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
|
// Copyright Yahoo. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
#include "splitjoinhandler.h"
#include "persistenceutil.h"
#include "bucketownershipnotifier.h"
#include "splitbitdetector.h"
#include "messages.h"
#include <vespa/persistence/spi/persistenceprovider.h>
#include <vespa/storageapi/message/bucket.h>
#include <vespa/log/log.h>
LOG_SETUP(".persistence.splitjoinhandler");
namespace storage {
SplitJoinHandler::SplitJoinHandler(PersistenceUtil & env, spi::PersistenceProvider & spi,
BucketOwnershipNotifier & notifier, bool enableMultibitSplitOptimalization)
: _env(env),
_spi(spi),
_bucketOwnershipNotifier(notifier),
_enableMultibitSplitOptimalization(enableMultibitSplitOptimalization)
{
}
MessageTracker::UP
SplitJoinHandler::handleSplitBucket(api::SplitBucketCommand& cmd, MessageTracker::UP tracker) const
{
tracker->setMetric(_env._metrics.splitBuckets);
NotificationGuard notifyGuard(_bucketOwnershipNotifier);
// Calculate the various bucket ids involved.
if (cmd.getBucketId().getUsedBits() >= 58) {
tracker->fail(api::ReturnCode::ILLEGAL_PARAMETERS,
"Can't split anymore since maximum split bits is already reached");
return tracker;
}
if (cmd.getMaxSplitBits() <= cmd.getBucketId().getUsedBits()) {
tracker->fail(api::ReturnCode::ILLEGAL_PARAMETERS,
"Max lit bits must be set higher than the number of bits used in the bucket to split");
return tracker;
}
spi::Bucket spiBucket(cmd.getBucket());
SplitBitDetector::Result targetInfo;
if (_enableMultibitSplitOptimalization) {
targetInfo = SplitBitDetector::detectSplit(_spi, spiBucket, cmd.getMaxSplitBits(),
tracker->context(), cmd.getMinDocCount(), cmd.getMinByteSize());
}
if (targetInfo.empty() || !_enableMultibitSplitOptimalization) {
document::BucketId src(cmd.getBucketId());
document::BucketId target1(src.getUsedBits() + 1, src.getId());
document::BucketId target2(src.getUsedBits() + 1, src.getId() | (uint64_t(1) << src.getUsedBits()));
targetInfo = SplitBitDetector::Result(target1, target2, false);
}
if (targetInfo.failed()) {
tracker->fail(api::ReturnCode::INTERNAL_FAILURE, targetInfo.getReason());
return tracker;
}
// If we get here, we're splitting data in two.
// (Possibly in special case where a target will be unused)
assert(targetInfo.success());
document::Bucket target1(spiBucket.getBucketSpace(), targetInfo.getTarget1());
document::Bucket target2(spiBucket.getBucketSpace(), targetInfo.getTarget2());
LOG(debug, "split(%s -> %s, %s)", cmd.getBucketId().toString().c_str(),
target1.getBucketId().toString().c_str(), target2.getBucketId().toString().c_str());
PersistenceUtil::LockResult lock1(_env.lockAndGetDisk(target1));
PersistenceUtil::LockResult lock2(_env.lockAndGetDisk(target2));
spi::Result result = _spi.split(spiBucket, spi::Bucket(target1), spi::Bucket(target2));
if (result.hasError()) {
tracker->fail(PersistenceUtil::convertErrorCode(result), result.getErrorMessage());
return tracker;
}
// After split we need to take all bucket db locks to update them.
// Ensure to take them in rising order.
