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
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
|
/**
* Copyright (C) 2018-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.
*/
#define MONGO_LOGV2_DEFAULT_COMPONENT ::mongo::logv2::LogComponent::kSharding
#include "mongo/platform/basic.h"
#include "mongo/db/s/split_vector.h"
#include "mongo/base/status_with.h"
#include "mongo/db/bson/dotted_path_support.h"
#include "mongo/db/catalog/index_catalog.h"
#include "mongo/db/catalog_raii.h"
#include "mongo/db/dbhelpers.h"
#include "mongo/db/exec/working_set_common.h"
#include "mongo/db/index/index_descriptor.h"
#include "mongo/db/keypattern.h"
#include "mongo/db/namespace_string.h"
#include "mongo/db/query/internal_plans.h"
#include "mongo/db/query/plan_executor.h"
#include "mongo/logv2/log.h"
namespace mongo {
namespace {
const int kMaxObjectPerChunk{250000};
const int estimatedAdditionalBytesPerItemInBSONArray{2};
BSONObj prettyKey(const BSONObj& keyPattern, const BSONObj& key) {
return key.replaceFieldNames(keyPattern).clientReadable();
}
} // namespace
std::vector<BSONObj> splitVector(OperationContext* opCtx,
const NamespaceString& nss,
const BSONObj& keyPattern,
const BSONObj& min,
const BSONObj& max,
bool force,
boost::optional<long long> maxSplitPoints,
boost::optional<long long> maxChunkObjects,
boost::optional<long long> maxChunkSizeBytes) {
std::vector<BSONObj> splitKeys;
std::size_t splitVectorResponseSize = 0;
// Always have a default value for maxChunkObjects
if (!maxChunkObjects) {
maxChunkObjects = kMaxObjectPerChunk;
}
{
AutoGetCollection autoColl(opCtx, nss, MODE_IS);
Collection* const collection = autoColl.getCollection();
uassert(ErrorCodes::NamespaceNotFound, "ns not found", collection);
// Allow multiKey based on the invariant that shard keys must be single-valued. Therefore,
// any multi-key index prefixed by shard key cannot be multikey over the shard key fields.
const IndexDescriptor* idx =
collection->getIndexCatalog()->findShardKeyPrefixedIndex(opCtx, keyPattern, false);
uassert(ErrorCodes::IndexNotFound,
str::stream() << "couldn't find index over splitting key "
<< keyPattern.clientReadable().toString(),
idx);
// extend min to get (min, MinKey, MinKey, ....)
KeyPattern kp(idx->keyPattern());
BSONObj minKey = Helpers::toKeyFormat(kp.extendRangeBound(min, false));
BSONObj maxKey;
if (max.isEmpty()) {
// if max not specified, make it (MaxKey, Maxkey, MaxKey...)
maxKey = Helpers::toKeyFormat(kp.extendRangeBound(max, true));
} else {
// otherwise make it (max,MinKey,MinKey...) so that bound is non-inclusive
maxKey = Helpers::toKeyFormat(kp.extendRangeBound(max, false));
}
// Get the size estimate for this namespace
const long long recCount = collection->numRecords(opCtx);
const long long dataSize = collection->dataSize(opCtx);
// Now that we have the size estimate, go over the remaining parameters and apply any
// maximum size restrictions specified there.
// Forcing a split is equivalent to having maxChunkSizeBytes be the size of the current
// chunk, i.e., the logic below will split that chunk in half
if (force) {
maxChunkSizeBytes = dataSize;
}
// We need a maximum size for the chunk.
if (!maxChunkSizeBytes || maxChunkSizeBytes.get() <= 0) {
uasserted(ErrorCodes::InvalidOptions, "need to specify the desired max chunk size");
}
// If there's not enough data for more than one chunk, no point continuing.
