      << "profile for query_id " << print_id(_query_id)
645
0
                      << ", fragment id: " << _fragment_id;
646
647
0
            std::stringstream ss;
648
0
            _runtime_state->runtime_profile()->compute_time_in_profile();
649
0
            _runtime_state->runtime_profile()->pretty_print(&ss);
650
0
            if (_runtime_state->load_channel_profile()) {
651
0
                _runtime_state->load_channel_profile()->pretty_print(&ss);
652
0
            }
653
654
0
            VLOG_FILE << "Query " << print_id(get_query_id()) << " fragment " << get_fragment_id()
655
0
                      << " profile:\n"
656
0
                      << ss.str();
657
0
        }
658
3.86k
        auto st = send_report(false);
659
3.86k
        if (!st.ok()) {
660
0
            disable = true;
661
0
            _disable_period_report.compare_exchange_strong(disable, false,
662
0
                                                           std::memory_order_acq_rel);
663
0
        }
664
3.86k
    }
665
406k
}
666
667
Status PipelineFragmentContext::_build_pipelines(ObjectPool* pool, const DescriptorTbl& descs,
668
320k
                                                 OperatorPtr* root, PipelinePtr cur_pipe) {
669
320k
    if (_params.fragment.plan.nodes.empty()) {
670
0
        throw Exception(ErrorCode::INTERNAL_ERROR, "Invalid plan which has no plan node!");
671
0
    }
672
673
320k
    int node_idx = 0;
674
675
320k
    RETURN_IF_ERROR(_create_tree_helper(pool, _params.fragment.plan.nodes, descs, nullptr,
676
320k
                                        &node_idx, root, cur_pipe, 0, false, false));
677
678
320k
    if (node_idx + 1 != _params.fragment.plan.nodes.size()) {
679
0
        return Status::InternalError(
680
0
                "Plan tree only partially reconstructed. Not all thrift nodes were used.");
681
0
    }
682
320k
    return Status::OK();
683
320k
}
684
685
Status PipelineFragmentContext::_create_tree_helper(
686
        ObjectPool* pool, const std::vector<TPlanNode>& tnodes, const DescriptorTbl& descs,
687
        OperatorPtr parent, int* node_idx, OperatorPtr* root, PipelinePtr& cur_pipe, int child_idx,
688
508k
        const bool followed_by_shuffled_operator, const bool require_bucket_distribution) {
689
    // propagate error case
690
508k
    if (*node_idx >= tnodes.size()) {
691
0
        return Status::InternalError(
692
0
                "Failed to reconstruct plan tree from thrift. Node id: {}, number of nodes: {}",
693
0
                *node_idx, tnodes.size());
694
0
    }
695
508k
    const TPlanNode& tnode = tnodes[*node_idx];
696
697
508k
    int num_children = tnodes[*node_idx].num_children;
698
508k
    bool current_followed_by_shuffled_operator = followed_by_shuffled_operator;
699
508k
    bool current_require_bucket_distribution = require_bucket_distribution;
700
    // TODO: Create CacheOperator is confused now
701
508k
    OperatorPtr op = nullptr;
702
508k
    OperatorPtr cache_op = nullptr;
703
508k
    RETURN_IF_ERROR(_create_operator(pool, tnodes[*node_idx], descs, op, cur_pipe,
704
508k
                                     parent == nullptr ? -1 : parent->node_id(), child_idx,
705
508k
                                     followed_by_shuffled_operator,
706
508k
                                     current_require_bucket_distribution, cache_op));
707
    // Initialization must be done here. For example, group by expressions in agg will be used to
708
    // decide if a local shuffle should be planed, so it must be initialized here.
709
508k
    RETURN_IF_ERROR(op->init(tnode, _runtime_state.get()));
710
    // assert(parent != nullptr || (node_idx == 0 && root_expr != nullptr));
711
508k
    if (parent != nullptr) {
712
        // add to parent's child(s)
713
188k
        RETURN_IF_ERROR(parent->set_child(cache_op ? cache_op : op));
714
319k
    } else {
715
319k
        *root = op;
716
319k
    }
717
    /**
718
     * `ExchangeType::HASH_SHUFFLE` should be used if an operator is followed by a shuffled operator (shuffled hash join, union operator followed by co-located operators).
719
     *
720
     * For plan:
721
     * LocalExchange(id=0) -> Aggregation(id=1) -> ShuffledHashJoin(id=2)
722
     *                           Exchange(id=3) -> ShuffledHashJoinBuild(id=2)
723
     * We must ensure data distribution of `LocalExchange(id=0)` is same as Exchange(id=3).
724
     *
725
     * If an operator's is followed by a local exchange without shuffle (e.g. passthrough), a
726
     * shuffled local exchanger will be used before join so it is not followed by shuffle join.
727
     */
728
508k
    auto required_data_distribution =
729
508k
            cur_pipe->operators().empty()
730
508k
                    ? cur_pipe->sink()->required_data_distribution(_runtime_state.get())
731
508k
                    : op->required_data_distribution(_runtime_state.get());
732
508k
    current_followed_by_shuffled_operator =
733
508k
            ((followed_by_shuffled_operator ||
734
508k
              (cur_pipe->operators().empty() ? cur_pipe->sink()->is_shuffled_operator()
735
451k
                                             : op->is_shuffled_operator())) &&
736
508k
             Pipeline::is_hash_exchange(required_data_distribution.distribution_type)) ||
737
508k
            (followed_by_shuffled_operator &&
738
400k
             required_data_distribution.distribution_type == ExchangeType::NOOP);
739
740
508k
    current_require_bucket_distribution =
741
508k
            ((require_bucket_distribution ||
742
508k
              (cur_pipe->operators().empty() ? cur_pipe->sink()->is_colocated_operator()
743
455k
                                             : op->is_colocated_operator())) &&
744
508k
             Pipeline::is_hash_exchange(required_data_distribution.distribution_type)) ||
745
508k
            (require_bucket_distribution &&
746
406k
             required_data_distribution.distribution_type == ExchangeType::NOOP);
747
748
508k
    if (num_children == 0) {
749
330k
        _use_serial_source = op->is_serial_operator();
750
330k
    }
751
    // rely on that tnodes is preorder of the plan
752
696k
    for (int i = 0; i < num_children; i++) {
753
188k
        ++*node_idx;
754
188k
        RETURN_IF_ERROR(_create_tree_helper(pool, tnodes, descs, op, node_idx, nullptr, cur_pipe, i,
755
188k
                                            current_followed_by_shuffled_operator,
756
188k
                                            current_require_bucket_distribution));
757
758
        // we are expecting a child, but have used all nodes
759
        // this means we have been given a bad tree and must fail
760
188k
        if (*node_idx >= tnodes.size()) {
761
0
            return Status::InternalError(
762
0
                    "Failed to reconstruct plan tree from thrift. Node id: {}, number of "
763
0
                    "nodes: {}",
764
0
                    *node_idx, tnodes.size());
765
0
        }
766
188k
    }
767
768
508k
    return Status::OK();
769
508k
}
770
771
void PipelineFragmentContext::_inherit_pipeline_properties(
772
        const DataDistribution& data_distribution, PipelinePtr pipe_with_source,
773
101k
        PipelinePtr pipe_with_sink) {
774
101k
    pipe_with_sink->set_num_tasks(pipe_with_source->num_tasks());
775
101k
    pipe_with_source->set_num_tasks(_num_instances);
776
101k
    pipe_with_source->set_data_distribution(data_distribution);
777
101k
}
778
779
Status PipelineFragmentContext::_add_local_exchange_impl(
780
        int idx, ObjectPool* pool, PipelinePtr cur_pipe, PipelinePtr new_pip,
781
        DataDistribution data_distribution, bool* do_local_exchange, int num_buckets,
782
        const std::map<int, int>& bucket_seq_to_instance_idx,
783
101k
        const std::map<int, int>& shuffle_idx_to_instance_idx) {
784
101k
    auto& operators = cur_pipe->operators();
785
101k
    const auto downstream_pipeline_id = cur_pipe->id();
786
101k
    auto local_exchange_id = next_operator_id();
787
    // 1. Create a new pipeline with local exchange sink.
788
101k
    DataSinkOperatorPtr sink;
789
101k
    auto sink_id = next_sink_operator_id();
790
791
    /**
792
     * `bucket_seq_to_instance_idx` is empty if no scan operator is contained in this fragment.
793
     * So co-located operators(e.g. Agg, Analytic) should use `HASH_SHUFFLE` instead of `BUCKET_HASH_SHUFFLE`.
794
     */
795
101k
    const bool followed_by_shuffled_operator =
796
101k
            operators.size() > idx ? operators[idx]->followed_by_shuffled_operator()
797
101k
                                   : cur_pipe->sink()->followed_by_shuffled_operator();
798
101k
    const bool use_global_hash_shuffle = bucket_seq_to_instance_idx.empty() &&
799
101k
                                         !shuffle_idx_to_instance_idx.contains(-1) &&
800
101k
                                         followed_by_shuffled_operator && !_use_serial_source;
801
101k
    sink = std::make_shared<LocalExchangeSinkOperatorX>(
802
101k
            sink_id, local_exchange_id, use_global_hash_shuffle ? _total_instances : _num_instances,
803
101k
            data_distribution.partition_exprs, bucket_seq_to_instance_idx);
804
101k
    if (bucket_seq_to_instance_idx.empty() &&
805
101k
        data_distribution.distribution_type == ExchangeType::BUCKET_HASH_SHUFFLE) {
806
4
        data_distribution.distribution_type = ExchangeType::HASH_SHUFFLE;
807
4
    }
808
101k
    RETURN_IF_ERROR(new_pip->set_sink(sink));
809
101k
    RETURN_IF_ERROR(new_pip->sink()->init(_runtime_state.get(), data_distribution.distribution_type,
810
101k
                                          num_buckets, use_global_hash_shuffle,
811
101k
                                          shuffle_idx_to_instance_idx));
812
813
    // 2. Create and initialize LocalExchangeSharedState.
