summaryrefslogtreecommitdiff
path: root/src/mongo/db/s/metadata_manager.h
blob: 2844d696874d751c7226eb19a549518580a760b9 (plain)
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
/**
 *    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.
 */

#pragma once

#include <list>
#include <memory>

#include "mongo/bson/simple_bsonobj_comparator.h"
#include "mongo/db/logical_time.h"
#include "mongo/db/namespace_string.h"
#include "mongo/db/range_arithmetic.h"
#include "mongo/db/s/scoped_collection_metadata.h"
#include "mongo/db/service_context.h"
#include "mongo/executor/task_executor.h"
#include "mongo/s/catalog/type_chunk.h"
#include "mongo/util/concurrency/with_lock.h"

namespace mongo {

class RangePreserver;

/**
 * Contains filtering metadata for a sharded collection.
 */
class MetadataManager : public std::enable_shared_from_this<MetadataManager> {
public:
    MetadataManager(ServiceContext* serviceContext,
                    NamespaceString nss,
                    std::shared_ptr<executor::TaskExecutor> executor,
                    CollectionMetadata initialMetadata);
    ~MetadataManager() = default;

    MetadataManager(const MetadataManager&) = delete;
    MetadataManager& operator=(const MetadataManager&) = delete;

    /**
     * Increments the usage counter of the active metadata and returns an RAII object, which
     * corresponds to it.
     *
     * Holding a reference on a particular instance of the metadata means that orphan cleanup is not
     * allowed to run and delete chunks which are covered by that metadata. When the returned
     * ScopedCollectionDescription goes out of scope, the reference counter on the metadata will be
     * decremented and if it reaches to zero, orphan cleanup may proceed.
     */
    std::shared_ptr<ScopedCollectionDescription::Impl> getActiveMetadata(
        const boost::optional<LogicalTime>& atClusterTime);

    /**
     * Returns the shard version of the active metadata object.
     */
    ChunkVersion getActiveShardVersion() {
        stdx::lock_guard<Latch> lg(_managerLock);
        invariant(!_metadata.empty());
        return _metadata.back()->metadata->getShardVersion();
    }

    /**
     * Returns the UUID of the collection tracked by this MetadataManager object.
     */
    UUID getCollectionUuid() const {
        return _collectionUuid;
    }

    /**
     * Returns the number of CollectionMetadata objects being maintained on behalf of running
     * queries.  The actual number may vary after it returns, so this is really only useful for unit
     * tests.
     */
    size_t numberOfMetadataSnapshots() const;

    /**
     * Returns the number of metadata objects that have been set to boost::none in
     * _retireExpiredMetadata(). The actual number may vary after it returns, so this is really only
     * useful for unit tests.
     */
    int numberOfEmptyMetadataSnapshots() const;

    void setFilteringMetadata(CollectionMetadata newMetadata);

    void toBSONPending(BSONArrayBuilder& bb) const;

    /**
     * Returns the number of items in the _receivingChunks list. Useful for unit tests.
     */
    size_t numberOfReceivingChunks() {
        return _receivingChunks.size();
    }

    /**
     * Clears the items in the _receivingChunks list.
     */
    void clearReceivingChunks();

    /**
     * Appends information on all the chunk ranges in rangesToClean to builder.
     */
    void append(BSONObjBuilder* builder) const;

    /**
     * Schedules any documents in `range` for immediate cleanup iff no running queries can depend
     * on them, and adds the range to the list of ranges currently being received.
     *
     * Returns a future that will be resolved when the deletion either completes or fail.
     */
    SharedSemiFuture<void> beginReceive(ChunkRange const& range);

    /**
     * Removes `range` from the list of ranges currently being received, and schedules any documents
     * in the range for immediate cleanup.
     */
    void forgetReceive(const ChunkRange& range);

    /**
     * Schedules documents in `range` for cleanup after any running queries that may depend on them
     * have terminated. Does not block. Fails if the range overlaps any current local shard chunk.
     *
     * If shouldDelayBeforeDeletion is false, deletion is scheduled immediately after the last
     * dependent query completes; otherwise, deletion is postponed until after
     * orphanCleanupDelaySecs after the last dependent query completes.
     *
     * Returns a future that will be fulfilled when the range deletion completes or fails.
     */
    SharedSemiFuture<void> cleanUpRange(ChunkRange const& range,
                                        boost::optional<UUID> migrationId,
                                        bool shouldDelayBeforeDeletion);

    /**
     * Returns the number of ranges scheduled to be cleaned, exclusive of such ranges that might
     * still be in use by running queries.  Outside of test drivers, the actual number may vary
     * after it returns, so this is really only useful for unit tests.
     */
    size_t numberOfRangesToClean() const;