if (dataSize < maxChunkSizeBytes.get() || recCount == 0) {
std::vector<BSONObj> emptyVector;
return emptyVector;
}
LOGV2(22107,
"Requested split points lookup for chunk {namespace} {minKey} -->> {maxKey}",
"Requested split points lookup for chunk",
"namespace"_attr = nss.toString(),
"minKey"_attr = redact(minKey),
"maxKey"_attr = redact(maxKey));
// We'll use the average object size and number of object to find approximately how many
// keys each chunk should have. We'll split at half the maxChunkSizeBytes or
// maxChunkObjects, if provided.
const long long avgRecSize = dataSize / recCount;
long long keyCount = maxChunkSizeBytes.get() / (2 * avgRecSize);
if (maxChunkObjects.get() && (maxChunkObjects.get() < keyCount)) {
LOGV2(22108,
"Limiting the number of documents per chunk to {maxChunkObjects} based "
"on the maxChunkObjects parameter for split vector command (compared to maximum "
"possible: {maxPossibleDocumentsPerChunk})",
"Limiting the number of documents per chunk for split vector command based on "
"the maxChunksObject parameter",
"maxChunkObjects"_attr = maxChunkObjects.get(),
"maxPossibleDocumentsPerChunk"_attr = keyCount);
keyCount = maxChunkObjects.get();
}
//
// Traverse the index and add the keyCount-th key to the result vector. If that key
// appeared in the vector before, we omit it. The invariant here is that all the
// instances of a given key value live in the same chunk.
//
Timer timer;
long long currCount = 0;
long long numChunks = 0;
auto exec = InternalPlanner::indexScan(opCtx,
collection,
idx,
minKey,
maxKey,
BoundInclusion::kIncludeStartKeyOnly,
PlanExecutor::YIELD_AUTO,
InternalPlanner::FORWARD);
BSONObj currKey;
PlanExecutor::ExecState state = exec->getNext(&currKey, nullptr);
uassert(ErrorCodes::OperationFailed,
"can't open a cursor to scan the range (desired range is possibly empty)",
state == PlanExecutor::ADVANCED);
// Get the final key in the range, and see if it's the same as the first key.
BSONObj maxKeyInChunk;
{
auto exec = InternalPlanner::indexScan(opCtx,
collection,
idx,
maxKey,
minKey,
BoundInclusion::kIncludeEndKeyOnly,
PlanExecutor::YIELD_AUTO,
InternalPlanner::BACKWARD);
PlanExecutor::ExecState state = exec->getNext(&maxKeyInChunk, nullptr);
uassert(
ErrorCodes::OperationFailed,
"can't open a cursor to find final key in range (desired range is possibly empty)",
state == PlanExecutor::ADVANCED);
}
if (currKey.woCompare(maxKeyInChunk) == 0) {
// Range contains only documents with a single key value. So we cannot possibly find a
// split point, and there is no need to scan any further.
LOGV2_WARNING(
22113,
"Possible low cardinality key detected in {namespace} - range {minKey} -->> "
"{maxKey} contains only the key {key}",
"Possible low cardinality key detected in range. Range contains only a single key.",
"namespace"_attr = nss.toString(),
"minKey"_attr = redact(minKey),
"maxKey"_attr = redact(maxKey),
"key"_attr = redact(prettyKey(idx->keyPattern(), currKey)));
std::vector<BSONObj> emptyVector;
return emptyVector;
}
// Use every 'keyCount'-th key as a split point. We add the initial key as a sentinel,
// to be removed at the end. If a key appears more times than entries allowed on a
// chunk, we issue a warning and split on the following key.
auto tooFrequentKeys = SimpleBSONObjComparator::kInstance.makeBSONObjSet();
splitKeys.push_back(dotted_path_support::extractElementsBasedOnTemplate(
prettyKey(idx->keyPattern(), currKey.getOwned()), keyPattern));
while (1) {
while (PlanExecutor::ADVANCED == state) {
currCount++;
if (currCount > keyCount && !force) {
currKey = dotted_path_support::extractElementsBasedOnTemplate(
prettyKey(idx->keyPattern(), currKey.getOwned()), keyPattern);
// Do not use this split key if it is the same used in the previous split
// point.