814
101k
    std::shared_ptr<LocalExchangeSharedState> shared_state =
815
101k
            LocalExchangeSharedState::create_shared(_num_instances);
816
101k
    switch (data_distribution.distribution_type) {
817
19.6k
    case ExchangeType::HASH_SHUFFLE:
818
19.6k
        shared_state->exchanger = ShuffleExchanger::create_unique(
819
19.6k
                std::max(cur_pipe->num_tasks(), _num_instances), _num_instances,
820
19.6k
                use_global_hash_shuffle ? _total_instances : _num_instances,
821
19.6k
                _runtime_state->query_options().__isset.local_exchange_free_blocks_limit
822
19.6k
                        ? cast_set<int>(
823
19.6k
                                  _runtime_state->query_options().local_exchange_free_blocks_limit)
824
19.6k
                        : 0);
825
19.6k
        break;
826
499
    case ExchangeType::BUCKET_HASH_SHUFFLE:
827
499
        shared_state->exchanger = BucketShuffleExchanger::create_unique(
828
499
                std::max(cur_pipe->num_tasks(), _num_instances), _num_instances, num_buckets,
829
499
                _runtime_state->query_options().__isset.local_exchange_free_blocks_limit
830
499
                        ? cast_set<int>(
831
499
                                  _runtime_state->query_options().local_exchange_free_blocks_limit)
832
499
                        : 0);
833
499
        break;
834
77.7k
    case ExchangeType::PASSTHROUGH:
835
77.7k
        shared_state->exchanger = PassthroughExchanger::create_unique(
836
77.7k
                cur_pipe->num_tasks(), _num_instances,
837
77.7k
                _runtime_state->query_options().__isset.local_exchange_free_blocks_limit
838
77.7k
                        ? cast_set<int>(
839
77.7k
                                  _runtime_state->query_options().local_exchange_free_blocks_limit)
840
77.7k
                        : 0);
841
77.7k
        break;
842
285
    case ExchangeType::BROADCAST:
843
285
        shared_state->exchanger = BroadcastExchanger::create_unique(
844
285
                cur_pipe->num_tasks(), _num_instances,
845
285
                _runtime_state->query_options().__isset.local_exchange_free_blocks_limit
846
285
                        ? cast_set<int>(
847
285
                                  _runtime_state->query_options().local_exchange_free_blocks_limit)
848
285
                        : 0);
849
285
        break;
850
2.09k
    case ExchangeType::PASS_TO_ONE:
851
2.09k
        if (_runtime_state->enable_share_hash_table_for_broadcast_join()) {
852
            // If shared hash table is enabled for BJ, hash table will be built by only one task
853
1.02k
            shared_state->exchanger = PassToOneExchanger::create_unique(
854
1.02k
                    cur_pipe->num_tasks(), _num_instances,
855
1.02k
                    _runtime_state->query_options().__isset.local_exchange_free_blocks_limit
856
1.02k
                            ? cast_set<int>(_runtime_state->query_options()
857
1.02k
                                                    .local_exchange_free_blocks_limit)
858
1.02k
                            : 0);
859
1.06k
        } else {
860
1.06k
            shared_state->exchanger = BroadcastExchanger::create_unique(
861
1.06k
                    cur_pipe->num_tasks(), _num_instances,
862
1.06k
                    _runtime_state->query_options().__isset.local_exchange_free_blocks_limit
863
1.06k
                            ? cast_set<int>(_runtime_state->query_options()
864
1.06k
                                                    .local_exchange_free_blocks_limit)
865
1.06k
                            : 0);
866
1.06k
        }
867
2.09k
        break;
868
882
    case ExchangeType::ADAPTIVE_PASSTHROUGH:
869
882
        shared_state->exchanger = AdaptivePassthroughExchanger::create_unique(
870
882
                std::max(cur_pipe->num_tasks(), _num_instances), _num_instances,
871
882
                _runtime_state->query_options().__isset.local_exchange_free_blocks_limit
872
882
                        ? cast_set<int>(
873
882
                                  _runtime_state->query_options().local_exchange_free_blocks_limit)
874
882
                        : 0);
875
882
        break;
876
0
    default:
877
0
        return Status::InternalError("Unsupported local exchange type : " +
878
0
                                     std::to_string((int)data_distribution.distribution_type));
879
101k
    }
880
101k
    shared_state->create_source_dependencies(_num_instances, local_exchange_id, local_exchange_id,
881
101k
                                             "LOCAL_EXCHANGE_OPERATOR");
882
101k
    shared_state->create_sink_dependency(sink_id, local_exchange_id, "LOCAL_EXCHANGE_SINK");
883
101k
    _op_id_to_shared_state.insert({local_exchange_id, {shared_state, shared_state->sink_deps}});
884
885
    // 3. Set two pipelines' operator list. For example, split pipeline [Scan - AggSink] to
886
    // pipeline1 [Scan - LocalExchangeSink] and pipeline2 [LocalExchangeSource - AggSink].
887
888
    // 3.1 Initialize new pipeline's operator list.
889
101k
    std::copy(operators.begin(), operators.begin() + idx,
890
101k
              std::inserter(new_pip->operators(), new_pip->operators().end()));
891
892
    // 3.2 Erase unused operators in previous pipeline.
893
101k
    operators.erase(operators.begin(), operators.begin() + idx);
894
895
    // 4. Initialize LocalExchangeSource and insert it into this pipeline.
896
101k
    OperatorPtr source_op;
897
101k
    source_op = std::make_shared<LocalExchangeSourceOperatorX>(pool, local_exchange_id);
898
101k
    RETURN_IF_ERROR(source_op->set_child(new_pip->operators().back()));
899
101k
    RETURN_IF_ERROR(source_op->init(data_distribution.distribution_type));
900
101k
    if (!operators.empty()) {
901
42.3k
        RETURN_IF_ERROR(operators.front()->set_child(nullptr));
902
42.3k
        RETURN_IF_ERROR(operators.front()->set_child(source_op));
903
42.3k
    }
904
101k
    operators.insert(operators.begin(), source_op);
905
906
    // 5. Set children for two pipelines separately.
907
101k
    std::vector<std::shared_ptr<Pipeline>> new_children;
908
101k
    std::vector<PipelineId> edges_with_source;
909
117k
    for (auto child : cur_pipe->children()) {
910
117k
        bool found = false;
911
131k
        for (auto op : new_pip->operators()) {
912
131k
            if (child->sink()->node_id() == op->node_id()) {
913
11.7k
                new_pip->set_children(child);
914
11.7k
                found = true;
915
11.7k
            };
916
131k
        }
917
117k
        if (!found) {
918
105k
            new_children.push_back(child);
919
105k
            edges_with_source.push_back(child->id());
920
105k
        }
921
117k
    }
922
101k
    new_children.push_back(new_pip);
923
101k
    edges_with_source.push_back(new_pip->id());
924
925
    // 6. Set DAG for new pipelines.
926
101k
    if (!new_pip->children().empty()) {
927
6.60k
        std::vector<PipelineId> edges_with_sink;
928
11.7k
        for (auto child : new_pip->children()) {
929
11.7k
            edges_with_sink.push_back(child->id());
930
11.7k
        }
931
6.60k
        _dag.insert({new_pip->id(), edges_with_sink});
932
6.60k
    }
933
101k
    cur_pipe->set_children(new_children);
934
101k
    _dag[downstream_pipeline_id] = edges_with_source;
935
101k
    RETURN_IF_ERROR(new_pip->sink()->set_child(new_pip->operators().back()));
936
101k
    RETURN_IF_ERROR(cur_pipe->sink()->set_child(nullptr));
937
101k
    RETURN_IF_ERROR(cur_pipe->sink()->set_child(cur_pipe->operators().back()));
938
939
    // 7. Inherit properties from current pipeline.
940
101k
    _inherit_pipeline_properties(data_distribution, cur_pipe, new_pip);
941
101k
    return Status::OK();
942
101k
}
943
944
Status PipelineFragmentContext::_add_local_exchange(
945
        int pip_idx, int idx, int node_id, ObjectPool* pool, PipelinePtr cur_pipe,
946
        DataDistribution data_distribution, bool* do_local_exchange, int num_buckets,
947
        const std::map<int, int>& bucket_seq_to_instance_idx,
948
157k
        const std::map<int, int>& shuffle_idx_to_instance_idx) {
949
157k
    if (_num_instances <= 1 || cur_pipe->num_tasks_of_parent() <= 1) {
950
28.8k
        return Status::OK();
951
28.8k
    }
952
953
128k
    if (!cur_pipe->need_to_local_exchange(data_distribution, idx)) {
954
47.4k
        return Status::OK();
955
47.4k
    }
956
81.3k
    *do_local_exchange = true;
957
958
81.3k
    auto& operators = cur_pipe->operators();
959
81.3k
    auto total_op_num = operators.size();
960
81.3k
    auto new_pip = add_pipeline(cur_pipe, pip_idx + 1);
961
81.3k
    RETURN_IF_ERROR(_add_local_exchange_impl(
962
81.3k
            idx, pool, cur_pipe, new_pip, data_distribution, do_local_exchange, num_buckets,
963
81.3k
            bucket_seq_to_instance_idx, shuffle_idx_to_instance_idx));
964
965
18.4E
    CHECK(total_op_num + 1 == cur_pipe->operators().size() + new_pip->operators().size())
966
18.4E
            << "total_op_num: " << total_op_num
967
18.4E
            << " cur_pipe->operators().size(): " << cur_pipe->operators().size()
968
18.4E
            << " new_pip->operators().size(): " << new_pip->operators().size();
969
970
    // There are some local shuffles with relatively heavy operations on the sink.
971
    // If the local sink concurrency is 1 and the local source concurrency is n, the sink becomes a bottleneck.
972
    // Therefore, local passthrough is used to increase the concurrency of the sink.
973
    // op -> local sink(1) -> local source (n)
974
    // op -> local passthrough(1) -> local passthrough(n) ->  local sink(n) -> local source (n)
975
81.4k
    if (cur_pipe->num_tasks() > 1 && new_pip->num_tasks() == 1 &&
976
81.3k
        Pipeline::heavy_operations_on_the_sink(data_distribution.distribution_type)) {
977
19.6k
        RETURN_IF_ERROR(_add_local_exchange_impl(
978
19.6k
                cast_set<int>(new_pip->operators().size()), pool, new_pip,
979
19.6k
                add_pipeline(new_pip, pip_idx + 2), DataDistribution(ExchangeType::PASSTHROUGH),
980
19.6k
                do_local_exchange, num_buckets, bucket_seq_to_instance_idx,
981
19.6k
                shuffle_idx_to_instance_idx));
982
19.6k
    }
983
81.3k
    return Status::OK();
984
81.3k
}
985
986
Status PipelineFragmentContext::_plan_local_exchange(
987
        int num_buckets, const std::map<int, int>& bucket_seq_to_instance_idx,
988
318k
        const std::map<int, int>& shuffle_idx_to_instance_idx) {
989
723k
    for (int pip_idx = cast_set<int>(_pipelines.size()) - 1; pip_idx >= 0; pip_idx--) {
990
405k
        _pipelines[pip_idx]->init_data_distribution(_runtime_state.get());
991
        // Set property if child pipeline is not join operator's child.
992
405k
        if (!_pipelines[pip_idx]->children().empty()) {
993
85.4k
            for (auto& child : _pipelines[pip_idx]->children()) {
994
85.4k
                if (child->sink()->node_id() ==
995
85.4k
                    _pipelines[pip_idx]->operators().front()->node_id()) {
996
76.7k
                    _pipelines[pip_idx]->set_data_distribution(child->data_distribution());
997
76.7k
                }
998
85.4k
            }
999
82.8k
        }
1000
1001
        // if 'num_buckets == 0' means the fragment is colocated by exchange node not the
1002
        // scan node. so here use `_num_instance` to replace the `num_buckets` to prevent dividing 0
1003
        // still keep colocate plan after local shuffle
1004
405k
        RETURN_IF_ERROR(_plan_local_exchange(num_buckets, pip_idx, _pipelines[pip_idx],
1005
405k
                                             bucket_seq_to_instance_idx,
1006
405k
                                             shuffle_idx_to_instance_idx));
1007
405k
    }
1008
318k
    return Status::OK();
1009
318k
}
1010
1011
Status PipelineFragmentContext::_plan_local_exchange(
1012
        int num_buckets, int pip_idx, PipelinePtr pip,
1013
        const std::map<int, int>& bucket_seq_to_instance_idx,
1014
404k
        const std::map<int, int>& shuffle_idx_to_instance_idx) {
1015
404k
    int idx = 1;
1016
404k
    bool do_local_exchange = false;
1017
446k
    do {
1018
446k
        auto& ops = pip->operators();
1019
446k
        do_local_exchange = false;
1020
        // Plan local exchange for each operator.
1021
507k
        for (; idx < ops.size();) {
1022
103k
            if (ops[idx]->required_data_distribution(_runtime_state.get()).need_local_exchange()) {
1023
98.2k
                RETURN_IF_ERROR(_add_local_exchange(
1024
98.2k
                        pip_idx, idx, ops[idx]->node_id(), _runtime_state->obj_pool(), pip,
1025
98.2k
                        ops[idx]->required_data_distribution(_runtime_state.get()),
1026
98.2k
                        &do_local_exchange, num_buckets, bucket_seq_to_instance_idx,
1027
98.2k
                        shuffle_idx_to_instance_idx));
1028
98.2k
            }
1029
103k
            if (do_local_exchange) {
1030
                // If local exchange is needed for current operator, we will split this pipeline to
1031
                // two pipelines by local exchange sink/source. And then we need to process remaining
1032
                // operators in this pipeline so we set idx to 2 (0 is local exchange source and 1
1033
                // is current operator was already processed) and continue to plan local exchange.
