Coverage Report

Created: 2026-03-17 03:12

next uncovered line (L), next uncovered region (R), next uncovered branch (B)
be/src/service/http/action/stream_load.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 "service/http/action/stream_load.h"
19
20
// use string iequal
21
#include <event2/buffer.h>
22
#include <event2/http.h>
23
#include <gen_cpp/FrontendService.h>
24
#include <gen_cpp/FrontendService_types.h>
25
#include <gen_cpp/HeartbeatService_types.h>
26
#include <gen_cpp/PaloInternalService_types.h>
27
#include <gen_cpp/PlanNodes_types.h>
28
#include <gen_cpp/Types_types.h>
29
#include <sys/time.h>
30
#include <thrift/protocol/TDebugProtocol.h>
31
32
#include <algorithm>
33
#include <cstdint>
34
#include <cstdlib>
35
#include <ctime>
36
#include <functional>
37
#include <future>
38
#include <sstream>
39
#include <stdexcept>
40
#include <utility>
41
42
#include "cloud/config.h"
43
#include "common/config.h"
44
#include "common/consts.h"
45
#include "common/logging.h"
46
#include "common/metrics/doris_metrics.h"
47
#include "common/metrics/metrics.h"
48
#include "common/status.h"
49
#include "common/utils.h"
50
#include "io/fs/stream_load_pipe.h"
51
#include "load/group_commit/group_commit_mgr.h"
52
#include "load/load_path_mgr.h"
53
#include "load/message_body_sink.h"
54
#include "load/stream_load/new_load_stream_mgr.h"
55
#include "load/stream_load/stream_load_context.h"
56
#include "load/stream_load/stream_load_executor.h"
57
#include "load/stream_load/stream_load_recorder.h"
58
#include "runtime/exec_env.h"
59
#include "service/http/http_channel.h"
60
#include "service/http/http_common.h"
61
#include "service/http/http_headers.h"
62
#include "service/http/http_request.h"
63
#include "service/http/utils.h"
64
#include "storage/storage_engine.h"
65
#include "util/byte_buffer.h"
66
#include "util/client_cache.h"
67
#include "util/load_util.h"
68
#include "util/string_util.h"
69
#include "util/thrift_rpc_helper.h"
70
#include "util/time.h"
71
#include "util/uid_util.h"
72
#include "util/url_coding.h"
73
74
namespace doris {
75
using namespace ErrorCode;
76
77
DEFINE_COUNTER_METRIC_PROTOTYPE_2ARG(streaming_load_requests_total, MetricUnit::REQUESTS);
78
DEFINE_COUNTER_METRIC_PROTOTYPE_2ARG(streaming_load_duration_ms, MetricUnit::MILLISECONDS);
79
DEFINE_GAUGE_METRIC_PROTOTYPE_2ARG(streaming_load_current_processing, MetricUnit::REQUESTS);
80
81
bvar::LatencyRecorder g_stream_load_receive_data_latency_ms("stream_load_receive_data_latency_ms");
82
bvar::LatencyRecorder g_stream_load_commit_and_publish_latency_ms("stream_load",
83
                                                                  "commit_and_publish_ms");
84
85
static constexpr size_t MIN_CHUNK_SIZE = 64 * 1024;
86
static const std::string CHUNK = "chunked";
87
static const std::string OFF_MODE = "off_mode";
88
static const std::string SYNC_MODE = "sync_mode";
89
static const std::string ASYNC_MODE = "async_mode";
90
91
#ifdef BE_TEST
92
TStreamLoadPutResult k_stream_load_put_result;
93
#endif
94
95
1
StreamLoadAction::StreamLoadAction(ExecEnv* exec_env) : _exec_env(exec_env) {
96
1
    _stream_load_entity =
97
1
            DorisMetrics::instance()->metric_registry()->register_entity("stream_load");
98
1
    INT_COUNTER_METRIC_REGISTER(_stream_load_entity, streaming_load_requests_total);
99
1
    INT_COUNTER_METRIC_REGISTER(_stream_load_entity, streaming_load_duration_ms);
100
1
    INT_GAUGE_METRIC_REGISTER(_stream_load_entity, streaming_load_current_processing);
101
1
}
102
103
1
StreamLoadAction::~StreamLoadAction() {
104
1
    DorisMetrics::instance()->metric_registry()->deregister_entity(_stream_load_entity);
105
1
}
106
107
0
void StreamLoadAction::handle(HttpRequest* req) {
108
0
    std::shared_ptr<StreamLoadContext> ctx =
109
0
            std::static_pointer_cast<StreamLoadContext>(req->handler_ctx());
110
0
    if (ctx == nullptr) {
111
0
        return;
112
0
    }
113
114
0
    {
115
0
        std::unique_lock<std::mutex> lock1(ctx->_send_reply_lock);
116
0
        ctx->_can_send_reply = true;
117
0
        ctx->_can_send_reply_cv.notify_all();
118
0
    }
119
120
    // status already set to fail
121
0
    if (ctx->status.ok()) {
122
0
        ctx->status = _handle(ctx, req);
123
0
        if (!ctx->status.ok() && !ctx->status.is<PUBLISH_TIMEOUT>()) {
124
0
            _send_reply(ctx, req);
125
0
        }
126
0
    }
127
0
}
128
129
0
Status StreamLoadAction::_handle(std::shared_ptr<StreamLoadContext> ctx, HttpRequest* req) {
130
0
    if (ctx->body_bytes > 0 && ctx->receive_bytes != ctx->body_bytes) {
131
0
        LOG(WARNING) << "recevie body don't equal with body bytes, body_bytes=" << ctx->body_bytes
132
0
                     << ", receive_bytes=" << ctx->receive_bytes << ", id=" << ctx->id;
133
0
        return Status::Error<ErrorCode::NETWORK_ERROR>("receive body don't equal with body bytes");
134
0
    }
135
136
    // if we use non-streaming, MessageBodyFileSink.finish will close the file
137
0
    RETURN_IF_ERROR(ctx->body_sink->finish());
138
0
    if (!ctx->use_streaming) {
139
        // we need to close file first, then execute_plan_fragment here
140
0
        ctx->body_sink.reset();
141
0
        TPipelineFragmentParamsList mocked;
142
0
        RETURN_IF_ERROR(_exec_env->stream_load_executor()->execute_plan_fragment(
143
0
                ctx, mocked,
144
0
                [req, this](std::shared_ptr<StreamLoadContext> ctx) { _on_finish(ctx, req); }));
145
0
    }
146
147
0
    return Status::OK();
148
0
}
149
150
0
void StreamLoadAction::_on_finish(std::shared_ptr<StreamLoadContext> ctx, HttpRequest* req) {
151
0
    ctx->status = ctx->load_status_future.get();
152
0
    if (ctx->status.ok()) {
153
0
        if (ctx->group_commit) {
154
0
            LOG(INFO) << "skip commit because this is group commit, pipe_id="
155
0
                      << ctx->id.to_string();
156
0
        } else if (ctx->two_phase_commit) {
157
0
            int64_t pre_commit_start_time = MonotonicNanos();
158
0
            ctx->status = _exec_env->stream_load_executor()->pre_commit_txn(ctx.get());
159
0
            ctx->pre_commit_txn_cost_nanos = MonotonicNanos() - pre_commit_start_time;
160
0
        } else {
161
            // If put file success we need commit this load
162
0
            int64_t commit_and_publish_start_time = MonotonicNanos();
163
0
            ctx->status = _exec_env->stream_load_executor()->commit_txn(ctx.get());
164
0
            ctx->commit_and_publish_txn_cost_nanos =
165
0
                    MonotonicNanos() - commit_and_publish_start_time;
166
0
            g_stream_load_commit_and_publish_latency_ms
167
0
                    << ctx->commit_and_publish_txn_cost_nanos / 1000000;
168
0
        }
169
0
    }
170
0
    _send_reply(ctx, req);
171
0
}
172
173
0
void StreamLoadAction::_send_reply(std::shared_ptr<StreamLoadContext> ctx, HttpRequest* req) {
174
0
    std::unique_lock<std::mutex> lock1(ctx->_send_reply_lock);
175
    // 1. _can_send_reply: ensure `send_reply` is invoked only after on_header/handle complete,
176
    //    avoid client errors (e.g., broken pipe).
