1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
|
/**
* Copyright (C) 2020-present MongoDB, Inc.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the Server Side Public License, version 1,
* as published by MongoDB, Inc.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* Server Side Public License for more details.
*
* You should have received a copy of the Server Side Public License
* along with this program. If not, see
* <http://www.mongodb.com/licensing/server-side-public-license>.
*
* As a special exception, the copyright holders give permission to link the
* code of portions of this program with the OpenSSL library under certain
* conditions as described in each individual source file and distribute
* linked combinations including the program with the OpenSSL library. You
* must comply with the Server Side Public License in all respects for
* all of the code used other than as permitted herein. If you modify file(s)
* with this exception, you may extend this exception to your version of the
* file(s), but you are not obligated to do so. If you do not wish to do so,
* delete this exception statement from your version. If you delete this
* exception statement from all source files in the program, then also delete
* it in the license file.
*/
#include "mongo/platform/basic.h"
#include "mongo/db/query/sbe_runtime_planner.h"
#include "mongo/db/catalog/collection.h"
#include "mongo/db/exec/sbe/expressions/expression.h"
#include "mongo/db/exec/trial_period_utils.h"
#include "mongo/db/exec/trial_run_tracker.h"
#include "mongo/db/query/plan_executor_sbe.h"
namespace mongo::sbe {
namespace {
/**
* Fetches a next document form the given plan stage tree and returns 'true' if the plan stage
* returns EOF, or throws 'TrialRunTracker::EarlyExitException' exception. Otherwise, the
* loaded document is placed into the candidate's plan result queue.
*
* If the plan stage throws a 'QueryExceededMemoryLimitNoDiskUseAllowed', it will be caught and the
* 'candidate->failed' flag will be set to 'true', and the 'numFailures' parameter incremented by 1.
* This failure is considered recoverable, as another candidate plan may require less memory, or may
* not contain a stage requiring spilling to disk at all.
*/
bool fetchNextDocument(plan_ranker::CandidatePlan* candidate,
const std::pair<value::SlotAccessor*, value::SlotAccessor*>& slots) {
try {
BSONObj obj;
RecordId recordId;
auto [resultSlot, recordIdSlot] = slots;
auto state = fetchNext(candidate->root.get(),
resultSlot,
recordIdSlot,
&obj,
recordIdSlot ? &recordId : nullptr,
true /* must return owned BSON */);
if (state == PlanState::IS_EOF) {
candidate->root->close();
return true;
}
invariant(state == PlanState::ADVANCED);
invariant(obj.isOwned());
candidate->results.push({obj, {recordIdSlot != nullptr, recordId}});
} catch (const ExceptionFor<ErrorCodes::QueryTrialRunCompleted>&) {
candidate->exitedEarly = true;
return true;
} catch (const ExceptionFor<ErrorCodes::QueryExceededMemoryLimitNoDiskUseAllowed>& ex) {
candidate->root->close();
candidate->status = ex.toStatus();
}
return false;
}
} // namespace
StatusWith<std::tuple<value::SlotAccessor*, value::SlotAccessor*, bool>>
BaseRuntimePlanner::prepareExecutionPlan(PlanStage* root,
stage_builder::PlanStageData* data) const {
invariant(root);
invariant(data);
root->prepare(data->ctx);
value::SlotAccessor* resultSlot{nullptr};
if (auto slot = data->outputs.getIfExists(stage_builder::PlanStageSlots::kResult); slot) {
resultSlot = root->getAccessor(data->ctx, *slot);
tassert(4822871, "Query does not have a result slot.", resultSlot);
}
value::SlotAccessor* recordIdSlot{nullptr};
if (auto slot = data->outputs.getIfExists(stage_builder::PlanStageSlots::kRecordId); slot) {
recordIdSlot = root->getAccessor(data->ctx, *slot);
tassert(4822872, "Query does not have a recordId slot.", recordIdSlot);
}
auto exitedEarly{false};
try {
root->open(false);
} catch (const ExceptionFor<ErrorCodes::QueryTrialRunCompleted>&) {
exitedEarly = true;
} catch (const ExceptionFor<ErrorCodes::QueryExceededMemoryLimitNoDiskUseAllowed>& ex) {