1034
42.4k
                idx = 2;
1035
42.4k
                break;
1036
42.4k
            }
1037
61.0k
            idx++;
1038
61.0k
        }
1039
446k
    } while (do_local_exchange);
1040
404k
    if (pip->sink()->required_data_distribution(_runtime_state.get()).need_local_exchange()) {
1041
59.5k
        RETURN_IF_ERROR(_add_local_exchange(
1042
59.5k
                pip_idx, idx, pip->sink()->node_id(), _runtime_state->obj_pool(), pip,
1043
59.5k
                pip->sink()->required_data_distribution(_runtime_state.get()), &do_local_exchange,
1044
59.5k
                num_buckets, bucket_seq_to_instance_idx, shuffle_idx_to_instance_idx));
1045
59.5k
    }
1046
404k
    return Status::OK();
1047
404k
}
1048
1049
Status PipelineFragmentContext::_create_data_sink(ObjectPool* pool, const TDataSink& thrift_sink,
1050
                                                  const std::vector<TExpr>& output_exprs,
1051
                                                  const TPipelineFragmentParams& params,
1052
                                                  const RowDescriptor& row_desc,
1053
                                                  RuntimeState* state, DescriptorTbl& desc_tbl,
1054
320k
                                                  PipelineId cur_pipeline_id) {
1055
320k
    switch (thrift_sink.type) {
1056
110k
    case TDataSinkType::DATA_STREAM_SINK: {
1057
110k
        if (!thrift_sink.__isset.stream_sink) {
1058
0
            return Status::InternalError("Missing data stream sink.");
1059
0
        }
1060
110k
        _sink = std::make_shared<ExchangeSinkOperatorX>(
1061
110k
                state, row_desc, next_sink_operator_id(), thrift_sink.stream_sink,
1062
110k
                params.destinations, _fragment_instance_ids);
1063
110k
        break;
1064
110k
    }
1065
180k
    case TDataSinkType::RESULT_SINK: {
1066
180k
        if (!thrift_sink.__isset.result_sink) {
1067
0
            return Status::InternalError("Missing data buffer sink.");
1068
0
        }
1069
1070
180k
        _sink = std::make_shared<ResultSinkOperatorX>(next_sink_operator_id(), row_desc,
1071
180k
                                                      output_exprs, thrift_sink.result_sink);
1072
180k
        break;
1073
180k
    }
1074
105
    case TDataSinkType::DICTIONARY_SINK: {
1075
105
        if (!thrift_sink.__isset.dictionary_sink) {
1076
0
            return Status::InternalError("Missing dict sink.");
1077
0
        }
1078
1079
105
        _sink = std::make_shared<DictSinkOperatorX>(next_sink_operator_id(), row_desc, output_exprs,
1080
105
                                                    thrift_sink.dictionary_sink);
1081
105
        break;
1082
105
    }
1083
0
    case TDataSinkType::GROUP_COMMIT_OLAP_TABLE_SINK:
1084
28.6k
    case TDataSinkType::OLAP_TABLE_SINK: {
1085
28.6k
        if (state->query_options().enable_memtable_on_sink_node &&
1086
28.6k
            !_has_inverted_index_v1_or_partial_update(thrift_sink.olap_table_sink) &&
1087
28.6k
            !config::is_cloud_mode()) {
1088
34
            _sink = std::make_shared<OlapTableSinkV2OperatorX>(pool, next_sink_operator_id(),
1089
34
                                                               row_desc, output_exprs);
1090
28.6k
        } else {
1091
28.6k
            _sink = std::make_shared<OlapTableSinkOperatorX>(pool, next_sink_operator_id(),
1092
28.6k
                                                             row_desc, output_exprs);
1093
28.6k
        }
1094
28.6k
        break;
1095
0
    }
1096
165
    case TDataSinkType::GROUP_COMMIT_BLOCK_SINK: {
1097
165
        DCHECK(thrift_sink.__isset.olap_table_sink);
1098
165
        DCHECK(state->get_query_ctx() != nullptr);
1099
165
        state->get_query_ctx()->query_mem_tracker()->is_group_commit_load = true;
1100
165
        _sink = std::make_shared<GroupCommitBlockSinkOperatorX>(next_sink_operator_id(), row_desc,
1101
165
                                                                output_exprs);
1102
165
        break;
1103
0
    }
1104
0
    case TDataSinkType::HIVE_TABLE_SINK: {
1105
0
        if (!thrift_sink.__isset.hive_table_sink) {
1106
0
            return Status::InternalError("Missing hive table sink.");
1107
0
        }
1108
0
        _sink = std::make_shared<HiveTableSinkOperatorX>(pool, next_sink_operator_id(), row_desc,
1109
0
                                                         output_exprs);
1110
0
        break;
1111
0
    }
1112
0
    case TDataSinkType::ICEBERG_TABLE_SINK: {
1113
0
        if (!thrift_sink.__isset.iceberg_table_sink) {
1114
0
            return Status::InternalError("Missing iceberg table sink.");
1115
0
        }
1116
0
        if (thrift_sink.iceberg_table_sink.__isset.sort_info) {
1117
0
            _sink = std::make_shared<SpillIcebergTableSinkOperatorX>(pool, next_sink_operator_id(),
1118
0
                                                                     row_desc, output_exprs);
1119
0
        } else {
1120
0
            _sink = std::make_shared<IcebergTableSinkOperatorX>(pool, next_sink_operator_id(),
1121
0
                                                                row_desc, output_exprs);
1122
0
        }
1123
0
        break;
1124
0
    }
1125
0
    case TDataSinkType::ICEBERG_DELETE_SINK: {
1126
0
        if (!thrift_sink.__isset.iceberg_delete_sink) {
1127
0
            return Status::InternalError("Missing iceberg delete sink.");
1128
0
        }
1129
0
        _sink = std::make_shared<IcebergDeleteSinkOperatorX>(pool, next_sink_operator_id(),
1130
0
                                                             row_desc, output_exprs);
1131
0
        break;
1132
0
    }
1133
0
    case TDataSinkType::ICEBERG_MERGE_SINK: {
1134
0
        if (!thrift_sink.__isset.iceberg_merge_sink) {
1135
0
            return Status::InternalError("Missing iceberg merge sink.");
1136
0
        }
1137
0
        _sink = std::make_shared<IcebergMergeSinkOperatorX>(pool, next_sink_operator_id(), row_desc,
1138
0
                                                            output_exprs);
1139
0
        break;
1140
0
    }
1141
0
    case TDataSinkType::MAXCOMPUTE_TABLE_SINK: {
1142
0
        if (!thrift_sink.__isset.max_compute_table_sink) {
1143
0
            return Status::InternalError("Missing max compute table sink.");
1144
0
        }
1145
0
        _sink = std::make_shared<MCTableSinkOperatorX>(pool, next_sink_operator_id(), row_desc,
1146
0
                                                       output_exprs);
1147
0
        break;
1148
0
    }
1149
0
    case TDataSinkType::JDBC_TABLE_SINK: {
1150
0
        if (!thrift_sink.__isset.jdbc_table_sink) {
1151
0
            return Status::InternalError("Missing data jdbc sink.");
1152
0
        }
1153
0
        if (config::enable_java_support) {
1154
0
            _sink = std::make_shared<JdbcTableSinkOperatorX>(row_desc, next_sink_operator_id(),
1155
0
                                                             output_exprs);
1156
0
        } else {
1157
0
            return Status::InternalError(
1158
0
                    "Jdbc table sink is not enabled, you can change be config "
1159
0
                    "enable_java_support to true and restart be.");
1160
0
        }
1161
0
        break;
1162
0
    }
1163
3
    case TDataSinkType::MEMORY_SCRATCH_SINK: {
1164
3
        if (!thrift_sink.__isset.memory_scratch_sink) {
1165
0
            return Status::InternalError("Missing data buffer sink.");
1166
0
        }
1167
1168
3
        _sink = std::make_shared<MemoryScratchSinkOperatorX>(row_desc, next_sink_operator_id(),
1169
3
                                                             output_exprs);
1170
3
        break;
1171
3
    }
1172
338
    case TDataSinkType::RESULT_FILE_SINK: {
1173
338
        if (!thrift_sink.__isset.result_file_sink) {
1174
0
            return Status::InternalError("Missing result file sink.");
1175
0
        }
1176
1177
        // Result file sink is not the top sink
1178
338
        if (params.__isset.destinations && !params.destinations.empty()) {
1179
0
            _sink = std::make_shared<ResultFileSinkOperatorX>(
1180
0
                    next_sink_operator_id(), row_desc, thrift_sink.result_file_sink,
1181
0
                    params.destinations, output_exprs, desc_tbl);
1182
338
        } else {
1183
338
            _sink = std::make_shared<ResultFileSinkOperatorX>(next_sink_operator_id(), row_desc,
1184
338
                                                              output_exprs);
1185
338
        }
1186
338
        break;
1187
338
    }
1188
572
    case TDataSinkType::MULTI_CAST_DATA_STREAM_SINK: {
1189
572
        DCHECK(thrift_sink.__isset.multi_cast_stream_sink);
1190
572
        DCHECK_GT(thrift_sink.multi_cast_stream_sink.sinks.size(), 0);
1191
572
        auto sink_id = next_sink_operator_id();
1192
572
        const int multi_cast_node_id = sink_id;
1193
572
        auto sender_size = thrift_sink.multi_cast_stream_sink.sinks.size();
1194
        // one sink has multiple sources.