177
    // 2. _finish_send_reply: Prevent duplicate reply sending; skip reply if HTTP request is canceled
178
    //    due to long import execution time.
179
0
    while (!ctx->_finish_send_reply && !ctx->_can_send_reply) {
180
0
        ctx->_can_send_reply_cv.wait(lock1);
181
0
    }
182
0
    if (ctx->_finish_send_reply) {
183
0
        return;
184
0
    }
185
0
    DCHECK(ctx->_can_send_reply);
186
0
    ctx->_finish_send_reply = true;
187
0
    ctx->_can_send_reply_cv.notify_all();
188
0
    ctx->load_cost_millis = UnixMillis() - ctx->start_millis;
189
190
0
    if (!ctx->status.ok() && !ctx->status.is<PUBLISH_TIMEOUT>()) {
191
0
        LOG(WARNING) << "handle streaming load failed, id=" << ctx->id
192
0
                     << ", errmsg=" << ctx->status;
193
0
        if (ctx->need_rollback) {
194
0
            _exec_env->stream_load_executor()->rollback_txn(ctx.get());
195
0
            ctx->need_rollback = false;
196
0
        }
197
0
        if (ctx->body_sink != nullptr) {
198
0
            ctx->body_sink->cancel(ctx->status.to_string());
199
0
        }
200
0
    }
201
202
0
    auto str = ctx->to_json();
203
    // add new line at end
204
0
    str = str + '\n';
205
206
#ifndef BE_TEST
207
    if (config::enable_stream_load_record || config::enable_stream_load_record_to_audit_log_table) {
208
        if (req->header(HTTP_SKIP_RECORD_TO_AUDIT_LOG_TABLE).empty()) {
209
            str = ctx->prepare_stream_load_record(str);
210
            _save_stream_load_record(ctx, str);
211
        }
212
    }
213
#endif
214
215
0
    HttpChannel::send_reply(req, str);
216
217
0
    LOG(INFO) << "finished to execute stream load. label=" << ctx->label
218
0
              << ", txn_id=" << ctx->txn_id << ", query_id=" << ctx->id
219
0
              << ", load_cost_ms=" << ctx->load_cost_millis << ", receive_data_cost_ms="
220
0
              << (ctx->receive_and_read_data_cost_nanos - ctx->read_data_cost_nanos) / 1000000
221
0
              << ", read_data_cost_ms=" << ctx->read_data_cost_nanos / 1000000
222
0
              << ", write_data_cost_ms=" << ctx->write_data_cost_nanos / 1000000
223
0
              << ", commit_and_publish_txn_cost_ms="
224
0
              << ctx->commit_and_publish_txn_cost_nanos / 1000000
225
0
              << ", number_total_rows=" << ctx->number_total_rows
226
0
              << ", number_loaded_rows=" << ctx->number_loaded_rows
227
0
              << ", receive_bytes=" << ctx->receive_bytes << ", loaded_bytes=" << ctx->loaded_bytes
228
0
              << ", error_url=" << ctx->error_url;
229
230
    // update statistics
231
0
    streaming_load_requests_total->increment(1);
232
0
    streaming_load_duration_ms->increment(ctx->load_cost_millis);
233
0
    if (!ctx->data_saved_path.empty()) {
234
0
        _exec_env->load_path_mgr()->clean_tmp_files(ctx->data_saved_path);
235
0
    }
236
0
}
237
238
0
int StreamLoadAction::on_header(HttpRequest* req) {
239
0
    req->mark_send_reply();
240
241
0
    streaming_load_current_processing->increment(1);
242
243
0
    std::shared_ptr<StreamLoadContext> ctx = std::make_shared<StreamLoadContext>(_exec_env);
244
0
    req->set_handler_ctx(ctx);
245
246
0
    ctx->load_type = TLoadType::MANUL_LOAD;
247
0
    ctx->load_src_type = TLoadSourceType::RAW;
248
249
0
    url_decode(req->param(HTTP_DB_KEY), &ctx->db);
250
0
    url_decode(req->param(HTTP_TABLE_KEY), &ctx->table);
251
0
    ctx->label = req->header(HTTP_LABEL_KEY);
252
0
    ctx->two_phase_commit = req->header(HTTP_TWO_PHASE_COMMIT) == "true";
253
0
    Status st = _handle_group_commit(req, ctx);
254
0
    if (!ctx->group_commit && ctx->label.empty()) {
255
0
        ctx->label = generate_uuid_string();
256
0
    }
257
258
0
    LOG(INFO) << "new income streaming load request." << ctx->brief() << ", db=" << ctx->db
259
0
              << ", tbl=" << ctx->table << ", group_commit=" << ctx->group_commit
260
0
              << ", group_commit_mode=" << ctx->group_commit_mode
261
0
              << ", HTTP headers=" << req->get_all_headers();
262
0
    ctx->begin_receive_and_read_data_cost_nanos = MonotonicNanos();
263
264
0
    if (st.ok()) {
265
0
        st = _on_header(req, ctx);
266
0
        LOG(INFO) << "finished to handle HTTP header, " << ctx->brief();
267
0
    }
268
0
    if (!st.ok()) {
269
0
        ctx->status = std::move(st);
270
0
        {
271
0
            std::unique_lock<std::mutex> lock1(ctx->_send_reply_lock);
272
0
            ctx->_can_send_reply = true;
273
0
            ctx->_can_send_reply_cv.notify_all();
274
0
        }
275
0
        _send_reply(ctx, req);
276
0
        return -1;
277
0
    }
278
0
    return 0;
279
0
}
280
281
0
Status StreamLoadAction::_on_header(HttpRequest* http_req, std::shared_ptr<StreamLoadContext> ctx) {
282
    // auth information
283
0
    if (!parse_basic_auth(*http_req, &ctx->auth)) {
284
0
        LOG(WARNING) << "parse basic authorization failed." << ctx->brief();
285
0
        return Status::NotAuthorized("no valid Basic authorization");
286
0
    }
287
288
    // get format of this put
289
0
    std::string format_str = http_req->header(HTTP_FORMAT_KEY);
290
0
    if (iequal(format_str, BeConsts::CSV_WITH_NAMES) ||
291
0
        iequal(format_str, BeConsts::CSV_WITH_NAMES_AND_TYPES)) {
292
0
        ctx->header_type = format_str;
293
        //treat as CSV
294
0
        format_str = BeConsts::CSV;
295
0
    }
296
0
    LoadUtil::parse_format(format_str, http_req->header(HTTP_COMPRESS_TYPE), &ctx->format,
297
0
                           &ctx->compress_type);
298
0
    if (ctx->format == TFileFormatType::FORMAT_UNKNOWN) {
299
0
        return Status::Error<ErrorCode::DATA_FILE_TYPE_ERROR>("unknown data format, format={}",
300
0
                                                              http_req->header(HTTP_FORMAT_KEY));
301
0
    }
302
303
