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
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
|
/**
* 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_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kSharding
#include "mongo/platform/basic.h"
#include "mongo/s/client/shard_connection.h"
#include <set>
#include "mongo/base/init.h"
#include "mongo/db/lasterror.h"
#include "mongo/db/server_parameters.h"
#include "mongo/s/chunk_manager.h"
#include "mongo/s/client/shard.h"
#include "mongo/s/client/shard_registry.h"
#include "mongo/s/client/version_manager.h"
#include "mongo/s/cluster_last_error_info.h"
#include "mongo/s/grid.h"
#include "mongo/s/is_mongos.h"
#include "mongo/s/transaction_router.h"
#include "mongo/util/concurrency/spin_lock.h"
#include "mongo/util/exit.h"
#include "mongo/util/log.h"
#include "mongo/util/stacktrace.h"
namespace mongo {
namespace {
class ClientConnections;
/**
* Class which tracks ClientConnections (the client connection pool) for each incoming connection,
* allowing stats access.
*/
class ActiveClientConnections {
public:
void add(const ClientConnections* cc) {
stdx::lock_guard<stdx::mutex> lg(_mutex);
_clientConnections.insert(cc);
}
void remove(const ClientConnections* cc) {
stdx::lock_guard<stdx::mutex> lg(_mutex);
_clientConnections.erase(cc);
}
void appendInfo(BSONObjBuilder* b) const;
private:
mutable stdx::mutex _mutex;
std::set<const ClientConnections*> _clientConnections;
} activeClientConnections;
/**
* Holds all the actual db connections for a client to various servers 1 per thread, so doesn't have
* to be thread safe.
*/
class ClientConnections {
MONGO_DISALLOW_COPYING(ClientConnections);
public:
struct Status {
// May be read concurrently, but only written from this thread
long long created = 0;
DBClientBase* avail = nullptr;
};
ClientConnections() {
// Start tracking client connections
activeClientConnections.add(this);
}
~ClientConnections() {
// Stop tracking these client connections
activeClientConnections.remove(this);
releaseAll(true);
}
static ClientConnections* threadInstance() {
if (!_perThread) {
_perThread = stdx::make_unique<ClientConnections>();
}
return _perThread.get();
}
DBClientBase* get(const std::string& addr, const std::string& ns) {
{
// We want to report ns stats
scoped_spinlock lock(_lock);
if (ns.size() > 0)
_seenNS.insert(ns);
}
Status* s = _getStatus(addr);
std::unique_ptr<DBClientBase> c;
if (s->avail) {
c.reset(s->avail);
s->avail = 0;
// May throw an exception
shardConnectionPool.onHandedOut(c.get());
} else {
c.reset(shardConnectionPool.get(addr));
// After, so failed creation doesn't get counted
s->created++;
}
return c.release();
}
void releaseAll(bool fromDestructor = false) {
// Don't need spinlock protection because if not in the destructor, we don't modify
// _hosts, and if in the destructor we are not accessible to external threads.
for (HostMap::iterator i = _hosts.begin(); i != _hosts.end(); ++i) {
const auto addr = i->first;
Status* ss = i->second;
invariant(ss);
if (ss->avail) {
// If we're shutting down, don't want to initiate release mechanism as it is
// slow, and isn't needed since all connections will be closed anyway.
if (globalInShutdownDeprecated()) {
if (versionManager.isVersionableCB(ss->avail)) {
versionManager.resetShardVersionCB(ss->avail);
}
delete ss->avail;
} else {
release(addr, ss->avail);
}
ss->avail = 0;
}
if (fromDestructor) {
delete ss;
}
}
if (fromDestructor) {
_hosts.clear();
}
}
void done(const std::string& addr, DBClientBase* conn) {
Status* s = _hosts[addr];
verify(s);
const bool isConnGood = shardConnectionPool.isConnectionGood(addr, conn);
if (s->avail != NULL) {
warning() << "Detected additional sharded connection in the "
<< "thread local pool for " << addr;
if (DBException::traceExceptions.load()) {
// There shouldn't be more than one connection checked out to the same
// host on the same thread.
printStackTrace();
}
if (!isConnGood) {
delete s->avail;
s->avail = NULL;
}
// Let the internal pool handle the bad connection, this can also
// update the lower bounds for the known good socket creation time
// for this host.
release(addr, conn);
return;
}
if (!isConnGood) {
// Let the internal pool handle the bad connection.