1195
572
        std::vector<int> sources;
1196
2.06k
        for (int i = 0; i < sender_size; ++i) {
1197
1.49k
            auto source_id = next_operator_id();
1198
1.49k
            sources.push_back(source_id);
1199
1.49k
        }
1200
1201
572
        _sink = std::make_shared<MultiCastDataStreamSinkOperatorX>(
1202
572
                sink_id, multi_cast_node_id, sources, pool, thrift_sink.multi_cast_stream_sink);
1203
2.06k
        for (int i = 0; i < sender_size; ++i) {
1204
1.49k
            auto new_pipeline = add_pipeline();
1205
            // use to exchange sink
1206
1.49k
            RowDescriptor* exchange_row_desc = nullptr;
1207
1.49k
            {
1208
1.49k
                const auto& tmp_row_desc =
1209
1.49k
                        !thrift_sink.multi_cast_stream_sink.sinks[i].output_exprs.empty()
1210
1.49k
                                ? RowDescriptor(state->desc_tbl(),
1211
1.49k
                                                {thrift_sink.multi_cast_stream_sink.sinks[i]
1212
1.49k
                                                         .output_tuple_id})
1213
1.49k
                                : row_desc;
1214
1.49k
                exchange_row_desc = pool->add(new RowDescriptor(tmp_row_desc));
1215
1.49k
            }
1216
1.49k
            auto source_id = sources[i];
1217
1.49k
            OperatorPtr source_op;
1218
            // 1. create and set the source operator of multi_cast_data_stream_source for new pipeline
1219
1.49k
            source_op = std::make_shared<MultiCastDataStreamerSourceOperatorX>(
1220
1.49k
                    /*node_id*/ source_id, /*consumer_id*/ i, pool,
1221
1.49k
                    thrift_sink.multi_cast_stream_sink.sinks[i], row_desc,
1222
1.49k
                    /*operator_id=*/source_id);
1223
1.49k
            RETURN_IF_ERROR(new_pipeline->add_operator(
1224
1.49k
                    source_op, params.__isset.parallel_instances ? params.parallel_instances : 0));
1225
            // 2. create and set sink operator of data stream sender for new pipeline
1226
1227
1.49k
            DataSinkOperatorPtr sink_op;
1228
1.49k
            sink_op = std::make_shared<ExchangeSinkOperatorX>(
1229
1.49k
                    state, *exchange_row_desc, next_sink_operator_id(),
1230
1.49k
                    thrift_sink.multi_cast_stream_sink.sinks[i],
1231
1.49k
                    thrift_sink.multi_cast_stream_sink.destinations[i], _fragment_instance_ids);
1232
1233
1.49k
            RETURN_IF_ERROR(new_pipeline->set_sink(sink_op));
1234
1.49k
            {
1235
1.49k
                TDataSink* t = pool->add(new TDataSink());
1236
1.49k
                t->stream_sink = thrift_sink.multi_cast_stream_sink.sinks[i];
1237
1.49k
                RETURN_IF_ERROR(sink_op->init(*t));
1238
1.49k
            }
1239
1240
            // 3. set dependency dag
1241
1.49k
            _dag[new_pipeline->id()].push_back(cur_pipeline_id);
1242
1.49k
        }
1243
572
        if (sources.empty()) {
1244
0
            return Status::InternalError("size of sources must be greater than 0");
1245
0
        }
1246
572
        break;
1247
572
    }
1248
572
    case TDataSinkType::BLACKHOLE_SINK: {
1249
3
        if (!thrift_sink.__isset.blackhole_sink) {
1250
0
            return Status::InternalError("Missing blackhole sink.");
1251
0
        }
1252
1253
3
        _sink.reset(new BlackholeSinkOperatorX(next_sink_operator_id()));
1254
3
        break;
1255
3
    }
1256
0
    case TDataSinkType::TVF_TABLE_SINK: {
1257
0
        if (!thrift_sink.__isset.tvf_table_sink) {
1258
0
            return Status::InternalError("Missing TVF table sink.");
1259
0
        }
1260
0
        _sink = std::make_shared<TVFTableSinkOperatorX>(pool, next_sink_operator_id(), row_desc,
1261
0
                                                        output_exprs);
1262
0
        break;
1263
0
    }
1264
0
    default:
1265
0
        return Status::InternalError("Unsuported sink type in pipeline: {}", thrift_sink.type);
1266
320k
    }
1267
320k
    return Status::OK();
1268
320k
}
1269
1270
// NOLINTBEGIN(readability-function-size)
1271
// NOLINTBEGIN(readability-function-cognitive-complexity)
1272
Status PipelineFragmentContext::_create_operator(ObjectPool* pool, const TPlanNode& tnode,
1273
                                                 const DescriptorTbl& descs, OperatorPtr& op,
1274
                                                 PipelinePtr& cur_pipe, int parent_idx,
1275
                                                 int child_idx,
1276
                                                 const bool followed_by_shuffled_operator,
1277
                                                 const bool require_bucket_distribution,
1278
510k
                                                 OperatorPtr& cache_op) {
1279
510k
    std::vector<DataSinkOperatorPtr> sink_ops;
1280
510k
    Defer defer = Defer([&]() {
1281
508k
        if (op) {
1282
508k
            op->update_operator(tnode, followed_by_shuffled_operator, require_bucket_distribution);
1283
508k
        }
1284
508k
        for (auto& s : sink_ops) {
1285
85.0k
            s->update_operator(tnode, followed_by_shuffled_operator, require_bucket_distribution);
1286
85.0k
        }
1287
508k
    });
1288
    // We directly construct the operator from Thrift because the given array is in the order of preorder traversal.
1289
    // Therefore, here we need to use a stack-like structure.
1290
510k
    _pipeline_parent_map.pop(cur_pipe, parent_idx, child_idx);
1291
510k
    std::stringstream error_msg;
1292
510k
    bool enable_query_cache = _params.fragment.__isset.query_cache_param;
1293
1294
510k
    bool fe_with_old_version = false;
1295
510k
    switch (tnode.node_type) {
1296
159k
    case TPlanNodeType::OLAP_SCAN_NODE: {
1297
159k
        op = std::make_shared<OlapScanOperatorX>(
1298
159k
                pool, tnode, next_operator_id(), descs, _num_instances,
1299
159k
                enable_query_cache ? _params.fragment.query_cache_param : TQueryCacheParam {});
1300
159k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1301
159k
        fe_with_old_version = !tnode.__isset.is_serial_operator;
1302
159k
        break;
1303
159k
    }
1304
78
    case TPlanNodeType::GROUP_COMMIT_SCAN_NODE: {
1305
78
        DCHECK(_query_ctx != nullptr);
1306
78
        _query_ctx->query_mem_tracker()->is_group_commit_load = true;
1307
78
        op = std::make_shared<GroupCommitOperatorX>(pool, tnode, next_operator_id(), descs,
1308
78
                                                    _num_instances);
1309
78
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1310
78
        fe_with_old_version = !tnode.__isset.is_serial_operator;
1311
78
        break;
1312
78
    }
1313
0
    case TPlanNodeType::JDBC_SCAN_NODE: {
1314
0
        if (config::enable_java_support) {
1315
0
            op = std::make_shared<JDBCScanOperatorX>(pool, tnode, next_operator_id(), descs,
1316
0
                                                     _num_instances);
1317
0
            RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1318
0
        } else {
1319
0
            return Status::InternalError(
1320
0
                    "Jdbc scan node is disabled, you can change be config enable_java_support "
1321
0
                    "to true and restart be.");
1322
0
        }
1323
0
        fe_with_old_version = !tnode.__isset.is_serial_operator;
1324
0
        break;
1325
0
    }
1326
2.64k
    case TPlanNodeType::FILE_SCAN_NODE: {
1327
2.64k
        op = std::make_shared<FileScanOperatorX>(pool, tnode, next_operator_id(), descs,
1328
2.64k
                                                 _num_instances);
1329
2.64k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1330
2.64k
        fe_with_old_version = !tnode.__isset.is_serial_operator;
1331
2.64k
        break;
1332
2.64k
    }
1333
0
    case TPlanNodeType::ES_SCAN_NODE:
1334
0
    case TPlanNodeType::ES_HTTP_SCAN_NODE: {
1335
0
        op = std::make_shared<EsScanOperatorX>(pool, tnode, next_operator_id(), descs,
1336
0
                                               _num_instances);
1337
0
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1338
0
        fe_with_old_version = !tnode.__isset.is_serial_operator;
1339
0
        break;
1340
0
    }
1341
109k
    case TPlanNodeType::EXCHANGE_NODE: {
1342
109k
        int num_senders = _params.per_exch_num_senders.contains(tnode.node_id)
1343
109k
                                  ? _params.per_exch_num_senders.find(tnode.node_id)->second
1344
18.4E
                                  : 0;
1345
109k
        DCHECK_GT(num_senders, 0);
1346
109k
        op = std::make_shared<ExchangeSourceOperatorX>(pool, tnode, next_operator_id(), descs,
1347
109k
                                                       num_senders);
1348
109k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1349
109k
        fe_with_old_version = !tnode.__isset.is_serial_operator;
1350
109k
        break;
1351
109k
    }
1352
133k
    case TPlanNodeType::AGGREGATION_NODE: {
1353
133k
        if (tnode.agg_node.grouping_exprs.empty() &&
1354
133k
            descs.get_tuple_descriptor(tnode.agg_node.output_tuple_id)->slots().empty()) {
1355
0
            return Status::InternalError("Illegal aggregate node " + std::to_string(tnode.node_id) +
1356
0
                                         ": group by and output is empty");
1357
0
        }
1358
133k
        bool need_create_cache_op =
1359
133k
                enable_query_cache && tnode.node_id == _params.fragment.query_cache_param.node_id;
1360
133k
        auto create_query_cache_operator = [&](PipelinePtr& new_pipe) {
1361
10
            auto cache_node_id = _params.local_params[0].per_node_scan_ranges.begin()->first;
1362
10
            auto cache_source_id = next_operator_id();
1363
10
            op = std::make_shared<CacheSourceOperatorX>(pool, cache_node_id, cache_source_id,
1364
10
                                                        _params.fragment.query_cache_param);
1365
10
            RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1366
1367
10
            const auto downstream_pipeline_id = cur_pipe->id();
1368
10
            if (!_dag.contains(downstream_pipeline_id)) {
1369
10
                _dag.insert({downstream_pipeline_id, {}});
1370
10
            }
1371
10
            new_pipe = add_pipeline(cur_pipe);
1372
10
            _dag[downstream_pipeline_id].push_back(new_pipe->id());
1373
1374
10
            DataSinkOperatorPtr cache_sink(new CacheSinkOperatorX(
1375
10
                    next_sink_operator_id(), op->node_id(), op->operator_id()));
1376
10
            RETURN_IF_ERROR(new_pipe->set_sink(cache_sink));
1377
10
            return Status::OK();
1378
10
        };
1379
133k
        const bool group_by_limit_opt =
1380
133k
                tnode.agg_node.__isset.agg_sort_info_by_group_key && tnode.limit > 0;
1381
1382
        /// PartitionedAggSourceOperatorX does not support "group by limit opt(#29641)" yet.
1383
        /// If `group_by_limit_opt` is true, then it might not need to spill at all.
1384
133k
        const bool enable_spill = _runtime_state->enable_spill() &&
1385
133k
                                  !tnode.agg_node.grouping_exprs.empty() && !group_by_limit_opt;
1386
133k
        const bool is_streaming_agg = tnode.agg_node.__isset.use_streaming_preaggregation &&
1387
133k
                                      tnode.agg_node.use_streaming_preaggregation &&
1388
133k
                                      !tnode.agg_node.grouping_exprs.empty();
1389
        // TODO: distinct streaming agg does not support spill.