    // check content length
304
0
    ctx->body_bytes = 0;
305
0
    size_t csv_max_body_bytes = config::streaming_load_max_mb * 1024 * 1024;
306
0
    size_t json_max_body_bytes = config::streaming_load_json_max_mb * 1024 * 1024;
307
0
    bool read_json_by_line = false;
308
0
    if (!http_req->header(HTTP_READ_JSON_BY_LINE).empty()) {
309
0
        if (iequal(http_req->header(HTTP_READ_JSON_BY_LINE), "true")) {
310
0
            read_json_by_line = true;
311
0
        }
312
0
    }
313
0
    if (!http_req->header(HttpHeaders::CONTENT_LENGTH).empty()) {
314
0
        try {
315
0
            ctx->body_bytes = std::stol(http_req->header(HttpHeaders::CONTENT_LENGTH));
316
0
        } catch (const std::exception& e) {
317
0
            return Status::InvalidArgument("invalid HTTP header CONTENT_LENGTH={}: {}",
318
0
                                           http_req->header(HttpHeaders::CONTENT_LENGTH), e.what());
319
0
        }
320
        // json max body size
321
0
        if ((ctx->format == TFileFormatType::FORMAT_JSON) &&
322
0
            (ctx->body_bytes > json_max_body_bytes) && !read_json_by_line) {
323
0
            return Status::Error<ErrorCode::EXCEEDED_LIMIT>(
324
0
                    "json body size {} exceed BE's conf `streaming_load_json_max_mb` {}. increase "
325
0
                    "it if you are sure this load is reasonable",
326
0
                    ctx->body_bytes, json_max_body_bytes);
327
0
        }
328
        // csv max body size
329
0
        else if (ctx->body_bytes > csv_max_body_bytes) {
330
0
            LOG(WARNING) << "body exceed max size." << ctx->brief();
331
0
            return Status::Error<ErrorCode::EXCEEDED_LIMIT>(
332
0
                    "body size {} exceed BE's conf `streaming_load_max_mb` {}. increase it if you "
333
0
                    "are sure this load is reasonable",
334
0
                    ctx->body_bytes, csv_max_body_bytes);
335
0
        }
336
0
    } else {
337
#ifndef BE_TEST
338
        evhttp_connection_set_max_body_size(
339
                evhttp_request_get_connection(http_req->get_evhttp_request()), csv_max_body_bytes);
340
#endif
341
0
    }
342
343
0
    if (!http_req->header(HttpHeaders::TRANSFER_ENCODING).empty()) {
344
0
        if (http_req->header(HttpHeaders::TRANSFER_ENCODING).find(CHUNK) != std::string::npos) {
345
0
            ctx->is_chunked_transfer = true;
346
0
        }
347
0
    }
348
0
    if (UNLIKELY((http_req->header(HttpHeaders::CONTENT_LENGTH).empty() &&
349
0
                  !ctx->is_chunked_transfer))) {
350
0
        LOG(WARNING) << "content_length is empty and transfer-encoding!=chunked, please set "
351
0
                        "content_length or transfer-encoding=chunked";
352
0
        return Status::InvalidArgument(
353
0
                "content_length is empty and transfer-encoding!=chunked, please set content_length "
354
0
                "or transfer-encoding=chunked");
355
0
    } else if (UNLIKELY(!http_req->header(HttpHeaders::CONTENT_LENGTH).empty() &&
356
0
                        ctx->is_chunked_transfer)) {
357
0
        LOG(WARNING) << "please do not set both content_length and transfer-encoding";
358
0
        return Status::InvalidArgument(
359
0
                "please do not set both content_length and transfer-encoding");
360
0
    }
361
362
0
    if (!http_req->header(HTTP_TIMEOUT).empty()) {
363
0
        ctx->timeout_second = DORIS_TRY(safe_stoi(http_req->header(HTTP_TIMEOUT), HTTP_TIMEOUT));
364
0
    }
365
0
    if (!http_req->header(HTTP_COMMENT).empty()) {
366
0
        ctx->load_comment = http_req->header(HTTP_COMMENT);
367
0
    }
368
    // begin transaction
369
0
    if (!ctx->group_commit) {
370
0
        int64_t begin_txn_start_time = MonotonicNanos();
371
0
        RETURN_IF_ERROR(_exec_env->stream_load_executor()->begin_txn(ctx.get()));
372
0
        ctx->begin_txn_cost_nanos = MonotonicNanos() - begin_txn_start_time;
373
0
        if (ctx->group_commit) {
374
0
            RETURN_IF_ERROR(_check_wal_space(ctx->group_commit_mode, ctx->body_bytes));
375
0
        }
376
0
    }
377
378
    // process put file
379
0
    return _process_put(http_req, ctx);
380
0
}
381
382
0
void StreamLoadAction::on_chunk_data(HttpRequest* req) {
383
0
    std::shared_ptr<StreamLoadContext> ctx =
384
0
            std::static_pointer_cast<StreamLoadContext>(req->handler_ctx());
385
0
    if (ctx == nullptr || !ctx->status.ok()) {
386
0
        return;
387
0
    }
388
389
0
    struct evhttp_request* ev_req = req->get_evhttp_request();
390
0
    auto evbuf = evhttp_request_get_input_buffer(ev_req);
391
392
0
    SCOPED_ATTACH_TASK(ExecEnv::GetInstance()->stream_load_pipe_tracker());
393
394
0
    int64_t start_read_data_time = MonotonicNanos();
395
0
    while (evbuffer_get_length(evbuf) > 0) {
396
0
        ByteBufferPtr bb;
397
0
        Status st = ByteBuffer::allocate(128 * 1024, &bb);
398
0
        if (!st.ok()) {
399
0
            ctx->status = st;
400
0
            return;
401
0
        }
402
0
        auto remove_bytes = evbuffer_remove(evbuf, bb->ptr, bb->capacity);
403
0
        bb->pos = remove_bytes;
404
0
        bb->flip();
405
0
        st = ctx->body_sink->append(bb);
406
0
        if (!st.ok()) {
407
0
            LOG(WARNING) << "append body content failed. errmsg=" << st << ", " << ctx->brief();
408
0
            ctx->status = st;
409
0
            return;
410
0
        }
411
0
        ctx->receive_bytes += remove_bytes;
412
0
    }
413
0
    int64_t read_data_time = MonotonicNanos() - start_read_data_time;
414
0
    int64_t last_receive_and_read_data_cost_nanos = ctx->receive_and_read_data_cost_nanos;
415
0
    ctx->read_data_cost_nanos += read_data_time;
416
0
    ctx->receive_and_read_data_cost_nanos =
417
0
            MonotonicNanos() - ctx->begin_receive_and_read_data_cost_nanos;
418