if (currKey.woCompare(splitKeys.back()) == 0) {
tooFrequentKeys.insert(currKey.getOwned());
} else {
auto additionalKeySize =
currKey.objsize() + estimatedAdditionalBytesPerItemInBSONArray;
if (splitVectorResponseSize + additionalKeySize > BSONObjMaxUserSize) {
if (splitKeys.empty()) {
// Keep trying until we get at least one split point that isn't
// above the max object user size.
state = exec->getNext(&currKey, nullptr);
continue;
}
LOGV2(22109,
"Max BSON response size reached for split vector before the end "
"of chunk {namespace} {minKey} -->> {maxKey}",
"Max BSON response size reached for split vector before the end "
"of chunk",
"namespace"_attr = nss.toString(),
"minKey"_attr = redact(minKey),
"maxKey"_attr = redact(maxKey));
break;
}
splitVectorResponseSize += additionalKeySize;
splitKeys.push_back(currKey.getOwned());
currCount = 0;
numChunks++;
LOGV2_DEBUG(22110,
4,
"Picked a split key: {key}",
"Picked a split key",
"key"_attr = redact(currKey));
}
}
// Stop if we have enough split points.
if (maxSplitPoints && maxSplitPoints.get() && (numChunks >= maxSplitPoints.get())) {
LOGV2(22111,
"Max number of requested split points reached ({numSplitPoints}) before "
"the end of chunk {namespace} {minKey} -->> {maxKey}",
"Max number of requested split points reached before the end of chunk",
"numSplitPoints"_attr = numChunks,
"namespace"_attr = nss.toString(),
"minKey"_attr = redact(minKey),
"maxKey"_attr = redact(maxKey));
break;
}
state = exec->getNext(&currKey, nullptr);
}
if (!force)
break;
//
// If we're forcing a split at the halfway point, then the first pass was just
// to count the keys, and we still need a second pass.
//
force = false;
keyCount = currCount / 2;
currCount = 0;
LOGV2(22112,
"splitVector doing another cycle because of force, keyCount now: {keyCount}",
"splitVector doing another cycle because of force",
"keyCount"_attr = keyCount);
exec = InternalPlanner::indexScan(opCtx,
collection,
idx,
minKey,
maxKey,
BoundInclusion::kIncludeStartKeyOnly,
PlanExecutor::YIELD_AUTO,
InternalPlanner::FORWARD);
state = exec->getNext(&currKey, nullptr);
}
//
// Format the result and issue any warnings about the data we gathered while traversing the
// index
//
// Warn for keys that are more numerous than maxChunkSizeBytes allows.
for (auto it = tooFrequentKeys.cbegin(); it != tooFrequentKeys.cend(); ++it) {
LOGV2_WARNING(22114,
"Possible low cardinality key detected in {namespace} - key is "
"{key}",
"Possible low cardinality key detected",
"namespace"_attr = nss.toString(),
"key"_attr = redact(prettyKey(idx->keyPattern(), *it)));
}
// Remove the sentinel at the beginning before returning
splitKeys.erase(splitKeys.begin());
if (timer.millis() > serverGlobalParams.slowMS) {
LOGV2_WARNING(
22115,
"Finding the split vector for {namespace} over {keyPattern} keyCount: {keyCount} "
"numSplits: {numSplits} lookedAt: {currCount} took {duration}",
"Finding the split vector completed",
"namespace"_attr = nss.toString(),
"keyPattern"_attr = redact(keyPattern),
"keyCount"_attr = keyCount,
"numSplits"_attr = splitKeys.size(),
"currCount"_attr = currCount,
"duration"_attr = Milliseconds(timer.millis()));
}
}
// Make sure splitKeys is in ascending order
std::sort(
splitKeys.begin(), splitKeys.end(), SimpleBSONObjComparator::kInstance.makeLessThan());
return splitKeys;
}
} // namespace mongo
|