1390
133k
        const bool can_use_distinct_streaming_agg =
1391
133k
                (!enable_spill || is_streaming_agg) && tnode.agg_node.aggregate_functions.empty() &&
1392
133k
                !tnode.agg_node.__isset.agg_sort_info_by_group_key &&
1393
133k
                _params.query_options.__isset.enable_distinct_streaming_aggregation &&
1394
133k
                _params.query_options.enable_distinct_streaming_aggregation;
1395
1396
133k
        if (can_use_distinct_streaming_agg) {
1397
91.2k
            if (need_create_cache_op) {
1398
8
                PipelinePtr new_pipe;
1399
8
                RETURN_IF_ERROR(create_query_cache_operator(new_pipe));
1400
1401
8
                cache_op = op;
1402
8
                op = std::make_shared<DistinctStreamingAggOperatorX>(pool, next_operator_id(),
1403
8
                                                                     tnode, descs);
1404
8
                RETURN_IF_ERROR(new_pipe->add_operator(op, _parallel_instances));
1405
8
                RETURN_IF_ERROR(cur_pipe->operators().front()->set_child(op));
1406
8
                cur_pipe = new_pipe;
1407
91.2k
            } else {
1408
91.2k
                op = std::make_shared<DistinctStreamingAggOperatorX>(pool, next_operator_id(),
1409
91.2k
                                                                     tnode, descs);
1410
91.2k
                RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1411
91.2k
            }
1412
91.2k
        } else if (is_streaming_agg) {
1413
1.24k
            if (need_create_cache_op) {
1414
0
                PipelinePtr new_pipe;
1415
0
                RETURN_IF_ERROR(create_query_cache_operator(new_pipe));
1416
0
                cache_op = op;
1417
0
                op = std::make_shared<StreamingAggOperatorX>(pool, next_operator_id(), tnode,
1418
0
                                                             descs);
1419
0
                RETURN_IF_ERROR(cur_pipe->operators().front()->set_child(op));
1420
0
                RETURN_IF_ERROR(new_pipe->add_operator(op, _parallel_instances));
1421
0
                cur_pipe = new_pipe;
1422
1.24k
            } else {
1423
1.24k
                op = std::make_shared<StreamingAggOperatorX>(pool, next_operator_id(), tnode,
1424
1.24k
                                                             descs);
1425
1.24k
                RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1426
1.24k
            }
1427
40.9k
        } else {
1428
            // create new pipeline to add query cache operator
1429
40.9k
            PipelinePtr new_pipe;
1430
40.9k
            if (need_create_cache_op) {
1431
2
                RETURN_IF_ERROR(create_query_cache_operator(new_pipe));
1432
2
                cache_op = op;
1433
2
            }
1434
1435
40.9k
            if (enable_spill) {
1436
105
                op = std::make_shared<PartitionedAggSourceOperatorX>(pool, tnode,
1437
105
                                                                     next_operator_id(), descs);
1438
40.8k
            } else {
1439
40.8k
                op = std::make_shared<AggSourceOperatorX>(pool, tnode, next_operator_id(), descs);
1440
40.8k
            }
1441
40.9k
            if (need_create_cache_op) {
1442
2
                RETURN_IF_ERROR(cur_pipe->operators().front()->set_child(op));
1443
2
                RETURN_IF_ERROR(new_pipe->add_operator(op, _parallel_instances));
1444
2
                cur_pipe = new_pipe;
1445
40.9k
            } else {
1446
40.9k
                RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1447
40.9k
            }
1448
1449
40.9k
            const auto downstream_pipeline_id = cur_pipe->id();
1450
40.9k
            if (!_dag.contains(downstream_pipeline_id)) {
1451
39.9k
                _dag.insert({downstream_pipeline_id, {}});
1452
39.9k
            }
1453
40.9k
            cur_pipe = add_pipeline(cur_pipe);
1454
40.9k
            _dag[downstream_pipeline_id].push_back(cur_pipe->id());
1455
1456
40.9k
            if (enable_spill) {
1457
105
                sink_ops.push_back(std::make_shared<PartitionedAggSinkOperatorX>(
1458
105
                        pool, next_sink_operator_id(), op->operator_id(), tnode, descs));
1459
40.8k
            } else {
1460
40.8k
                sink_ops.push_back(std::make_shared<AggSinkOperatorX>(
1461
40.8k
                        pool, next_sink_operator_id(), op->operator_id(), tnode, descs));
1462
40.8k
            }
1463
40.9k
            RETURN_IF_ERROR(cur_pipe->set_sink(sink_ops.back()));
1464
40.9k
            RETURN_IF_ERROR(cur_pipe->sink()->init(tnode, _runtime_state.get()));
1465
40.9k
        }
1466
133k
        break;
1467
133k
    }
1468
133k
    case TPlanNodeType::HASH_JOIN_NODE: {
1469
6.91k
        const auto is_broadcast_join = tnode.hash_join_node.__isset.is_broadcast_join &&
1470
6.91k
                                       tnode.hash_join_node.is_broadcast_join;
1471
6.91k
        const auto enable_spill = _runtime_state->enable_spill();
1472
6.91k
        if (enable_spill && !is_broadcast_join) {
1473
0
            auto tnode_ = tnode;
1474
0
            tnode_.runtime_filters.clear();
1475
0
            auto inner_probe_operator =
1476
0
                    std::make_shared<HashJoinProbeOperatorX>(pool, tnode_, 0, descs);
1477
1478
            // probe side inner sink operator is used to build hash table on probe side when data is spilled.
1479
            // So here use `tnode_` which has no runtime filters.
1480
0
            auto probe_side_inner_sink_operator =
1481
0
                    std::make_shared<HashJoinBuildSinkOperatorX>(pool, 0, 0, tnode_, descs);
1482
1483
0
            RETURN_IF_ERROR(inner_probe_operator->init(tnode_, _runtime_state.get()));
1484
0
            RETURN_IF_ERROR(probe_side_inner_sink_operator->init(tnode_, _runtime_state.get()));
1485
1486
0
            auto probe_operator = std::make_shared<PartitionedHashJoinProbeOperatorX>(
1487
0
                    pool, tnode_, next_operator_id(), descs);
1488
0
            probe_operator->set_inner_operators(probe_side_inner_sink_operator,
1489
0
                                                inner_probe_operator);
1490
0
            op = std::move(probe_operator);
1491
0
            RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1492
1493
0
            const auto downstream_pipeline_id = cur_pipe->id();
1494
0
            if (!_dag.contains(downstream_pipeline_id)) {
1495
0
                _dag.insert({downstream_pipeline_id, {}});
1496
0
            }
1497
0
            PipelinePtr build_side_pipe = add_pipeline(cur_pipe);
1498
0
            _dag[downstream_pipeline_id].push_back(build_side_pipe->id());
1499
1500
0
            auto inner_sink_operator =
1501
0
                    std::make_shared<HashJoinBuildSinkOperatorX>(pool, 0, 0, tnode, descs);
1502
0
            auto sink_operator = std::make_shared<PartitionedHashJoinSinkOperatorX>(
1503
0
                    pool, next_sink_operator_id(), op->operator_id(), tnode_, descs);
1504
0
            RETURN_IF_ERROR(inner_sink_operator->init(tnode, _runtime_state.get()));
1505
1506
0
            sink_operator->set_inner_operators(inner_sink_operator, inner_probe_operator);
1507
0
            sink_ops.push_back(std::move(sink_operator));
1508
0
            RETURN_IF_ERROR(build_side_pipe->set_sink(sink_ops.back()));
1509
0
            RETURN_IF_ERROR(build_side_pipe->sink()->init(tnode_, _runtime_state.get()));
1510
1511
0
            _pipeline_parent_map.push(op->node_id(), cur_pipe);
1512
0
            _pipeline_parent_map.push(op->node_id(), build_side_pipe);
1513
6.91k
        } else {
1514
6.91k
            op = std::make_shared<HashJoinProbeOperatorX>(pool, tnode, next_operator_id(), descs);
1515
6.91k
            RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1516
1517
6.91k
            const auto downstream_pipeline_id = cur_pipe->id();
1518
6.91k
            if (!_dag.contains(downstream_pipeline_id)) {
1519
6.14k
                _dag.insert({downstream_pipeline_id, {}});
1520
6.14k
            }
1521
6.91k
            PipelinePtr build_side_pipe = add_pipeline(cur_pipe);
1522
6.91k
            _dag[downstream_pipeline_id].push_back(build_side_pipe->id());
1523
1524
6.91k
            sink_ops.push_back(std::make_shared<HashJoinBuildSinkOperatorX>(
1525
6.91k
                    pool, next_sink_operator_id(), op->operator_id(), tnode, descs));
1526
6.91k
            RETURN_IF_ERROR(build_side_pipe->set_sink(sink_ops.back()));
1527
6.91k
            RETURN_IF_ERROR(build_side_pipe->sink()->init(tnode, _runtime_state.get()));
1528
1529
6.91k
            _pipeline_parent_map.push(op->node_id(), cur_pipe);
1530
6.91k
            _pipeline_parent_map.push(op->node_id(), build_side_pipe);
1531
6.91k
        }
1532
6.91k
        if (is_broadcast_join && _runtime_state->enable_share_hash_table_for_broadcast_join()) {
1533
2.03k
            std::shared_ptr<HashJoinSharedState> shared_state =
1534
2.03k
                    HashJoinSharedState::create_shared(_num_instances);
1535
14.8k
            for (int i = 0; i < _num_instances; i++) {
1536
12.8k
                auto sink_dep = std::make_shared<Dependency>(op->operator_id(), op->node_id(),
1537
12.8k
                                                             "HASH_JOIN_BUILD_DEPENDENCY");
1538
12.8k
                sink_dep->set_shared_state(shared_state.get());
1539
12.8k
                shared_state->sink_deps.push_back(sink_dep);
1540
12.8k
            }
1541
2.03k
            shared_state->create_source_dependencies(_num_instances, op->operator_id(),
1542
2.03k
                                                     op->node_id(), "HASH_JOIN_PROBE");
1543
2.03k
            _op_id_to_shared_state.insert(
1544
2.03k
                    {op->operator_id(), {shared_state, shared_state->sink_deps}});
1545
2.03k
        }
1546
6.91k
        break;
1547
6.91k
    }
1548
1.78k
    case TPlanNodeType::CROSS_JOIN_NODE: {
1549
1.78k
        op = std::make_shared<NestedLoopJoinProbeOperatorX>(pool, tnode, next_operator_id(), descs);
1550
1.78k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1551
1552
1.78k
        const auto downstream_pipeline_id = cur_pipe->id();
1553
1.78k
        if (!_dag.contains(downstream_pipeline_id)) {
1554
1.55k
            _dag.insert({downstream_pipeline_id, {}});
1555
1.55k
        }
1556
1.78k
        PipelinePtr build_side_pipe = add_pipeline(cur_pipe);
1557
1.78k
        _dag[downstream_pipeline_id].push_back(build_side_pipe->id());
1558
1559
1.78k
        sink_ops.push_back(std::make_shared<NestedLoopJoinBuildSinkOperatorX>(
1560
1.78k
                pool, next_sink_operator_id(), op->operator_id(), tnode, descs));
1561
1.78k
        RETURN_IF_ERROR(build_side_pipe->set_sink(sink_ops.back()));
1562
1.78k
        RETURN_IF_ERROR(build_side_pipe->sink()->init(tnode, _runtime_state.get()));
1563
1.78k
        _pipeline_parent_map.push(op->node_id(), cur_pipe);
1564
1.78k
        _pipeline_parent_map.push(op->node_id(), build_side_pipe);
1565
1.78k
        break;
1566
1.78k
    }
1567
49.1k
    case TPlanNodeType::UNION_NODE: {
1568
49.1k
        int child_count = tnode.num_children;
1569
49.1k
        op = std::make_shared<UnionSourceOperatorX>(pool, tnode, next_operator_id(), descs);
1570
49.1k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1571
1572
49.1k
        const auto downstream_pipeline_id = cur_pipe->id();
1573
49.1k
        if (!_dag.contains(downstream_pipeline_id)) {
1574
48.6k
            _dag.insert({downstream_pipeline_id, {}});
1575
48.6k
        }
1576
50.2k
        for (int i = 0; i < child_count; i++) {
1577
1.01k
            PipelinePtr build_side_pipe = add_pipeline(cur_pipe);
1578
1.01k
            _dag[downstream_pipeline_id].push_back(build_side_pipe->id());
1579
1.01k
            sink_ops.push_back(std::make_shared<UnionSinkOperatorX>(
1580
1.01k
                    i, next_sink_operator_id(), op->operator_id(), pool, tnode, descs));
1581
1.01k
            RETURN_IF_ERROR(build_side_pipe->set_sink(sink_ops.back()));
1582
1.01k
            RETURN_IF_ERROR(build_side_pipe->sink()->init(tnode, _runtime_state.get()));
1583
            // preset children pipelines. if any pipeline found this as its father, will use the prepared pipeline to build.