0
    g_stream_load_receive_data_latency_ms
419
0
            << (ctx->receive_and_read_data_cost_nanos - last_receive_and_read_data_cost_nanos -
420
0
                read_data_time) /
421
0
                       1000000;
422
0
}
423
424
0
void StreamLoadAction::free_handler_ctx(std::shared_ptr<void> param) {
425
0
    std::shared_ptr<StreamLoadContext> ctx = std::static_pointer_cast<StreamLoadContext>(param);
426
0
    if (ctx == nullptr) {
427
0
        return;
428
0
    }
429
    // sender is gone, make receiver know it
430
0
    if (ctx->body_sink != nullptr) {
431
0
        ctx->body_sink->cancel("sender is gone");
432
0
    }
433
    // remove stream load context from stream load manager and the resource will be released
434
0
    ctx->exec_env()->new_load_stream_mgr()->remove(ctx->id);
435
0
    streaming_load_current_processing->increment(-1);
436
0
}
437
438
Status StreamLoadAction::_process_put(HttpRequest* http_req,
439
0
                                      std::shared_ptr<StreamLoadContext> ctx) {
440
    // Now we use stream
441
0
    ctx->use_streaming = LoadUtil::is_format_support_streaming(ctx->format);
442
443
    // put request
444
0
    TStreamLoadPutRequest request;
445
0
    set_request_auth(&request, ctx->auth);
446
0
    request.db = ctx->db;
447
0
    request.tbl = ctx->table;
448
0
    request.txnId = ctx->txn_id;
449
0
    request.formatType = ctx->format;
450
0
    request.__set_compress_type(ctx->compress_type);
451
0
    request.__set_header_type(ctx->header_type);
452
0
    request.__set_loadId(ctx->id.to_thrift());
453
0
    if (ctx->use_streaming) {
454
0
        std::shared_ptr<io::StreamLoadPipe> pipe;
455
0
        if (ctx->is_chunked_transfer) {
456
0
            pipe = std::make_shared<io::StreamLoadPipe>(
457
0
                    io::kMaxPipeBufferedBytes /* max_buffered_bytes */);
458
0
            pipe->set_is_chunked_transfer(true);
459
0
        } else {
460
0
            pipe = std::make_shared<io::StreamLoadPipe>(
461
0
                    io::kMaxPipeBufferedBytes /* max_buffered_bytes */,
462
0
                    MIN_CHUNK_SIZE /* min_chunk_size */, ctx->body_bytes /* total_length */);
463
0
        }
464
0
        request.fileType = TFileType::FILE_STREAM;
465
0
        ctx->body_sink = pipe;
466
0
        ctx->pipe = pipe;
467
0
        RETURN_IF_ERROR(_exec_env->new_load_stream_mgr()->put(ctx->id, ctx));
468
0
    } else {
469
0
        RETURN_IF_ERROR(_data_saved_path(http_req, &request.path, ctx->body_bytes));
470
0
        auto file_sink = std::make_shared<MessageBodyFileSink>(request.path);
471
0
        RETURN_IF_ERROR(file_sink->open());
472
0
        request.__isset.path = true;
473
0
        request.fileType = TFileType::FILE_LOCAL;
474
0
        request.__set_file_size(ctx->body_bytes);
475
0
        ctx->body_sink = file_sink;
476
0
        ctx->data_saved_path = request.path;
477
0
    }
478
0
    if (!http_req->header(HTTP_COLUMNS).empty()) {
479
0
        request.__set_columns(http_req->header(HTTP_COLUMNS));
480
0
    }
481
0
    if (!http_req->header(HTTP_WHERE).empty()) {
482
0
        request.__set_where(http_req->header(HTTP_WHERE));
483
0
    }
484
0
    if (!http_req->header(HTTP_COLUMN_SEPARATOR).empty()) {
485
0
        request.__set_columnSeparator(http_req->header(HTTP_COLUMN_SEPARATOR));
486
0
    }
487
0
    if (!http_req->header(HTTP_LINE_DELIMITER).empty()) {
488
0
        request.__set_line_delimiter(http_req->header(HTTP_LINE_DELIMITER));
489
0
    }
490
0
    if (!http_req->header(HTTP_ENCLOSE).empty() && !http_req->header(HTTP_ENCLOSE).empty()) {
491
0
        const auto& enclose_str = http_req->header(HTTP_ENCLOSE);
492
0
        if (enclose_str.length() != 1) {
493
0
            return Status::InvalidArgument("enclose must be single-char, actually is {}",
494
0
                                           enclose_str);
495
0
        }
496
0
        request.__set_enclose(http_req->header(HTTP_ENCLOSE)[0]);
497
0
    }
498
0
    if (!http_req->header(HTTP_ESCAPE).empty() && !http_req->header(HTTP_ESCAPE).empty()) {
499
0
        const auto& escape_str = http_req->header(HTTP_ESCAPE);
500
0
        if (escape_str.length() != 1) {
501
0
            return Status::InvalidArgument("escape must be single-char, actually is {}",
502
0
                                           escape_str);
503
0
        }
504
0
        request.__set_escape(http_req->header(HTTP_ESCAPE)[0]);
505
0
    }
506
0
    if (!http_req->header(HTTP_PARTITIONS).empty()) {
507
0
        request.__set_partitions(http_req->header(HTTP_PARTITIONS));
508
0
        request.__set_isTempPartition(false);
509
0
        if (!http_req->header(HTTP_TEMP_PARTITIONS).empty()) {
510
0
            return Status::InvalidArgument(
511
0
                    "Can not specify both partitions and temporary partitions");
512
0
        }
513
0
    }
514
0
    if (!http_req->header(HTTP_TEMP_PARTITIONS).empty()) {
515
0
        request.__set_partitions(http_req->header(HTTP_TEMP_PARTITIONS));
516
0
        request.__set_isTempPartition(true);
517
0
        if (!http_req->header(HTTP_PARTITIONS).empty()) {
518
0
            return Status::InvalidArgument(
519
0
                    "Can not specify both partitions and temporary partitions");
520
0
        }
521
0
    }
522
0
    if (!http_req->header(HTTP_NEGATIVE).empty() && http_req->header(HTTP_NEGATIVE) == "true") {
523
0
        request.__set_negative(true);
524
0
    } else {
525
0
        request.__set_negative(false);
526
0
    }
527
0
    bool strictMode = false;
528
0
    if (!http_req->header(HTTP_STRICT_MODE).empty()) {
529
0
        if (iequal(http_req->header(HTTP_STRICT_MODE), "false")) {
530
0
            strictMode = false;
531
0
        } else if (iequal(http_req->header(HTTP_STRICT_MODE), "true")) {
532
0
            strictMode = true;