    /**
     * Returns the number of ranges scheduled to be cleaned once all queries that could depend on
     * them have terminated. The actual number may vary after it returns, so this is really only
     * useful for unit tests.
     */
    size_t numberOfRangesToCleanStillInUse() const;

    /**
     * Returns the number of ranges scheduled for deletion, regardless of whether they may still be
     * in use by running queries.
     */
    size_t numberOfRangesScheduledForDeletion() const;

    /**
     * Reports whether any range still scheduled for deletion overlaps the argument range. If so,
     * returns a future that will be resolved when the newest overlapping range's deletion (possibly
     * the one of interest) completes or fails.
     */
    boost::optional<SharedSemiFuture<void>> trackOrphanedDataCleanup(
        ChunkRange const& orphans) const;

    boost::optional<ChunkRange> getNextOrphanRange(BSONObj const& from) const;

private:
    // Management of the _metadata list is implemented in RangePreserver
    friend class RangePreserver;

    /**
     * Represents an instance of what the filtering metadata for this collection was at a particular
     * point in time along with a counter of how many queries are still using it.
     */
    struct CollectionMetadataTracker {
        CollectionMetadataTracker(const CollectionMetadataTracker&) = delete;
        CollectionMetadataTracker& operator=(const CollectionMetadataTracker&) = delete;

        CollectionMetadataTracker(CollectionMetadata inMetadata)
            : metadata(std::move(inMetadata)) {}

        ~CollectionMetadataTracker() {
            invariant(!usageCounter);
            onDestructionPromise.emplaceValue();
        }

        boost::optional<CollectionMetadata> metadata;

        /**
         * Number of range deletion tasks waiting on this CollectionMetadataTracker to be destroyed
         * before deleting documents.
         */
        uint32_t numContingentRangeDeletionTasks{0};

        /**
         * Promise that will be signaled when this object is destroyed.
         *
         * In the case where this CollectionMetadataTracker may refer to orphaned documents for one
         * or more ranges, the corresponding futures from this promise are used as barriers to
         * prevent range deletion tasks for those ranges from proceeding until this object is
         * destroyed, to guarantee that ranges aren't deleted while active queries can still access
         * them.
         */
        SharedPromise<void> onDestructionPromise;

        uint32_t usageCounter{0};
    };

    /**
     * Retires any metadata that has fallen out of use, potentially allowing range deletions to
     * proceed which were waiting for active queries using these metadata objects to complete.
     */
    void _retireExpiredMetadata(WithLock);

    /**
     * Pushes current set of chunks, if any, to _metadataInUse, replaces it with newMetadata.
     */
    void _setActiveMetadata(WithLock wl, CollectionMetadata newMetadata);

    /**
     * Finds the most-recently pushed metadata that might depend on `range`, or nullptr if none.
     * The result is usable until the lock is released.
     */
    CollectionMetadataTracker* _findNewestOverlappingMetadata(WithLock, ChunkRange const& range);

    /**
     * Returns true if the specified range overlaps any chunk that might be currently in use by a
     * running query.
     */

    bool _overlapsInUseChunk(WithLock, ChunkRange const& range);

    /**
     * Schedule a task to delete the given range of documents once waitForActiveQueriesToComplete
     * has been signaled, and store the resulting future for the task in
     * _rangesScheduledForDeletion.
     */
    SharedSemiFuture<void> _submitRangeForDeletion(
        const WithLock&,
        SemiFuture<void> waitForActiveQueriesToComplete,
        const ChunkRange& range,
        boost::optional<UUID> migrationId,
        Seconds delayForActiveQueriesOnSecondariesToComplete);

    // ServiceContext from which to obtain instances of global support objects
    ServiceContext* const _serviceContext;

    // Namespace for which this manager object applies
    const NamespaceString _nss;

    // The UUID for the collection tracked by this manager object.
    const UUID _collectionUuid;

    // The background task that deletes documents from orphaned chunk ranges.
    std::shared_ptr<executor::TaskExecutor> const _executor;

    // Mutex to protect the state below
    mutable Mutex _managerLock = MONGO_MAKE_LATCH("MetadataManager::_managerLock");

    // Contains a list of collection metadata for the same collection uuid, ordered in
    // chronological order based on the refreshes that occurred. The entry at _metadata.back() is
    // the most recent metadata and is what is returned to new queries. The rest are previously
    // active collection metadata instances still in use by active server operations or cursors.
    std::list<std::shared_ptr<CollectionMetadataTracker>> _metadata;

    // Chunk ranges being migrated into to the shard. Indexed by the min key of the range.
    RangeMap _receivingChunks;

    // Ranges being deleted, or scheduled to be deleted, by a background task.
    std::list<std::pair<ChunkRange, SharedSemiFuture<void>>> _rangesScheduledForDeletion;
};

}  // namespace mongo