1584
1.01k
            _pipeline_parent_map.push(op->node_id(), build_side_pipe);
1585
1.01k
        }
1586
49.1k
        break;
1587
49.1k
    }
1588
49.1k
    case TPlanNodeType::SORT_NODE: {
1589
32.6k
        const auto should_spill = _runtime_state->enable_spill() &&
1590
32.6k
                                  tnode.sort_node.algorithm == TSortAlgorithm::FULL_SORT;
1591
32.6k
        const bool use_local_merge =
1592
32.6k
                tnode.sort_node.__isset.use_local_merge && tnode.sort_node.use_local_merge;
1593
32.6k
        if (should_spill) {
1594
9
            op = std::make_shared<SpillSortSourceOperatorX>(pool, tnode, next_operator_id(), descs);
1595
32.6k
        } else if (use_local_merge) {
1596
30.7k
            op = std::make_shared<LocalMergeSortSourceOperatorX>(pool, tnode, next_operator_id(),
1597
30.7k
                                                                 descs);
1598
30.7k
        } else {
1599
1.88k
            op = std::make_shared<SortSourceOperatorX>(pool, tnode, next_operator_id(), descs);
1600
1.88k
        }
1601
32.6k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1602
1603
32.6k
        const auto downstream_pipeline_id = cur_pipe->id();
1604
32.6k
        if (!_dag.contains(downstream_pipeline_id)) {
1605
32.6k
            _dag.insert({downstream_pipeline_id, {}});
1606
32.6k
        }
1607
32.6k
        cur_pipe = add_pipeline(cur_pipe);
1608
32.6k
        _dag[downstream_pipeline_id].push_back(cur_pipe->id());
1609
1610
32.6k
        if (should_spill) {
1611
9
            sink_ops.push_back(std::make_shared<SpillSortSinkOperatorX>(
1612
9
                    pool, next_sink_operator_id(), op->operator_id(), tnode, descs));
1613
32.6k
        } else {
1614
32.6k
            sink_ops.push_back(std::make_shared<SortSinkOperatorX>(
1615
32.6k
                    pool, next_sink_operator_id(), op->operator_id(), tnode, descs));
1616
32.6k
        }
1617
32.6k
        RETURN_IF_ERROR(cur_pipe->set_sink(sink_ops.back()));
1618
32.6k
        RETURN_IF_ERROR(cur_pipe->sink()->init(tnode, _runtime_state.get()));
1619
32.6k
        break;
1620
32.6k
    }
1621
32.6k
    case TPlanNodeType::PARTITION_SORT_NODE: {
1622
62
        op = std::make_shared<PartitionSortSourceOperatorX>(pool, tnode, next_operator_id(), descs);
1623
62
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1624
1625
62
        const auto downstream_pipeline_id = cur_pipe->id();
1626
62
        if (!_dag.contains(downstream_pipeline_id)) {
1627
62
            _dag.insert({downstream_pipeline_id, {}});
1628
62
        }
1629
62
        cur_pipe = add_pipeline(cur_pipe);
1630
62
        _dag[downstream_pipeline_id].push_back(cur_pipe->id());
1631
1632
62
        sink_ops.push_back(std::make_shared<PartitionSortSinkOperatorX>(
1633
62
                pool, next_sink_operator_id(), op->operator_id(), tnode, descs));
1634
62
        RETURN_IF_ERROR(cur_pipe->set_sink(sink_ops.back()));
1635
62
        RETURN_IF_ERROR(cur_pipe->sink()->init(tnode, _runtime_state.get()));
1636
62
        break;
1637
62
    }
1638
1.60k
    case TPlanNodeType::ANALYTIC_EVAL_NODE: {
1639
1.60k
        op = std::make_shared<AnalyticSourceOperatorX>(pool, tnode, next_operator_id(), descs);
1640
1.60k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1641
1642
1.60k
        const auto downstream_pipeline_id = cur_pipe->id();
1643
1.60k
        if (!_dag.contains(downstream_pipeline_id)) {
1644
1.58k
            _dag.insert({downstream_pipeline_id, {}});
1645
1.58k
        }
1646
1.60k
        cur_pipe = add_pipeline(cur_pipe);
1647
1.60k
        _dag[downstream_pipeline_id].push_back(cur_pipe->id());
1648
1649
1.60k
        sink_ops.push_back(std::make_shared<AnalyticSinkOperatorX>(
1650
1.60k
                pool, next_sink_operator_id(), op->operator_id(), tnode, descs));
1651
1.60k
        RETURN_IF_ERROR(cur_pipe->set_sink(sink_ops.back()));
1652
1.60k
        RETURN_IF_ERROR(cur_pipe->sink()->init(tnode, _runtime_state.get()));
1653
1.60k
        break;
1654
1.60k
    }
1655
1.60k
    case TPlanNodeType::MATERIALIZATION_NODE: {
1656
662
        op = std::make_shared<MaterializationOperator>(pool, tnode, next_operator_id(), descs);
1657
662
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1658
662
        break;
1659
662
    }
1660
662
    case TPlanNodeType::INTERSECT_NODE: {
1661
114
        RETURN_IF_ERROR(_build_operators_for_set_operation_node<true>(pool, tnode, descs, op,
1662
114
                                                                      cur_pipe, sink_ops));
1663
114
        break;
1664
114
    }
1665
125
    case TPlanNodeType::EXCEPT_NODE: {
1666
125
        RETURN_IF_ERROR(_build_operators_for_set_operation_node<false>(pool, tnode, descs, op,
1667
125
                                                                       cur_pipe, sink_ops));
1668
125
        break;
1669
125
    }
1670
298
    case TPlanNodeType::REPEAT_NODE: {
1671
298
        op = std::make_shared<RepeatOperatorX>(pool, tnode, next_operator_id(), descs);
1672
298
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1673
298
        break;
1674
298
    }
1675
917
    case TPlanNodeType::TABLE_FUNCTION_NODE: {
1676
917
        op = std::make_shared<TableFunctionOperatorX>(pool, tnode, next_operator_id(), descs);
1677
917
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1678
917
        break;
1679
917
    }
1680
917
    case TPlanNodeType::ASSERT_NUM_ROWS_NODE: {
1681
18
        op = std::make_shared<AssertNumRowsOperatorX>(pool, tnode, next_operator_id(), descs);
1682
18
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1683
18
        break;
1684
18
    }
1685
1.46k
    case TPlanNodeType::EMPTY_SET_NODE: {
1686
1.46k
        op = std::make_shared<EmptySetSourceOperatorX>(pool, tnode, next_operator_id(), descs);
1687
1.46k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1688
1.46k
        break;
1689
1.46k
    }
1690
1.46k
    case TPlanNodeType::DATA_GEN_SCAN_NODE: {
1691
265
        op = std::make_shared<DataGenSourceOperatorX>(pool, tnode, next_operator_id(), descs);
1692
265
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1693
265
        fe_with_old_version = !tnode.__isset.is_serial_operator;
1694
265
        break;
1695
265
    }
1696
1.52k
    case TPlanNodeType::SCHEMA_SCAN_NODE: {
1697
1.52k
        op = std::make_shared<SchemaScanOperatorX>(pool, tnode, next_operator_id(), descs);
1698
1.52k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1699
1.52k
        break;
1700
1.52k
    }
1701
5.13k
    case TPlanNodeType::META_SCAN_NODE: {
1702
5.13k
        op = std::make_shared<MetaScanOperatorX>(pool, tnode, next_operator_id(), descs);
1703
5.13k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1704
5.13k
        break;
1705
5.13k
    }
1706
5.13k
    case TPlanNodeType::SELECT_NODE: {
1707
326
        op = std::make_shared<SelectOperatorX>(pool, tnode, next_operator_id(), descs);
1708
326
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1709
326
        break;
1710
326
    }
1711
326
    case TPlanNodeType::REC_CTE_NODE: {
1712
151
        op = std::make_shared<RecCTESourceOperatorX>(pool, tnode, next_operator_id(), descs);
1713
151
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1714
1715
151
        const auto downstream_pipeline_id = cur_pipe->id();
1716
151
        if (!_dag.contains(downstream_pipeline_id)) {
1717
148
            _dag.insert({downstream_pipeline_id, {}});
1718
148
        }
1719
1720
151
        PipelinePtr anchor_side_pipe = add_pipeline(cur_pipe);
1721
151
        _dag[downstream_pipeline_id].push_back(anchor_side_pipe->id());
1722
1723
151
        DataSinkOperatorPtr anchor_sink;
1724
151
        anchor_sink = std::make_shared<RecCTEAnchorSinkOperatorX>(next_sink_operator_id(),
1725
151
                                                                  op->operator_id(), tnode, descs);
1726
151
        RETURN_IF_ERROR(anchor_side_pipe->set_sink(anchor_sink));
1727
151
        RETURN_IF_ERROR(anchor_side_pipe->sink()->init(tnode, _runtime_state.get()));
1728
151
        _pipeline_parent_map.push(op->node_id(), anchor_side_pipe);
1729
1730
151
        PipelinePtr rec_side_pipe = add_pipeline(cur_pipe);
1731
151
        _dag[downstream_pipeline_id].push_back(rec_side_pipe->id());
1732
1733
151
        DataSinkOperatorPtr rec_sink;
1734
151
        rec_sink = std::make_shared<RecCTESinkOperatorX>(next_sink_operator_id(), op->operator_id(),
1735
151
                                                         tnode, descs);
1736
151
        RETURN_IF_ERROR(rec_side_pipe->set_sink(rec_sink));
1737
151
        RETURN_IF_ERROR(rec_side_pipe->sink()->init(tnode, _runtime_state.get()));
1738
151
        _pipeline_parent_map.push(op->node_id(), rec_side_pipe);
1739
1740
151
        break;
1741
151
    }
1742
1.95k
    case TPlanNodeType::REC_CTE_SCAN_NODE: {
1743
1.95k
        op = std::make_shared<RecCTEScanOperatorX>(pool, tnode, next_operator_id(), descs);
1744
1.95k
        RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1745
1.95k
        break;
1746
1.95k
    }
1747
1.95k
    default:
1748
0
        return Status::InternalError("Unsupported exec type in pipeline: {}",
1749
0
                                     print_plan_node_type(tnode.node_type));
1750
510k
    }
1751
507k
    if (_params.__isset.parallel_instances && fe_with_old_version) {
1752
0
        cur_pipe->set_num_tasks(_params.parallel_instances);
1753
0
        op->set_serial_operator();
1754
0
    }
1755
1756
507k
    return Status::OK();
1757
510k
}
1758
// NOLINTEND(readability-function-cognitive-complexity)
1759
// NOLINTEND(readability-function-size)
1760
1761
template <bool is_intersect>
1762
Status PipelineFragmentContext::_build_operators_for_set_operation_node(
1763
        ObjectPool* pool, const TPlanNode& tnode, const DescriptorTbl& descs, OperatorPtr& op,
1764
239
        PipelinePtr& cur_pipe, std::vector<DataSinkOperatorPtr>& sink_ops) {
1765
239
    op.reset(new SetSourceOperatorX<is_intersect>(pool, tnode, next_operator_id(), descs));
1766
239
    RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1767
1768
239
    const auto downstream_pipeline_id = cur_pipe->id();
1769
239
    if (!_dag.contains(downstream_pipeline_id)) {
1770
222
        _dag.insert({downstream_pipeline_id, {}});
1771
222
    }
1772
1773
810
    for (int child_id = 0; child_id < tnode.num_children; child_id++) {
1774
571
        PipelinePtr probe_side_pipe = add_pipeline(cur_pipe);
1775
571
        _dag[downstream_pipeline_id].push_back(probe_side_pipe->id());
1776
1777
571
        if (child_id == 0) {
1778
239
            sink_ops.push_back(std::make_shared<SetSinkOperatorX<is_intersect>>(
1779
239
                    child_id, next_sink_operator_id(), op->operator_id(), pool, tnode, descs));
1780
332
        } else {
1781
332
            sink_ops.push_back(std::make_shared<SetProbeSinkOperatorX<is_intersect>>(
1782
332
                    child_id, next_sink_operator_id(), op->operator_id(), pool, tnode, descs));
1783
332
        }
1784
571
        RETURN_IF_ERROR(probe_side_pipe->set_sink(sink_ops.back()));
1785
571
        RETURN_IF_ERROR(probe_side_pipe->sink()->init(tnode, _runtime_state.get()));
1786
        // prepare children pipelines. if any pipeline found this as its father, will use the prepared pipeline to build.