533
0
        } else {
534
0
            return Status::InvalidArgument("Invalid strict mode format. Must be bool type");
535
0
        }
536
0
        request.__set_strictMode(strictMode);
537
0
    }
538
    // timezone first. if not, try system time_zone
539
0
    if (!http_req->header(HTTP_TIMEZONE).empty()) {
540
0
        request.__set_timezone(http_req->header(HTTP_TIMEZONE));
541
0
    } else if (!http_req->header(HTTP_TIME_ZONE).empty()) {
542
0
        request.__set_timezone(http_req->header(HTTP_TIME_ZONE));
543
0
    }
544
0
    if (!http_req->header(HTTP_EXEC_MEM_LIMIT).empty()) {
545
0
        try {
546
0
            request.__set_execMemLimit(std::stoll(http_req->header(HTTP_EXEC_MEM_LIMIT)));
547
0
        } catch (const std::invalid_argument& e) {
548
0
            return Status::InvalidArgument("Invalid mem limit format, {}", e.what());
549
0
        }
550
0
    }
551
0
    if (!http_req->header(HTTP_JSONPATHS).empty()) {
552
0
        request.__set_jsonpaths(http_req->header(HTTP_JSONPATHS));
553
0
    }
554
0
    if (!http_req->header(HTTP_JSONROOT).empty()) {
555
0
        request.__set_json_root(http_req->header(HTTP_JSONROOT));
556
0
    }
557
0
    if (!http_req->header(HTTP_STRIP_OUTER_ARRAY).empty()) {
558
0
        if (iequal(http_req->header(HTTP_STRIP_OUTER_ARRAY), "true")) {
559
0
            request.__set_strip_outer_array(true);
560
0
        } else {
561
0
            request.__set_strip_outer_array(false);
562
0
        }
563
0
    } else {
564
0
        request.__set_strip_outer_array(false);
565
0
    }
566
567
0
    if (!http_req->header(HTTP_READ_JSON_BY_LINE).empty()) {
568
0
        if (iequal(http_req->header(HTTP_READ_JSON_BY_LINE), "true")) {
569
0
            request.__set_read_json_by_line(true);
570
0
        } else {
571
0
            request.__set_read_json_by_line(false);
572
0
        }
573
0
    } else {
574
0
        request.__set_read_json_by_line(false);
575
0
    }
576
577
0
    if (http_req->header(HTTP_READ_JSON_BY_LINE).empty() &&
578
0
        http_req->header(HTTP_STRIP_OUTER_ARRAY).empty()) {
579
0
        request.__set_read_json_by_line(true);
580
0
        request.__set_strip_outer_array(false);
581
0
    }
582
583
0
    if (!http_req->header(HTTP_NUM_AS_STRING).empty()) {
584
0
        if (iequal(http_req->header(HTTP_NUM_AS_STRING), "true")) {
585
0
            request.__set_num_as_string(true);
586
0
        } else {
587
0
            request.__set_num_as_string(false);
588
0
        }
589
0
    } else {
590
0
        request.__set_num_as_string(false);
591
0
    }
592
0
    if (!http_req->header(HTTP_FUZZY_PARSE).empty()) {
593
0
        if (iequal(http_req->header(HTTP_FUZZY_PARSE), "true")) {
594
0
            request.__set_fuzzy_parse(true);
595
0
        } else {
596
0
            request.__set_fuzzy_parse(false);
597
0
        }
598
0
    } else {
599
0
        request.__set_fuzzy_parse(false);
600
0
    }
601
602
0
    if (!http_req->header(HTTP_FUNCTION_COLUMN + "." + HTTP_SEQUENCE_COL).empty()) {
603
0
        request.__set_sequence_col(
604
0
                http_req->header(HTTP_FUNCTION_COLUMN + "." + HTTP_SEQUENCE_COL));
605
0
    }
606
607
0
    if (!http_req->header(HTTP_SEND_BATCH_PARALLELISM).empty()) {
608
0
        int parallelism = DORIS_TRY(safe_stoi(http_req->header(HTTP_SEND_BATCH_PARALLELISM),
609
0
                                              HTTP_SEND_BATCH_PARALLELISM));
610
0
        request.__set_send_batch_parallelism(parallelism);
611
0
    }
612
613
0
    if (!http_req->header(HTTP_LOAD_TO_SINGLE_TABLET).empty()) {
614
0
        if (iequal(http_req->header(HTTP_LOAD_TO_SINGLE_TABLET), "true")) {
615
0
            request.__set_load_to_single_tablet(true);
616
0
        } else {
617
0
            request.__set_load_to_single_tablet(false);
618
0
        }
619
0
    }
620
621
0
    if (ctx->timeout_second != -1) {
622
0
        request.__set_timeout(ctx->timeout_second);
623
0
    }
624
0
    request.__set_thrift_rpc_timeout_ms(config::thrift_rpc_timeout_ms);
625
0
    TMergeType::type merge_type = TMergeType::APPEND;
626
0
    StringCaseMap<TMergeType::type> merge_type_map = {{"APPEND", TMergeType::APPEND},
627
0
                                                      {"DELETE", TMergeType::DELETE},
628
0
                                                      {"MERGE", TMergeType::MERGE}};
629
0
    if (!http_req->header(HTTP_MERGE_TYPE).empty()) {
630
0
        std::string merge_type_str = http_req->header(HTTP_MERGE_TYPE);
631
0
        auto iter = merge_type_map.find(merge_type_str);
632
0
        if (iter != merge_type_map.end()) {
633
0
            merge_type = iter->second;
634
0
        } else {
635
0
            return Status::InvalidArgument("Invalid merge type {}", merge_type_str);
636
0
        }
637
0
        if (merge_type == TMergeType::MERGE && http_req->header(HTTP_DELETE_CONDITION).empty()) {
638
0
            return Status::InvalidArgument("Excepted DELETE ON clause when merge type is MERGE.");
639
0
        } else if (merge_type != TMergeType::MERGE &&
640
0
                   !http_req->header(HTTP_DELETE_CONDITION).empty()) {
641
0
            return Status::InvalidArgument(
642
0
                    "Not support DELETE ON clause when merge type is not MERGE.");
643
0
        }
644
0
    }
645
0
    request.__set_merge_type(merge_type);
646
0
    if (!http_req->header(HTTP_DELETE_CONDITION).empty()) {
647
0
        request.__set_delete_condition(http_req->header(HTTP_DELETE_CONDITION));
648
0
    }
649
650
0
    if (!http_req->header(HTTP_MAX_FILTER_RATIO).empty()) {
651
0
        ctx->max_filter_ratio = strtod(http_req->header(HTTP_MAX_FILTER_RATIO).c_str(), nullptr);
652
0
        request.__set_max_filter_ratio(ctx->max_filter_ratio);
653
0
    }
654
655
0
    if (!http_req->header(HTTP_HIDDEN_COLUMNS).empty()) {
