summaryrefslogtreecommitdiff
path: root/src/mongo/db/s/sharding_state.h
blob: c6efdeaae30961a73e32c949cb61751721d9a463 (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
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
/**
 *    Copyright (C) 2015 MongoDB Inc.
 *
 *    This program is free software: you can redistribute it and/or  modify
 *    it under the terms of the GNU Affero General Public License, version 3,
 *    as published by the Free Software Foundation.
 *
 *    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
 *    GNU Affero General Public License for more details.
 *
 *    You should have received a copy of the GNU Affero General Public License
 *    along with this program.  If not, see <http://www.gnu.org/licenses/>.
 *
 *    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 GNU Affero General 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 <string>
#include <vector>

#include "mongo/base/disallow_copying.h"
#include "mongo/bson/oid.h"
#include "mongo/db/namespace_string.h"
#include "mongo/db/s/active_migrations_registry.h"
#include "mongo/db/s/collection_range_deleter.h"
#include "mongo/db/s/migration_destination_manager.h"
#include "mongo/executor/task_executor.h"
#include "mongo/executor/thread_pool_task_executor.h"
#include "mongo/stdx/functional.h"
#include "mongo/stdx/memory.h"
#include "mongo/stdx/mutex.h"
#include "mongo/stdx/unordered_map.h"

namespace mongo {

class BSONObj;
class BSONObjBuilder;
struct ChunkVersion;
class CollectionMetadata;
class CollectionShardingState;
class ConnectionString;
class OperationContext;
class ScopedCollectionMetadata;
class ServiceContext;
class ShardIdentityType;
class Status;

namespace repl {
class OpTime;
}  // namespace repl

/**
 * Contains the global sharding state for a running mongod. There is one instance of this object per
 * service context and it is never destroyed for the lifetime of the context.
 */
class ShardingState {
    MONGO_DISALLOW_COPYING(ShardingState);

public:
    using GlobalInitFunc =
        stdx::function<Status(OperationContext*, const ConnectionString&, StringData)>;

    ShardingState();
    ~ShardingState();

    /**
     * Retrieves the sharding state object associated with the specified service context. This
     * method must only be called if ShardingState decoration has been created on the service
     * context, otherwise it will fassert. In other words, it may only be called on MongoD and
     * tests, which specifically require and instantiate ShardingState.
     *
     * Returns the instance's ShardingState.
     */
    static ShardingState* get(ServiceContext* serviceContext);
    static ShardingState* get(OperationContext* operationContext);

    /**
     * Returns true if ShardingState has been successfully initialized.
     *
     * Code that needs to perform extra actions if sharding is initialized, but does not need to
     * error if not, should use this. Alternatively, see ShardingState::canAcceptShardedCommands().
     */
    bool enabled() const;

    /**
     * Returns Status::OK if the ShardingState is enabled; if not, returns an error describing
     * whether the ShardingState is just not yet initialized, or if this shard is not running with
     * --shardsvr at all.
     *
     * Code that should error if sharding state has not been initialized should use this to report
     * a more descriptive error. Alternatively, see ShardingState::enabled().
     */
    Status canAcceptShardedCommands() const;

    ConnectionString getConfigServer(OperationContext* opCtx);

    std::string getShardName();

    MigrationDestinationManager* migrationDestinationManager() {
        return &_migrationDestManager;
    }

    /**
     * Initializes the sharding state of this server from the shard identity document argument.
     */
    Status initializeFromShardIdentity(OperationContext* opCtx,
                                       const ShardIdentityType& shardIdentity);

    /**
     * Shuts down sharding machinery on the shard.
     */
    void shutDown(OperationContext* opCtx);

    /**
     * Updates the ShardRegistry's stored notion of the config server optime based on the
     * ConfigServerMetadata decoration attached to the OperationContext.
     */
    Status updateConfigServerOpTimeFromMetadata(OperationContext* opCtx);

    CollectionShardingState* getNS(const std::string& ns, OperationContext* opCtx);

    /**
     * Iterates through all known sharded collections and marks them (in memory only) as not sharded
     * so that no filtering will be happening for slaveOk queries.
     */
    void markCollectionsNotShardedAtStepdown();

    /**
     * Refreshes the local metadata based on whether the expected version is higher than what we
     * have cached.
     */
    Status onStaleShardVersion(OperationContext* opCtx,
                               const NamespaceString& nss,
                               const ChunkVersion& expectedVersion);

    /**
     * Refreshes collection metadata by asking the config server for the latest information.
     * Starts a new config server request.
     *
     * Locking Notes:
     *   + Must NOT be called with the write lock because this call may go into the network,
     *     and deadlocks may occur with shard-as-a-config.  Therefore, nothing here guarantees
     *     that 'latestShardVersion' is indeed the current one on return.
     *
     *   + Because this call must not be issued with the DBLock held, by the time the config
     *     server sent us back the collection metadata information, someone else may have
     *     updated the previously stored collection metadata.  There are cases when one can't
     *     tell which of updated or loaded metadata are the freshest. There are also cases where
     *     the data coming from configs do not correspond to a consistent snapshot.
     *     In these cases, return RemoteChangeDetected. (This usually means this call needs to
     *     be issued again, at caller discretion)
     *
     * @return OK if remote metadata successfully loaded (may or may not have been installed)
     * @return RemoteChangeDetected if something changed while reloading and we may retry
     * @return !OK if something else went wrong during reload
     * @return latestShardVersion the version that is now stored for this collection
     */
    Status refreshMetadataNow(OperationContext* opCtx,
                              const NamespaceString& nss,
                              ChunkVersion* latestShardVersion);

    void appendInfo(OperationContext* opCtx, BSONObjBuilder& b);

    bool needCollectionMetadata(OperationContext* opCtx, const std::string& ns);