1787
571
        _pipeline_parent_map.push(op->node_id(), probe_side_pipe);
1788
571
    }
1789
1790
239
    return Status::OK();
1791
239
}
_ZN5doris23PipelineFragmentContext39_build_operators_for_set_operation_nodeILb1EEENS_6StatusEPNS_10ObjectPoolERKNS_9TPlanNodeERKNS_13DescriptorTblERSt10shared_ptrINS_13OperatorXBaseEERSB_INS_8PipelineEERSt6vectorISB_INS_21DataSinkOperatorXBaseEESaISK_EE
Line
Count
Source
1764
114
        PipelinePtr& cur_pipe, std::vector<DataSinkOperatorPtr>& sink_ops) {
1765
114
    op.reset(new SetSourceOperatorX<is_intersect>(pool, tnode, next_operator_id(), descs));
1766
114
    RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1767
1768
114
    const auto downstream_pipeline_id = cur_pipe->id();
1769
114
    if (!_dag.contains(downstream_pipeline_id)) {
1770
105
        _dag.insert({downstream_pipeline_id, {}});
1771
105
    }
1772
1773
420
    for (int child_id = 0; child_id < tnode.num_children; child_id++) {
1774
306
        PipelinePtr probe_side_pipe = add_pipeline(cur_pipe);
1775
306
        _dag[downstream_pipeline_id].push_back(probe_side_pipe->id());
1776
1777
306
        if (child_id == 0) {
1778
114
            sink_ops.push_back(std::make_shared<SetSinkOperatorX<is_intersect>>(
1779
114
                    child_id, next_sink_operator_id(), op->operator_id(), pool, tnode, descs));
1780
192
        } else {
1781
192
            sink_ops.push_back(std::make_shared<SetProbeSinkOperatorX<is_intersect>>(
1782
192
                    child_id, next_sink_operator_id(), op->operator_id(), pool, tnode, descs));
1783
192
        }
1784
306
        RETURN_IF_ERROR(probe_side_pipe->set_sink(sink_ops.back()));
1785
306
        RETURN_IF_ERROR(probe_side_pipe->sink()->init(tnode, _runtime_state.get()));
1786
        // prepare children pipelines. if any pipeline found this as its father, will use the prepared pipeline to build.
1787
306
        _pipeline_parent_map.push(op->node_id(), probe_side_pipe);
1788
306
    }
1789
1790
114
    return Status::OK();
1791
114
}
_ZN5doris23PipelineFragmentContext39_build_operators_for_set_operation_nodeILb0EEENS_6StatusEPNS_10ObjectPoolERKNS_9TPlanNodeERKNS_13DescriptorTblERSt10shared_ptrINS_13OperatorXBaseEERSB_INS_8PipelineEERSt6vectorISB_INS_21DataSinkOperatorXBaseEESaISK_EE
Line
Count
Source
1764
125
        PipelinePtr& cur_pipe, std::vector<DataSinkOperatorPtr>& sink_ops) {
1765
125
    op.reset(new SetSourceOperatorX<is_intersect>(pool, tnode, next_operator_id(), descs));
1766
125
    RETURN_IF_ERROR(cur_pipe->add_operator(op, _parallel_instances));
1767
1768
125
    const auto downstream_pipeline_id = cur_pipe->id();
1769
125
    if (!_dag.contains(downstream_pipeline_id)) {
1770
117
        _dag.insert({downstream_pipeline_id, {}});
1771
117
    }
1772
1773
390
    for (int child_id = 0; child_id < tnode.num_children; child_id++) {
1774
265
        PipelinePtr probe_side_pipe = add_pipeline(cur_pipe);
1775
265
        _dag[downstream_pipeline_id].push_back(probe_side_pipe->id());
1776
1777
265
        if (child_id == 0) {
1778
125
            sink_ops.push_back(std::make_shared<SetSinkOperatorX<is_intersect>>(
1779
125
                    child_id, next_sink_operator_id(), op->operator_id(), pool, tnode, descs));
1780
140
        } else {
1781
140
            sink_ops.push_back(std::make_shared<SetProbeSinkOperatorX<is_intersect>>(
1782
140
                    child_id, next_sink_operator_id(), op->operator_id(), pool, tnode, descs));
1783
140
        }
1784
265
        RETURN_IF_ERROR(probe_side_pipe->set_sink(sink_ops.back()));
1785
265
        RETURN_IF_ERROR(probe_side_pipe->sink()->init(tnode, _runtime_state.get()));
1786
        // prepare children pipelines. if any pipeline found this as its father, will use the prepared pipeline to build.
1787
265
        _pipeline_parent_map.push(op->node_id(), probe_side_pipe);
1788
265
    }
1789
1790
125
    return Status::OK();
1791
125
}
1792
1793
318k
Status PipelineFragmentContext::submit() {
1794
318k
    if (_submitted) {
1795
0
        return Status::InternalError("submitted");
1796
0
    }
1797
318k
    _submitted = true;
1798
1799
318k
    int submit_tasks = 0;
1800
318k
    Status st;
1801
318k
    auto* scheduler = _query_ctx->get_pipe_exec_scheduler();
1802
992k
    for (auto& task : _tasks) {
1803
1.71M
        for (auto& t : task) {
1804
1.71M
            st = scheduler->submit(t.first);
1805
1.71M
            DBUG_EXECUTE_IF("PipelineFragmentContext.submit.failed",
1806
1.71M
                            { st = Status::Aborted("PipelineFragmentContext.submit.failed"); });
1807
1.71M
            if (!st) {
1808
0
                cancel(Status::InternalError("submit context to executor fail"));
1809
0
                std::lock_guard<std::mutex> l(_task_mutex);
1810
0
                _total_tasks = submit_tasks;
1811
0
                break;
1812
0
            }
1813
1.71M
            submit_tasks++;
1814
1.71M
        }
1815
992k
    }
1816
318k
    if (!st.ok()) {
1817
0
        bool need_remove = false;
1818
0
        {
1819
0
            std::lock_guard<std::mutex> l(_task_mutex);
1820
0
            if (_closed_tasks >= _total_tasks) {
1821
0
                need_remove = _close_fragment_instance();
1822
0
            }
1823
0
        }
1824
        // Call remove_pipeline_context() outside _task_mutex to avoid ABBA deadlock.
1825
0
        if (need_remove) {
1826
0
            _exec_env->fragment_mgr()->remove_pipeline_context({_query_id, _fragment_id});
1827
0
        }
1828
0
        return Status::InternalError("Submit pipeline failed. err = {}, BE: {}", st.to_string(),
1829
0
                                     BackendOptions::get_localhost());
1830
318k
    } else {
1831
318k
        return st;
1832
318k
    }
1833
318k
}
1834
1835
0
void PipelineFragmentContext::print_profile(const std::string& extra_info) {
1836
0
    if (_runtime_state->enable_profile()) {
1837
0
        std::stringstream ss;
1838
0
        for (auto runtime_profile_ptr : _runtime_state->pipeline_id_to_profile()) {
1839
0
            runtime_profile_ptr->pretty_print(&ss);
1840
0
        }
1841
1842
0
        if (_runtime_state->load_channel_profile()) {
1843
0
            _runtime_state->load_channel_profile()->pretty_print(&ss);
1844
0
        }
1845
1846
0
        auto profile_str =
1847
0
                fmt::format("Query {} fragment {} {}, profile, {}", print_id(this->_query_id),
1848
0
                            this->_fragment_id, extra_info, ss.str());
1849
0
        LOG_LONG_STRING(INFO, profile_str);
1850
0
    }
1851
0
}
1852
// If all pipeline tasks binded to the fragment instance are finished, then we could
1853
// close the fragment instance.
1854
// Returns true if the caller should call remove_pipeline_context() **after** releasing
1855
// _task_mutex. We must not call remove_pipeline_context() here because it acquires
1856
// _pipeline_map's shard lock, and this function is called while _task_mutex is held.
1857
// Acquiring _pipeline_map while holding _task_mutex creates an ABBA deadlock with
1858
// dump_pipeline_tasks(), which acquires _pipeline_map first and then _task_mutex
1859
// (via debug_string()).
1860
320k
bool PipelineFragmentContext::_close_fragment_instance() {
1861
320k
    if (_is_fragment_instance_closed) {
1862
0
        return false;
1863
0
    }
1864
320k
    Defer defer_op {[&]() { _is_fragment_instance_closed = true; }};
1865
320k
    _fragment_level_profile->total_time_counter()->update(_fragment_watcher.elapsed_time());
1866
320k
    if (!_need_notify_close) {
1867
317k
        auto st = send_report(true);
1868
317k
        if (!st) {
1869
0
            LOG(WARNING) << fmt::format("Failed to send report for query {}, fragment {}: {}",
1870
0
                                        print_id(_query_id), _fragment_id, st.to_string());
1871
0
        }
1872
317k
    }
1873
    // Print profile content in info log is a tempoeray solution for stream load and external_connector.
1874
    // Since stream load does not have someting like coordinator on FE, so
1875
    // backend can not report profile to FE, ant its profile can not be shown
1876
    // in the same way with other query. So we print the profile content to info log.
1877
1878
320k
    if (_runtime_state->enable_profile() &&
1879
320k
        (_query_ctx->get_query_source() == QuerySource::STREAM_LOAD ||
1880
2.40k
         _query_ctx->get_query_source() == QuerySource::EXTERNAL_CONNECTOR ||
1881
2.40k
         _query_ctx->get_query_source() == QuerySource::GROUP_COMMIT_LOAD)) {
1882
0
        std::stringstream ss;
1883
        // Compute the _local_time_percent before pretty_print the runtime_profile
1884
        // Before add this operation, the print out like that:
1885
        // UNION_NODE (id=0):(Active: 56.720us, non-child: 00.00%)
1886
        // After add the operation, the print out like that:
1887
        // UNION_NODE (id=0):(Active: 56.720us, non-child: 82.53%)
1888
        // We can easily know the exec node execute time without child time consumed.
1889
0
        for (auto runtime_profile_ptr : _runtime_state->pipeline_id_to_profile()) {
1890
0
            runtime_profile_ptr->pretty_print(&ss);
1891
0
        }
1892
1893
0
        if (_runtime_state->load_channel_profile()) {
1894
0
            _runtime_state->load_channel_profile()->pretty_print(&ss);
1895
0
        }
1896
1897
0
        LOG_INFO("Query {} fragment {} profile:\n {}", print_id(_query_id), _fragment_id, ss.str());
1898
0
    }
1899
1900
320k
    if (_query_ctx->enable_profile()) {
1901
2.40k
        _query_ctx->add_fragment_profile(_fragment_id, collect_realtime_profile(),
1902
2.40k
                                         collect_realtime_load_channel_profile());
1903
2.40k
    }
1904
1905
    // Return whether the caller needs to remove from the pipeline map.
1906
    // The caller must do this after releasing _task_mutex.
1907
320k
    return !_need_notify_close;
1908
320k
}
1909
1910
1.70M
void PipelineFragmentContext::decrement_running_task(PipelineId pipeline_id) {
1911
    // If all tasks of this pipeline has been closed, upstream tasks is never needed, and we just make those runnable here
1912
1.70M
    DCHECK(_pip_id_to_pipeline.contains(pipeline_id));
1913
1.70M
    if (_pip_id_to_pipeline[pipeline_id]->close_task()) {
1914
508k
        if (_dag.contains(pipeline_id)) {
1915
289k
            for (auto dep : _dag[pipeline_id]) {
1916
289k
                _pip_id_to_pipeline[dep]->make_all_runnable(pipeline_id);
1917
289k
            }
1918
229k
        }
1919
508k
    }
1920
1.70M
    bool need_remove = false;
1921
1.70M
    {
1922
1.70M
        std::lock_guard<std::mutex> l(_task_mutex);
1923
1.70M
        ++_closed_tasks;
1924
1.70M
        if (_closed_tasks >= _total_tasks) {
1925
320k
            need_remove = _close_fragment_instance();
1926
320k
        }
1927
1.70M
    }
1928
    // Call remove_pipeline_context() outside _task_mutex to avoid ABBA deadlock.
1929
1.70M
    if (need_remove) {
1930
317k
        _exec_env->fragment_mgr()->remove_pipeline_context({_query_id, _fragment_id});
1931
317k
    }
1932
1.70M
}
1933
1934
42.4k
std::string PipelineFragmentContext::get_load_error_url() {
1935
42.4k
    if (const auto& str = _runtime_state->get_error_log_file_path(); !str.empty()) {
1936
0
        return to_load_error_http_path(str);
1937
0
    }
1938
119k
    for (auto& tasks : _tasks) {
1939
207k
        for (auto& task : tasks) {
1940
207k
            if (const auto& str = task.second->get_error_log_file_path(); !str.empty()) {
1941
150
                return to_load_error_http_path(str);
1942
150
            }
1943
207k
        }
1944
119k
    }
1945
42.2k
    return "";
1946
42.4k
}
1947
1948
42.3k
std::string PipelineFragmentContext::get_first_error_msg() {
1949
42.3k
    if (const auto& str = _runtime_state->get_first_error_msg(); !str.empty()) {
1950
0
        return str;
1951
0
    }
1952
119k
    for (auto& tasks : _tasks) {
1953
208k
        for (auto& task : tasks) {
1954
208k
            if (const auto& str = task.second->get_first_error_msg(); !str.empty()) {
1955
150
                return str;
1956
150
            }
1957
208k
        }
1958
119k
    }
1959
42.2k
    return "";
1960
42.3k
}
1961
1962
321k
Status PipelineFragmentContext::send_report(bool done) {
1963
321k
    Status exec_status = _query_ctx->exec_status();
1964
    // If plan is done successfully, but _is_report_success is false,
1965
    // no need to send report.