656
0
        request.__set_hidden_columns(http_req->header(HTTP_HIDDEN_COLUMNS));
657
0
    }
658
0
    if (!http_req->header(HTTP_TRIM_DOUBLE_QUOTES).empty()) {
659
0
        if (iequal(http_req->header(HTTP_TRIM_DOUBLE_QUOTES), "true")) {
660
0
            request.__set_trim_double_quotes(true);
661
0
        } else {
662
0
            request.__set_trim_double_quotes(false);
663
0
        }
664
0
    }
665
0
    if (!http_req->header(HTTP_SKIP_LINES).empty()) {
666
0
        int skip_lines = DORIS_TRY(safe_stoi(http_req->header(HTTP_SKIP_LINES), HTTP_SKIP_LINES));
667
0
        if (skip_lines < 0) {
668
0
            return Status::InvalidArgument("Invalid 'skip_lines': {}", skip_lines);
669
0
        }
670
0
        request.__set_skip_lines(skip_lines);
671
0
    }
672
0
    if (!http_req->header(HTTP_ENABLE_PROFILE).empty()) {
673
0
        if (iequal(http_req->header(HTTP_ENABLE_PROFILE), "true")) {
674
0
            request.__set_enable_profile(true);
675
0
        } else {
676
0
            request.__set_enable_profile(false);
677
0
        }
678
0
    }
679
680
0
    if (!http_req->header(HTTP_UNIQUE_KEY_UPDATE_MODE).empty()) {
681
0
        static const StringCaseMap<TUniqueKeyUpdateMode::type> unique_key_update_mode_map = {
682
0
                {"UPSERT", TUniqueKeyUpdateMode::UPSERT},
683
0
                {"UPDATE_FIXED_COLUMNS", TUniqueKeyUpdateMode::UPDATE_FIXED_COLUMNS},
684
0
                {"UPDATE_FLEXIBLE_COLUMNS", TUniqueKeyUpdateMode::UPDATE_FLEXIBLE_COLUMNS}};
685
0
        std::string unique_key_update_mode_str = http_req->header(HTTP_UNIQUE_KEY_UPDATE_MODE);
686
0
        auto iter = unique_key_update_mode_map.find(unique_key_update_mode_str);
687
0
        if (iter != unique_key_update_mode_map.end()) {
688
0
            TUniqueKeyUpdateMode::type unique_key_update_mode = iter->second;
689
0
            if (unique_key_update_mode == TUniqueKeyUpdateMode::UPDATE_FLEXIBLE_COLUMNS) {
690
                // check constraints when flexible partial update is enabled
691
0
                if (ctx->format != TFileFormatType::FORMAT_JSON) {
692
0
                    return Status::InvalidArgument(
693
0
                            "flexible partial update only support json format as input file "
694
0
                            "currently");
695
0
                }
696
0
                if (!http_req->header(HTTP_FUZZY_PARSE).empty() &&
697
0
                    iequal(http_req->header(HTTP_FUZZY_PARSE), "true")) {
698
0
                    return Status::InvalidArgument(
699
0
                            "Don't support flexible partial update when 'fuzzy_parse' is enabled");
700
0
                }
701
0
                if (!http_req->header(HTTP_COLUMNS).empty()) {
702
0
                    return Status::InvalidArgument(
703
0
                            "Don't support flexible partial update when 'columns' is specified");
704
0
                }
705
0
                if (!http_req->header(HTTP_JSONPATHS).empty()) {
706
0
                    return Status::InvalidArgument(
707
0
                            "Don't support flexible partial update when 'jsonpaths' is specified");
708
0
                }
709
0
                if (!http_req->header(HTTP_HIDDEN_COLUMNS).empty()) {
710
0
                    return Status::InvalidArgument(
711
0
                            "Don't support flexible partial update when 'hidden_columns' is "
712
0
                            "specified");
713
0
                }
714
0
                if (!http_req->header(HTTP_FUNCTION_COLUMN + "." + HTTP_SEQUENCE_COL).empty()) {
715
0
                    return Status::InvalidArgument(
716
0
                            "Don't support flexible partial update when "
717
0
                            "'function_column.sequence_col' is specified");
718
0
                }
719
0
                if (!http_req->header(HTTP_MERGE_TYPE).empty()) {
720
0
                    return Status::InvalidArgument(
721
0
                            "Don't support flexible partial update when "
722
0
                            "'merge_type' is specified");
723
0
                }
724
0
                if (!http_req->header(HTTP_WHERE).empty()) {
725
0
                    return Status::InvalidArgument(
726
0
                            "Don't support flexible partial update when "
727
0
                            "'where' is specified");
728
0
                }
729
0
            }
730
0
            request.__set_unique_key_update_mode(unique_key_update_mode);
731
0
        } else {
732
0
            return Status::InvalidArgument(
733
0
                    "Invalid unique_key_partial_mode {}, must be one of 'UPSERT', "
734
0
                    "'UPDATE_FIXED_COLUMNS' or 'UPDATE_FLEXIBLE_COLUMNS'",
735
0
                    unique_key_update_mode_str);
736
0
        }
737
0
    }
738
739
0
    if (http_req->header(HTTP_UNIQUE_KEY_UPDATE_MODE).empty() &&
740
0
        !http_req->header(HTTP_PARTIAL_COLUMNS).empty()) {
741
        // only consider `partial_columns` parameter when `unique_key_update_mode` is not set
742
0
        if (iequal(http_req->header(HTTP_PARTIAL_COLUMNS), "true")) {
743
0
            request.__set_unique_key_update_mode(TUniqueKeyUpdateMode::UPDATE_FIXED_COLUMNS);
744
            // for backward compatibility
745
0
            request.__set_partial_update(true);
746
0
        }
747
0
    }
748
749
0
    if (!http_req->header(HTTP_PARTIAL_UPDATE_NEW_ROW_POLICY).empty()) {
750
0
        static const std::map<std::string, TPartialUpdateNewRowPolicy::type> policy_map {
751
0
                {"APPEND", TPartialUpdateNewRowPolicy::APPEND},
752
0
                {"ERROR", TPartialUpdateNewRowPolicy::ERROR}};
753
754
0
        auto policy_name = http_req->header(HTTP_PARTIAL_UPDATE_NEW_ROW_POLICY);
755
0