release(addr, conn);
return;
}
// Note: Although we try our best to clear bad connections as much as possible, some of them
// can still slip through because of how ClientConnections are being used - as thread local
// variables. This means that threads won't be able to see the s->avail connection of other
// threads.
s->avail = conn;
}
void checkVersions(OperationContext* opCtx, const std::string& ns) {
auto const shardRegistry = Grid::get(opCtx)->shardRegistry();
std::vector<ShardId> all;
shardRegistry->getAllShardIdsNoReload(&all);
// Don't report exceptions here as errors in GetLastError
LastError::Disabled ignoreForGLE(&LastError::get(cc()));
// Now only check top-level shard connections
for (const ShardId& shardId : all) {
try {
auto shardStatus = shardRegistry->getShard(opCtx, shardId);
if (!shardStatus.isOK()) {
invariant(shardStatus == ErrorCodes::ShardNotFound);
continue;
}
const auto shard = shardStatus.getValue();
const auto sconnString = shard->getConnString().toString();
Status* s = _getStatus(sconnString);
if (!s->avail) {
s->avail = shardConnectionPool.get(sconnString);
s->created++; // After, so failed creation doesn't get counted
}
versionManager.checkShardVersionCB(opCtx, s->avail, ns, false, 1);
} catch (const DBException& ex) {
warning() << "Problem while initially checking shard versions on"
<< " " << shardId << causedBy(redact(ex));
// NOTE: This is only a heuristic, to avoid multiple stale version retries across
// multiple shards, and does not affect correctness.
}
}
}
void release(const std::string& addr, DBClientBase* conn) {
shardConnectionPool.release(addr, conn);
}
void forgetNS(const std::string& ns) {
scoped_spinlock lock(_lock);
_seenNS.erase(ns);
}
/**
* Clears the connections kept by this pool (ie, not including the global pool)
*/
void clearPool() {
for (HostMap::iterator iter = _hosts.begin(); iter != _hosts.end(); ++iter) {
if (iter->second->avail != NULL) {
delete iter->second->avail;
}
delete iter->second;
}
_hosts.clear();
}
/**
* Appends info about the client connection pool to a BSONObjBuilder. Safe to call with
* activeClientConnections lock.
*/
void appendInfo(BSONObjBuilder& b) const {
scoped_spinlock lock(_lock);
BSONArrayBuilder hostsArrB(b.subarrayStart("hosts"));
for (HostMap::const_iterator i = _hosts.begin(); i != _hosts.end(); ++i) {
BSONObjBuilder bb(hostsArrB.subobjStart());
bb.append("host", i->first);
bb.append("created", i->second->created);
bb.appendBool("avail", static_cast<bool>(i->second->avail));
bb.done();
}
hostsArrB.done();
BSONArrayBuilder nsArrB(b.subarrayStart("seenNS"));
for (const auto& ns : _seenNS) {
nsArrB.append(ns);
}
nsArrB.done();
}
private:
/**
* Gets or creates the status object for the host.
*/
Status* _getStatus(const std::string& addr) {
scoped_spinlock lock(_lock);
Status*& temp = _hosts[addr];
if (!temp) {
temp = new Status();
}
return temp;
}
static thread_local std::unique_ptr<ClientConnections> _perThread;
// Protects only the creation of new entries in the _hosts and _seenNS map from external
// threads. Reading _hosts/_seenNS in this thread doesn't need protection.