1966
    // Load will set _is_report_success to true because load wants to know
1967
    // the process.
1968
321k
    if (!_is_report_success && done && exec_status.ok()) {
1969
284k
        return Status::OK();
1970
284k
    }
1971
1972
    // If both _is_report_success and _is_report_on_cancel are false,
1973
    // which means no matter query is success or failed, no report is needed.
1974
    // This may happen when the query limit reached and
1975
    // a internal cancellation being processed
1976
    // When limit is reached the fragment is also cancelled, but _is_report_on_cancel will
1977
    // be set to false, to avoid sending fault report to FE.
1978
37.0k
    if (!_is_report_success && !_is_report_on_cancel) {
1979
107
        if (done) {
1980
            // if done is true, which means the query is finished successfully, we can safely close the fragment instance without sending report to FE, and just return OK status here.
1981
107
            return Status::OK();
1982
107
        }
1983
0
        return Status::NeedSendAgain("");
1984
107
    }
1985
1986
36.9k
    std::vector<RuntimeState*> runtime_states;
1987
1988
90.8k
    for (auto& tasks : _tasks) {
1989
144k
        for (auto& task : tasks) {
1990
144k
            runtime_states.push_back(task.second.get());
1991
144k
        }
1992
90.8k
    }
1993
1994
36.9k
    std::string load_eror_url = _query_ctx->get_load_error_url().empty()
1995
36.9k
                                        ? get_load_error_url()
1996
18.4E
                                        : _query_ctx->get_load_error_url();
1997
36.9k
    std::string first_error_msg = _query_ctx->get_first_error_msg().empty()
1998
36.9k
                                          ? get_first_error_msg()
1999
18.4E
                                          : _query_ctx->get_first_error_msg();
2000
2001
36.9k
    ReportStatusRequest req {.status = exec_status,
2002
36.9k
                             .runtime_states = runtime_states,
2003
36.9k
                             .done = done || !exec_status.ok(),
2004
36.9k
                             .coord_addr = _query_ctx->coord_addr,
2005
36.9k
                             .query_id = _query_id,
2006
36.9k
                             .fragment_id = _fragment_id,
2007
36.9k
                             .fragment_instance_id = TUniqueId(),
2008
36.9k
                             .backend_num = -1,
2009
36.9k
                             .runtime_state = _runtime_state.get(),
2010
36.9k
                             .load_error_url = load_eror_url,
2011
36.9k
                             .first_error_msg = first_error_msg,
2012
36.9k
                             .cancel_fn = [this](const Status& reason) { cancel(reason); }};
2013
2014
36.9k
    return _report_status_cb(
2015
36.9k
            req, std::dynamic_pointer_cast<PipelineFragmentContext>(shared_from_this()));
2016
37.0k
}
2017
2018
0
size_t PipelineFragmentContext::get_revocable_size(bool* has_running_task) const {
2019
0
    size_t res = 0;
2020
    // _tasks will be cleared during ~PipelineFragmentContext, so that it's safe
2021
    // here to traverse the vector.
2022
0
    for (const auto& task_instances : _tasks) {
2023
0
        for (const auto& task : task_instances) {
2024
0
            if (task.first->is_running()) {
2025
0
                LOG_EVERY_N(INFO, 50) << "Query: " << print_id(_query_id)
2026
0
                                      << " is running, task: " << (void*)task.first.get()
2027
0
                                      << ", is_running: " << task.first->is_running();
2028
0
                *has_running_task = true;
2029
0
                return 0;
2030
0
            }
2031
2032
0
            size_t revocable_size = task.first->get_revocable_size();
2033
0
            if (revocable_size >= SpillFile::MIN_SPILL_WRITE_BATCH_MEM) {
2034
0
                res += revocable_size;
2035
0
            }
2036
0
        }
2037
0
    }
2038
0
    return res;
2039
0
}
2040
2041
0
std::vector<PipelineTask*> PipelineFragmentContext::get_revocable_tasks() const {
2042
0
    std::vector<PipelineTask*> revocable_tasks;
2043
0
    for (const auto& task_instances : _tasks) {
2044
0
        for (const auto& task : task_instances) {
2045
0
            size_t revocable_size_ = task.first->get_revocable_size();
2046
2047
0
            if (revocable_size_ >= SpillFile::MIN_SPILL_WRITE_BATCH_MEM) {
2048
0
                revocable_tasks.emplace_back(task.first.get());
2049
0
            }
2050
0
        }
2051
0
    }
2052
0
    return revocable_tasks;
2053
0
}
2054
2055
74
std::string PipelineFragmentContext::debug_string() {
2056
74
    std::lock_guard<std::mutex> l(_task_mutex);
2057
74
    fmt::memory_buffer debug_string_buffer;
2058
74
    fmt::format_to(debug_string_buffer,
2059
74
                   "PipelineFragmentContext Info: _closed_tasks={}, _total_tasks={}, "
2060
74
                   "need_notify_close={}, fragment_id={}, _rec_cte_stage={}\n",
2061
74
                   _closed_tasks, _total_tasks, _need_notify_close, _fragment_id, _rec_cte_stage);
2062
304
    for (size_t j = 0; j < _tasks.size(); j++) {
2063
230
        fmt::format_to(debug_string_buffer, "Tasks in instance {}:\n", j);
2064
556
        for (size_t i = 0; i < _tasks[j].size(); i++) {
2065
326
            fmt::format_to(debug_string_buffer, "Task {}: {}\n", i,
2066
326
                           _tasks[j][i].first->debug_string());
2067
326
        }
2068
230
    }
2069
2070
74
    return fmt::to_string(debug_string_buffer);
2071
74
}
2072
2073
std::vector<std::shared_ptr<TRuntimeProfileTree>>
2074
2.40k
PipelineFragmentContext::collect_realtime_profile() const {
2075
2.40k
    std::vector<std::shared_ptr<TRuntimeProfileTree>> res;
2076
2077
    // we do not have mutex to protect pipeline_id_to_profile
2078
    // so we need to make sure this funciton is invoked after fragment context
2079
    // has already been prepared.
2080
2.40k
    if (!_prepared) {
2081
0
        std::string msg =
2082
0
                "Query " + print_id(_query_id) + " collecting profile, but its not prepared";
2083
0
        DCHECK(false) << msg;
2084
0
        LOG_ERROR(msg);
2085
0
        return res;
2086
0
    }
2087
2088
    // Make sure first profile is fragment level profile
2089
2.40k
    auto fragment_profile = std::make_shared<TRuntimeProfileTree>();
2090
2.40k
    _fragment_level_profile->to_thrift(fragment_profile.get(), _runtime_state->profile_level());
2091
2.40k
    res.push_back(fragment_profile);
2092
2093
    // pipeline_id_to_profile is initialized in prepare stage
2094
4.69k
    for (auto pipeline_profile : _runtime_state->pipeline_id_to_profile()) {
2095
4.69k
        auto profile_ptr = std::make_shared<TRuntimeProfileTree>();
2096
4.69k
        pipeline_profile->to_thrift(profile_ptr.get(), _runtime_state->profile_level());
2097
4.69k
        res.push_back(profile_ptr);
2098
4.69k
    }
2099
2100
2.40k
    return res;
2101
2.40k
}
2102
2103
std::shared_ptr<TRuntimeProfileTree>
2104
2.40k
PipelineFragmentContext::collect_realtime_load_channel_profile() const {
2105
    // we do not have mutex to protect pipeline_id_to_profile
2106
    // so we need to make sure this funciton is invoked after fragment context
2107
    // has already been prepared.
2108
2.40k
    if (!_prepared) {
2109
0
        std::string msg =
2110
0
                "Query " + print_id(_query_id) + " collecting profile, but its not prepared";
2111
0
        DCHECK(false) << msg;
2112
0
        LOG_ERROR(msg);
2113
0
        return nullptr;
2114
0
    }
2115
2116
8.75k
    for (const auto& tasks : _tasks) {
2117
17.5k
        for (const auto& task : tasks) {
2118
17.5k
            if (task.second->load_channel_profile() == nullptr) {
2119
0
                continue;
2120
0
            }
2121
2122
17.5k
            auto tmp_load_channel_profile = std::make_shared<TRuntimeProfileTree>();
2123
2124
17.5k
            task.second->load_channel_profile()->to_thrift(tmp_load_channel_profile.get(),
2125
17.5k
                                                           _runtime_state->profile_level());
2126
17.5k
            _runtime_state->load_channel_profile()->update(*tmp_load_channel_profile);
2127
17.5k
        }
2128
8.75k
    }
2129
2130
2.40k
    auto load_channel_profile = std::make_shared<TRuntimeProfileTree>();
2131
2.40k
    _runtime_state->load_channel_profile()->to_thrift(load_channel_profile.get(),
2132
2.40k
                                                      _runtime_state->profile_level());
2133
2.40k
    return load_channel_profile;
2134
2.40k
}
2135
2136
// Collect runtime filter IDs registered by all tasks in this PFC.
2137
// Used during recursive CTE stage transitions to know which filters to deregister
2138
// before creating the new PFC for the next recursion round.
2139
// Called from rerun_fragment(wait_for_destroy) while tasks are still closing.
2140
// Thread safety: safe because _tasks is structurally immutable after prepare() —
2141
// the vector sizes do not change, and individual RuntimeState filter sets are
2142
// written only during open() which has completed by the time we reach rerun.
2143
3.28k
std::set<int> PipelineFragmentContext::get_deregister_runtime_filter() const {
2144
3.28k
    std::set<int> result;
2145
12.1k
    for (const auto& _task : _tasks) {
2146
20.1k
        for (const auto& task : _task) {
2147
20.1k
            auto set = task.first->runtime_state()->get_deregister_runtime_filter();
2148
20.1k
            result.merge(set);
2149
20.1k
        }
2150
12.1k
    }
2151
3.28k
    if (_runtime_state) {
2152
3.28k
        auto set = _runtime_state->get_deregister_runtime_filter();
2153
3.28k
        result.merge(set);
2154
3.28k
    }
2155
3.28k
    return result;
2156
3.28k
}
2157
2158
321k
void PipelineFragmentContext::_release_resource() {
2159
321k
    std::lock_guard<std::mutex> l(_task_mutex);
2160
    // The memory released by the query end is recorded in the query mem tracker.
2161
321k
    SCOPED_SWITCH_THREAD_MEM_TRACKER_LIMITER(_query_ctx->query_mem_tracker());
2162
321k
    auto st = _query_ctx->exec_status();
2163
993k
    for (auto& _task : _tasks) {
2164
993k
        if (!_task.empty()) {
2165
993k
            _call_back(_task.front().first->runtime_state(), &st);
2166
993k
        }
2167
993k
    }
2168
321k
    _tasks.clear();
2169
321k
    _dag.clear();
2170
321k
    _pip_id_to_pipeline.clear();
2171
321k
    _pipelines.clear();
2172
321k
    _sink.reset();
2173
321k
    _root_op.reset();
2174
321k
    _runtime_filter_mgr_map.clear();
2175
321k
    _op_id_to_shared_state.clear();
2176
321k
}
2177
2178
#include "common/compile_check_end.h"
2179
} // namespace doris