        std::transform(policy_name.begin(), policy_name.end(), policy_name.begin(),
756
0
                       [](unsigned char c) { return std::toupper(c); });
757
0
        auto it = policy_map.find(policy_name);
758
0
        if (it == policy_map.end()) {
759
0
            return Status::InvalidArgument(
760
0
                    "Invalid partial_update_new_key_behavior {}, must be one of {'APPEND', "
761
0
                    "'ERROR'}",
762
0
                    policy_name);
763
0
        }
764
0
        request.__set_partial_update_new_key_policy(it->second);
765
0
    }
766
767
0
    if (!http_req->header(HTTP_MEMTABLE_ON_SINKNODE).empty()) {
768
0
        bool value = iequal(http_req->header(HTTP_MEMTABLE_ON_SINKNODE), "true");
769
0
        request.__set_memtable_on_sink_node(value);
770
0
    }
771
0
    if (!http_req->header(HTTP_LOAD_STREAM_PER_NODE).empty()) {
772
0
        int stream_per_node = DORIS_TRY(
773
0
                safe_stoi(http_req->header(HTTP_LOAD_STREAM_PER_NODE), HTTP_LOAD_STREAM_PER_NODE));
774
0
        request.__set_stream_per_node(stream_per_node);
775
0
    }
776
0
    if (ctx->group_commit) {
777
0
        request.__set_group_commit_mode(ctx->group_commit_mode);
778
0
    }
779
780
0
    if (!http_req->header(HTTP_COMPUTE_GROUP).empty()) {
781
0
        request.__set_cloud_cluster(http_req->header(HTTP_COMPUTE_GROUP));
782
0
    } else if (!http_req->header(HTTP_CLOUD_CLUSTER).empty()) {
783
0
        request.__set_cloud_cluster(http_req->header(HTTP_CLOUD_CLUSTER));
784
0
    }
785
786
0
    if (!http_req->header(HTTP_EMPTY_FIELD_AS_NULL).empty()) {
787
0
        if (iequal(http_req->header(HTTP_EMPTY_FIELD_AS_NULL), "true")) {
788
0
            request.__set_empty_field_as_null(true);
789
0
        }
790
0
    }
791
792
#ifndef BE_TEST
793
    // plan this load
794
    TNetworkAddress master_addr = _exec_env->cluster_info()->master_fe_addr;
795
    int64_t stream_load_put_start_time = MonotonicNanos();
796
    RETURN_IF_ERROR(ThriftRpcHelper::rpc<FrontendServiceClient>(
797
            master_addr.hostname, master_addr.port,
798
            [&request, ctx](FrontendServiceConnection& client) {
799
                client->streamLoadPut(ctx->put_result, request);
800
            }));
801
    ctx->stream_load_put_cost_nanos = MonotonicNanos() - stream_load_put_start_time;
802
#else
803
0
    ctx->put_result = k_stream_load_put_result;
804
0
#endif
805
0
    Status plan_status(Status::create(ctx->put_result.status));
806
0
    if (!plan_status.ok()) {
807
0
        LOG(WARNING) << "plan streaming load failed. errmsg=" << plan_status << ctx->brief();
808
0
        return plan_status;
809
0
    }
810
0
    DCHECK(ctx->put_result.__isset.pipeline_params);
811
0
    ctx->put_result.pipeline_params.query_options.__set_enable_strict_cast(false);
812
0
    ctx->put_result.pipeline_params.query_options.__set_enable_insert_strict(strictMode);
813
0
    if (config::is_cloud_mode() && ctx->two_phase_commit && ctx->is_mow_table()) {
814
0
        return Status::NotSupported("stream load 2pc is unsupported for mow table");
815
0
    }
816
0
    if (iequal(ctx->group_commit_mode, ASYNC_MODE)) {
817
        // FIXME find a way to avoid chunked stream load write large WALs
818
0
        size_t content_length = 0;
819
0
        if (!http_req->header(HttpHeaders::CONTENT_LENGTH).empty()) {
820
0
            try {
821
0
                content_length = std::stol(http_req->header(HttpHeaders::CONTENT_LENGTH));
822
0
            } catch (const std::exception& e) {
823
0
                return Status::InvalidArgument("invalid HTTP header CONTENT_LENGTH={}: {}",
824
0
                                               http_req->header(HttpHeaders::CONTENT_LENGTH),
825
0
                                               e.what());
826
0
            }
827
0
            if (ctx->format == TFileFormatType::FORMAT_CSV_GZ ||
828
0
                ctx->format == TFileFormatType::FORMAT_CSV_LZO ||
829
0
                ctx->format == TFileFormatType::FORMAT_CSV_BZ2 ||
830
0
                ctx->format == TFileFormatType::FORMAT_CSV_LZ4FRAME ||
831
0
                ctx->format == TFileFormatType::FORMAT_CSV_LZOP ||
832
0
                ctx->format == TFileFormatType::FORMAT_CSV_LZ4BLOCK ||
833
0
                ctx->format == TFileFormatType::FORMAT_CSV_SNAPPYBLOCK) {
834
0
                content_length *= 3;
835
0
            }
836
0
        }
837
0
        ctx->put_result.pipeline_params.__set_content_length(content_length);
838
0
    }
839
840
0
    VLOG_NOTICE << "params is "
841
0
                << apache::thrift::ThriftDebugString(ctx->put_result.pipeline_params);
842
    // if we not use streaming, we must download total content before we begin
843
    // to process this load
844
0
    if (!ctx->use_streaming) {
845
0
        return Status::OK();
846
0
    }
847
848
0
    TPipelineFragmentParamsList mocked;
849
0
    return _exec_env->stream_load_executor()->execute_plan_fragment(
850
0
            ctx, mocked, [http_req, this](std::shared_ptr<StreamLoadContext> ctx) {
851
0
                _on_finish(ctx, http_req);
852
0
            });
853
0
}
854
855
Status StreamLoadAction::_data_saved_path(HttpRequest* req, std::string* file_path,
856
0
                                          int64_t file_bytes) {
857
0
    std::string prefix;
858
0
    RETURN_IF_ERROR(_exec_env->load_path_mgr()->allocate_dir(req->param(HTTP_DB_KEY), "", &prefix,
859
0
                                                             file_bytes));
860
0
    timeval tv;
861
0
    gettimeofday(&tv, nullptr);
862
0
    struct tm tm;
863
0
    time_t cur_sec = tv.tv_sec;
864
0
    localtime_r(&cur_sec, &tm);
865
0
    char buf[64];
866
0
    strftime(buf, 64, "%Y%m%d%H%M%S", &tm);
867
0
    std::stringstream ss;
868
0
    ss << prefix << "/" << req->param(HTTP_TABLE_KEY) << "." << buf << "." << tv.tv_usec;