    /**
     * Updates the config server field of the shardIdentity document with the given connection
     * string.
     *
     * Note: this can return NotMaster error.
     */
    Status updateShardIdentityConfigString(OperationContext* opCtx,
                                           const std::string& newConnectionString);

    /**
     * If there are no migrations running on this shard, registers an active migration with the
     * specified arguments and returns a ScopedRegisterDonateChunk, which must be signaled by the
     * caller before it goes out of scope.
     *
     * If there is an active migration already running on this shard and it has the exact same
     * arguments, returns a ScopedRegisterDonateChunk, which can be used to join the existing one.
     *
     * Othwerwise returns a ConflictingOperationInProgress error.
     */
    StatusWith<ScopedRegisterDonateChunk> registerDonateChunk(const MoveChunkRequest& args);

    /**
     * If there are no migrations running on this shard, registers an active receive operation with
     * the specified session id and returns a ScopedRegisterReceiveChunk, which will unregister it
     * when it goes out of scope.
     *
     * Otherwise returns a ConflictingOperationInProgress error.
     */
    StatusWith<ScopedRegisterReceiveChunk> registerReceiveChunk(const NamespaceString& nss,
                                                                const ChunkRange& chunkRange,
                                                                const ShardId& fromShardId);

    /**
     * If a migration has been previously registered through a call to registerDonateChunk returns
     * that namespace. Otherwise returns boost::none.
     *
     * This method can be called without any locks, but once the namespace is fetched it needs to be
     * re-checked after acquiring some intent lock on that namespace.
     */
    boost::optional<NamespaceString> getActiveDonateChunkNss();

    /**
     * Get a migration status report from the migration registry. If no migration is active, this
     * returns an empty BSONObj.
     *
     * Takes an IS lock on the namespace of the active migration, if one is active.
     */
    BSONObj getActiveMigrationStatusReport(OperationContext* opCtx);

    /**
     * For testing only. Mock the initialization method used by initializeFromConfigConnString and
     * initializeFromShardIdentity after all checks are performed.
     */
    void setGlobalInitMethodForTest(GlobalInitFunc func);

    /**
     * Schedules for the range to clean of the given namespace to be deleted.
     * Behavior can be modified through setScheduleCleanupFunctionForTest.
     */
    void scheduleCleanup(const NamespaceString& nss);

    /**
     * If started with --shardsvr, initializes sharding awareness from the shardIdentity document
     * on disk, if there is one.
     * If started with --shardsvr in queryableBackupMode, initializes sharding awareness from the
     * shardIdentity document passed through the --overrideShardIdentity startup parameter.
     *
     * If returns true, the ShardingState::_globalInit method was called, meaning all the core
     * classes for sharding were initialized, but no networking calls were made yet (with the
     * exception of the duplicate ShardRegistry reload in ShardRegistry::startup() (see
     * SERVER-26123). Outgoing networking calls to cluster members can now be made.
     */
    StatusWith<bool> initializeShardingAwarenessIfNeeded(OperationContext* opCtx);

    /**
     * Return the task executor to be shared by the range deleters for all collections.
     */
    executor::TaskExecutor* getRangeDeleterTaskExecutor();

private:
    // Map from a namespace into the sharding state for each collection we have
    typedef stdx::unordered_map<std::string, std::unique_ptr<CollectionShardingState>>
        CollectionShardingStateMap;

    // Progress of the sharding state initialization
    enum class InitializationState : uint32_t {
        // Initial state. The server must be under exclusive lock when this state is entered. No
        // metadata is available yet and it is not known whether there is any min optime metadata,
        // which needs to be recovered. From this state, the server may enter INITIALIZING, if a
        // recovey document is found or stay in it until initialize has been called.
        kNew,

        // Sharding state is fully usable.
        kInitialized,

        // Some initialization error occurred. The _initializationStatus variable will contain the
        // error.
        kError,
    };

    /**
     * Returns the initialization state.
     */
    InitializationState _getInitializationState() const;

    /**
     * Updates the initialization state.
     */
    void _setInitializationState(InitializationState newState);

    /**
     * Refreshes collection metadata by asking the config server for the latest information and
     * returns the latest version at the time the reload was done. This call does network I/O and
     * should never be called with a lock.
     */
    ChunkVersion _refreshMetadata(OperationContext* opCtx, const NamespaceString& nss);

    // Manages the state of the migration recipient shard
    MigrationDestinationManager _migrationDestManager;

    // Tracks the active move chunk operations running on this shard
    ActiveMigrationsRegistry _activeMigrationsRegistry;

    // Protects state below
    stdx::mutex _mutex;

    // State of the initialization of the sharding state along with any potential errors
    AtomicUInt32 _initializationState;

    // Only valid if _initializationState is kError. Contains the reason for initialization failure.
    Status _initializationStatus;

    // Signaled when ::initialize finishes.
    stdx::condition_variable _initializationFinishedCondition;

    // Sets the shard name for this host (comes through setShardVersion)
    std::string _shardName;

    // Cache of collection metadata on this shard. It is not safe to look-up values from this map
    // without holding some form of collection lock. It is only safe to add/remove values when
    // holding X lock on the respective namespace.
    CollectionShardingStateMap _collections;

    // The id for the cluster this shard belongs to.
    OID _clusterId;

    // Function for initializing the external sharding state components not owned here.
    GlobalInitFunc _globalInit;

    // Task executor shared by the collection range deleters.
    struct RangeDeleterExecutor {
        stdx::mutex lock{};
        std::unique_ptr<executor::TaskExecutor> taskExecutor{nullptr};
        ~RangeDeleterExecutor();
    };
    RangeDeleterExecutor _rangeDeleterExecutor;
};

}  // namespace mongo