root->close();
return ex.toStatus();
}
return std::make_tuple(resultSlot, recordIdSlot, exitedEarly);
}
bool BaseRuntimePlanner::executeCandidateTrial(plan_ranker::CandidatePlan* candidate,
size_t maxNumResults) {
_indexExistenceChecker.check();
auto status = prepareExecutionPlan(candidate->root.get(), &candidate->data);
if (!status.isOK()) {
candidate->status = status.getStatus();
return candidate;
}
auto [resultAccessor, recordIdAccessor, exitedEarly] = status.getValue();
candidate->exitedEarly = exitedEarly;
for (size_t it = 0; it < maxNumResults && candidate->status.isOK() && !candidate->exitedEarly &&
!candidate->needsReplanning;
++it) {
if (fetchNextDocument(candidate, std::make_pair(resultAccessor, recordIdAccessor))) {
return true;
}
}
return false;
}
std::vector<plan_ranker::CandidatePlan> BaseRuntimePlanner::collectExecutionStats(
std::vector<std::unique_ptr<QuerySolution>> solutions,
std::vector<std::pair<std::unique_ptr<PlanStage>, stage_builder::PlanStageData>> roots,
size_t maxTrialPeriodNumReads) {
invariant(solutions.size() == roots.size());
std::vector<plan_ranker::CandidatePlan> candidates;
std::vector<std::pair<value::SlotAccessor*, value::SlotAccessor*>> accessors;
const auto maxNumResults{trial_period::getTrialPeriodNumToReturn(_cq)};
// Determine which plans are blocking and which are non blocking. The non blocking plans will
// be run first in order to provide an upper bound on the number of reads allowed for the
// blocking plans.
std::vector<size_t> nonBlockingPlanIndexes;
std::vector<size_t> blockingPlanIndexes;
for (size_t index = 0; index < solutions.size(); ++index) {
if (solutions[index]->hasBlockingStage) {
blockingPlanIndexes.push_back(index);
} else {
nonBlockingPlanIndexes.push_back(index);
}
}
// If all the plans are blocking, then the trial period risks going on for too long. Because the
// plans are blocking, they may not provide 'maxNumResults' within the allotted budget of reads.
// We could end up in a situation where each plan's trial period runs for a long time,
// substantially slowing down the multi-planning process. For this reason, when all the plans
// are blocking, we pass 'maxNumResults' to the trial run tracker. This causes the sort stage to
// exit early as soon as it sees 'maxNumResults' _input_ values, which keeps the trial period
// shorter.
//
// On the other hand, if we have a mix of blocking and non-blocking plans, we don't want the
// sort stage to exit early based on the number of input rows it observes. This could cause the
// trial period for the blocking plans to run for a much shorter timeframe than the non-blocking
// plans. This leads to an apples-to-oranges comparison between the blocking and non-blocking
// plans which could artificially favor the blocking plans.
const size_t trackerResultsBudget = nonBlockingPlanIndexes.empty() ? maxNumResults : 0;
auto runPlans = [&](const std::vector<size_t>& planIndexes, size_t& maxNumReads) -> void {
for (auto planIndex : planIndexes) {
// Prepare the plan.
auto&& [root, data] = roots[planIndex];
// Attach a unique TrialRunTracker to the plan, which is configured to use at most
// 'maxNumReads' reads.
auto tracker = std::make_unique<TrialRunTracker>(trackerResultsBudget, maxNumReads);
ON_BLOCK_EXIT([rootPtr = root.get()] { rootPtr->detachFromTrialRunTracker(); });
root->attachToTrialRunTracker(tracker.get());
candidates.push_back({std::move(solutions[planIndex]),
std::move(root),
std::move(data),
false /* exitedEarly */,
false /* needsReplanning */,
Status::OK()});
auto& currentCandidate = candidates.back();
bool candidateDone = executeCandidateTrial(¤tCandidate, maxNumResults);
if (candidateDone ||
(currentCandidate.status.isOK() && !currentCandidate.exitedEarly)) {
maxNumReads = std::min(
maxNumReads, tracker->getMetric<TrialRunTracker::TrialRunMetric::kNumReads>());
}
}
};
runPlans(nonBlockingPlanIndexes, maxTrialPeriodNumReads);
runPlans(blockingPlanIndexes, maxTrialPeriodNumReads);
return candidates;
}
} // namespace mongo::sbe
|