869
0
    *file_path = ss.str();
870
0
    return Status::OK();
871
0
}
872
873
void StreamLoadAction::_save_stream_load_record(std::shared_ptr<StreamLoadContext> ctx,
874
0
                                                const std::string& str) {
875
0
    std::shared_ptr<StreamLoadRecorder> stream_load_recorder =
876
0
            ExecEnv::GetInstance()->storage_engine().get_stream_load_recorder();
877
878
0
    if (stream_load_recorder != nullptr) {
879
0
        std::string key =
880
0
                std::to_string(ctx->start_millis + ctx->load_cost_millis) + "_" + ctx->label;
881
0
        auto st = stream_load_recorder->put(key, str);
882
0
        if (st.ok()) {
883
0
            LOG(INFO) << "put stream_load_record rocksdb successfully. label: " << ctx->label
884
0
                      << ", key: " << key;
885
0
        }
886
0
    } else {
887
0
        LOG(WARNING) << "put stream_load_record rocksdb failed. stream_load_recorder is null.";
888
0
    }
889
0
}
890
891
Status StreamLoadAction::_check_wal_space(const std::string& group_commit_mode,
892
0
                                          int64_t content_length) {
893
0
    if (iequal(group_commit_mode, ASYNC_MODE) &&
894
0
        !load_size_smaller_than_wal_limit(content_length)) {
895
0
        std::stringstream ss;
896
0
        ss << "There is no space for group commit stream load async WAL. This stream load "
897
0
              "size is "
898
0
           << content_length
899
0
           << ". WAL dir info: " << ExecEnv::GetInstance()->wal_mgr()->get_wal_dirs_info_string();
900
0
        LOG(WARNING) << ss.str();
901
0
        return Status::Error<EXCEEDED_LIMIT>(ss.str());
902
0
    }
903
0
    return Status::OK();
904
0
}
905
906
Status StreamLoadAction::_can_group_commit(HttpRequest* req, std::shared_ptr<StreamLoadContext> ctx,
907
                                           std::string& group_commit_header,
908
0
                                           bool& can_group_commit) {
909
0
    int64_t content_length = req->header(HttpHeaders::CONTENT_LENGTH).empty()
910
0
                                     ? 0
911
0
                                     : std::stoll(req->header(HttpHeaders::CONTENT_LENGTH));
912
0
    if (content_length < 0) {
913
0
        std::stringstream ss;
914
0
        ss << "This stream load content length <0 (" << content_length
915
0
           << "), please check your content length.";
916
0
        LOG(WARNING) << ss.str();
917
0
        return Status::InvalidArgument(ss.str());
918
0
    }
919
0
    auto is_chunk = !req->header(HttpHeaders::TRANSFER_ENCODING).empty() &&
920
0
                    req->header(HttpHeaders::TRANSFER_ENCODING).find(CHUNK) != std::string::npos;
921
0
    if (content_length == 0 && !is_chunk) {
922
        // off_mode and empty
923
0
        can_group_commit = false;
924
0
        return Status::OK();
925
0
    }
926
0
    if (is_chunk) {
927
0
        ctx->label = "";
928
0
    }
929
930
0
    auto partial_columns = !req->header(HTTP_PARTIAL_COLUMNS).empty() &&
931
0
                           iequal(req->header(HTTP_PARTIAL_COLUMNS), "true");
932
0
    auto temp_partitions = !req->header(HTTP_TEMP_PARTITIONS).empty();
933
0
    auto partitions = !req->header(HTTP_PARTITIONS).empty();
934
0
    auto update_mode =
935
0
            !req->header(HTTP_UNIQUE_KEY_UPDATE_MODE).empty() &&
936
0
            (iequal(req->header(HTTP_UNIQUE_KEY_UPDATE_MODE), "UPDATE_FIXED_COLUMNS") ||
937
0
             iequal(req->header(HTTP_UNIQUE_KEY_UPDATE_MODE), "UPDATE_FLEXIBLE_COLUMNS"));
938
0
    if (!partial_columns && !partitions && !temp_partitions && !ctx->two_phase_commit &&
939
0
        !update_mode) {
940
0
        if (!config::wait_internal_group_commit_finish && !group_commit_header.empty() &&
941
0
            !ctx->label.empty()) {
942
0
            return Status::InvalidArgument("label and group_commit can't be set at the same time");
943
0
        }
944
0
        RETURN_IF_ERROR(_check_wal_space(group_commit_header, content_length));
945
0
        can_group_commit = true;
946
0
    }
947
0
    return Status::OK();
948
0
}
949
950
Status StreamLoadAction::_handle_group_commit(HttpRequest* req,
951
0
                                              std::shared_ptr<StreamLoadContext> ctx) {
952
0
    std::string group_commit_header = req->header(HTTP_GROUP_COMMIT);
953
0
    if (!group_commit_header.empty() && !iequal(group_commit_header, SYNC_MODE) &&
954
0
        !iequal(group_commit_header, ASYNC_MODE) && !iequal(group_commit_header, OFF_MODE)) {
955
0
        return Status::InvalidArgument(
956
0
                "group_commit can only be [async_mode, sync_mode, off_mode]");
957
0
    }
958
0
    if (config::wait_internal_group_commit_finish) {
959
0
        group_commit_header = SYNC_MODE;
960
0
    }
961
962
    // if group_commit_header is off_mode, we will not use group commit
963
0
    if (iequal(group_commit_header, OFF_MODE)) {
964
0
        ctx->group_commit_mode = OFF_MODE;
965
0
        ctx->group_commit = false;
966
0
        return Status::OK();
967
0
    }
968
0
    bool can_group_commit = false;
969
0
    RETURN_IF_ERROR(_can_group_commit(req, ctx, group_commit_header, can_group_commit));
970
0
    if (!can_group_commit) {
971
0
        ctx->group_commit_mode = OFF_MODE;
972
0
        ctx->group_commit = false;
973
0
    } else {
974
0
        if (!group_commit_header.empty()) {
975
0
            ctx->group_commit_mode = group_commit_header;
976
0
            ctx->group_commit = true;
977
0
        } else {
978
            // use table property to decide group commit or not
979
0
            ctx->group_commit_mode = "";
980
0
            ctx->group_commit = false;
981
0
        }
982
0
    }
983
0
    return Status::OK();
984
0
}
985
986
} // namespace doris