mutable SpinLock _lock;
using HostMap = std::map<std::string, Status*, DBConnectionPool::serverNameCompare>;
HostMap _hosts;
std::set<std::string> _seenNS;
};
void ActiveClientConnections::appendInfo(BSONObjBuilder* b) const {
// Preallocate the buffer because there may be quite a few threads to report
BSONArrayBuilder arr(64 * 1024);
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
for (const auto* conn : _clientConnections) {
BSONObjBuilder bb(arr.subobjStart());
conn->appendInfo(bb);
bb.doneFast();
}
}
b->appendArray("threads", arr.obj());
}
thread_local std::unique_ptr<ClientConnections> ClientConnections::_perThread;
// Maximum connections per host the sharded conn pool should store
int maxShardedConnsPerHost(200);
ExportedServerParameter<int, ServerParameterType::kStartupOnly> //
maxShardedConnsPerHostParameter(ServerParameterSet::getGlobal(),
"connPoolMaxShardedConnsPerHost",
&maxShardedConnsPerHost);
// Maximum in-use connections per host in the sharded connection pool
int maxShardedInUseConnsPerHost(std::numeric_limits<int>::max());
ExportedServerParameter<int, ServerParameterType::kStartupOnly> //
maxShardedInUseConnsPerHostParameter(ServerParameterSet::getGlobal(),
"connPoolMaxShardedInUseConnsPerHost",
&maxShardedInUseConnsPerHost);
// Amount of time, in minutes, to keep idle connections in the sharded connection pool
int shardedConnPoolIdleTimeout(std::numeric_limits<int>::max());
ExportedServerParameter<int, ServerParameterType::kStartupOnly> //
shardedConnPoolIdleTimeoutParameter(ServerParameterSet::getGlobal(),
"shardedConnPoolIdleTimeoutMinutes",
&shardedConnPoolIdleTimeout);
MONGO_INITIALIZER(InitializeShardedConnectionPool)(InitializerContext* context) {
shardConnectionPool.setName("sharded connection pool");
shardConnectionPool.setMaxPoolSize(maxShardedConnsPerHost);
shardConnectionPool.setMaxInUse(maxShardedInUseConnsPerHost);
shardConnectionPool.setIdleTimeout(shardedConnPoolIdleTimeout);
return Status::OK();
}
} // namespace
DBConnectionPool shardConnectionPool;
ShardConnection::ShardConnection(OperationContext* opCtx,
const ConnectionString& connectionString,
const std::string& ns,
std::shared_ptr<ChunkManager> manager)
: _cs(connectionString), _ns(ns), _manager(manager) {
invariant(_cs.isValid());
// This code should never run under a cross-shard transaction
invariant(!TransactionRouter::get(opCtx));
// Make sure we specified a manager for the correct namespace
if (_ns.size() && _manager) {
invariant(_manager->getns().ns() == _ns);
}
auto csString = _cs.toString();
_conn = ClientConnections::threadInstance()->get(csString, _ns);
if (isMongos()) {
// In mongos, we record this connection as having been used for useful work to provide
// useful information in getLastError.
ClusterLastErrorInfo::get(opCtx->getClient())->addShardHost(csString);
}
}
ShardConnection::~ShardConnection() {
if (_conn) {
if (_conn->isFailed()) {
if (_conn->getSockCreationMicroSec() == DBClientBase::INVALID_SOCK_CREATION_TIME) {
kill();
} else {
// The pool takes care of deleting the failed connection - this
// will also trigger disposal of older connections in the pool
done();
}
} else {
// see done() comments above for why we log this line
log() << "sharded connection to " << _conn->getServerAddress()
<< " not being returned to the pool";
kill();
}
}
}
void ShardConnection::_finishInit() {
if (_finishedInit)
return;
_finishedInit = true;
if (versionManager.isVersionableCB(_conn)) {
auto& client = cc();
auto opCtx = client.getOperationContext();
invariant(opCtx);
_setVersion = versionManager.checkShardVersionCB(opCtx, this, false, 1);
} else {
// Make sure we didn't specify a manager for a non-versionable connection (i.e. config)
invariant(!_manager);
_setVersion = false;
}
}
void ShardConnection::done() {
if (_conn) {
ClientConnections::threadInstance()->done(_cs.toString(), _conn);
_conn = nullptr;
_finishedInit = true;
}
}
void ShardConnection::kill() {
if (_conn) {
if (versionManager.isVersionableCB(_conn)) {
versionManager.resetShardVersionCB(_conn);
}
if (_conn->isFailed()) {
// Let the pool know about the bad connection and also delegate disposal to it.
ClientConnections::threadInstance()->done(_cs.toString(), _conn);
} else {
delete _conn;
}
_conn = 0;
_finishedInit = true;
}
}
void ShardConnection::reportActiveClientConnections(BSONObjBuilder* builder) {
activeClientConnections.appendInfo(builder);
}
void ShardConnection::checkMyConnectionVersions(OperationContext* opCtx, const std::string& ns) {
ClientConnections::threadInstance()->checkVersions(opCtx, ns);
}
void ShardConnection::releaseMyConnections() {
ClientConnections::threadInstance()->releaseAll();
}
void ShardConnection::clearPool() {
shardConnectionPool.clear();
ClientConnections::threadInstance()->clearPool();
}
void ShardConnection::forgetNS(const std::string& ns) {
ClientConnections::threadInstance()->forgetNS(ns);
}
} // namespace mongo
|