StorBucketDatabase::WrappedEntry sourceEntry(_env.getBucketDatabase(spiBucket.getBucket().getBucketSpace()).get(
cmd.getBucketId(), "PersistenceThread::handleSplitBucket-source"));
auto reply = std::make_shared<api::SplitBucketReply>(cmd);
api::SplitBucketReply & splitReply = *reply;
tracker->setReply(std::move(reply));
using TargetInfo = std::pair<StorBucketDatabase::WrappedEntry, FileStorHandler::RemapInfo>;
std::vector<TargetInfo> targets;
for (uint32_t i = 0; i < 2; i++) {
const document::Bucket &target(i == 0 ? target1 : target2);
assert(target.getBucketId().getRawId() != 0);
targets.emplace_back(_env.getBucketDatabase(target.getBucketSpace()).get(
target.getBucketId(), "PersistenceThread::handleSplitBucket - Target",
StorBucketDatabase::CREATE_IF_NONEXISTING),
FileStorHandler::RemapInfo(target));
targets.back().first->setBucketInfo(_env.getBucketInfo(target));
}
if (LOG_WOULD_LOG(spam)) {
api::BucketInfo targ1(targets[0].first->getBucketInfo());
api::BucketInfo targ2(targets[1].first->getBucketInfo());
LOG(spam, "split(%s - %u -> %s - %u, %s - %u)",
cmd.getBucketId().toString().c_str(),
targ1.getMetaCount() + targ2.getMetaCount(),
target1.getBucketId().toString().c_str(),
targ1.getMetaCount(),
target2.getBucketId().toString().c_str(),
targ2.getMetaCount());
}
FileStorHandler::RemapInfo source(cmd.getBucket());
_env._fileStorHandler.remapQueueAfterSplit(source, targets[0].second, targets[1].second);
bool ownershipChanged(!_bucketOwnershipNotifier.distributorOwns(cmd.getSourceIndex(), cmd.getBucket()));
// Now release all the bucketdb locks.
for (auto & target : targets) {
if (ownershipChanged) {
notifyGuard.notifyAlways(target.second.bucket, target.first->getBucketInfo());
}
// The entries vector has the source bucket in element zero, so indexing
// that with i+1
if (target.second.foundInQueue || target.first->getMetaCount() > 0) {
if (target.first->getMetaCount() == 0) {
// Fake that the bucket has content so it is not deleted.
target.first->info.setMetaCount(1);
// Must make sure target bucket exists when we have pending ops
// to an empty target bucket, since the provider will have
// implicitly erased it by this point.
spi::Bucket createTarget(target.second.bucket);
LOG(debug, "Split target %s was empty, but re-creating it since there are remapped operations queued to it",
createTarget.toString().c_str());
_spi.createBucket(createTarget);
}
splitReply.getSplitInfo().emplace_back(target.second.bucket.getBucketId(),
target.first->getBucketInfo());
target.first.write();
} else {
target.first.remove();
}
}
if (sourceEntry.exists()) {
if (ownershipChanged) {
notifyGuard.notifyAlways(cmd.getBucket(), sourceEntry->getBucketInfo());
}
// Delete the old entry.
sourceEntry.remove();
}
return tracker;
}
MessageTracker::UP
SplitJoinHandler::handleRecheckBucketInfo(RecheckBucketInfoCommand& cmd, MessageTracker::UP tracker) const
{
tracker->setMetric(_env._metrics.recheckBucketInfo);
document::Bucket bucket(cmd.getBucket());
api::BucketInfo info(_env.getBucketInfo(bucket));
NotificationGuard notifyGuard(_bucketOwnershipNotifier);
{
// Update bucket database
StorBucketDatabase::WrappedEntry entry(
_env.getBucketDatabase(bucket.getBucketSpace()).get(bucket.getBucketId(), "handleRecheckBucketInfo"));
if (entry.exists()) {
api::BucketInfo prevInfo(entry->getBucketInfo());
if (!(prevInfo == info)) {
notifyGuard.notifyAlways(bucket, info);
entry->info = info;
entry.write();
}
}
// else: there is a race condition where concurrent execution of
// DeleteBucket in the FileStorManager and this function can cause it
// to look like the provider has a bucket we do not know about, simply
// because this function was executed before the actual
// DeleteBucketCommand in the persistence thread (see ticket 6143025).
}
return tracker;
}
MessageTracker::UP
SplitJoinHandler::handleJoinBuckets(api::JoinBucketsCommand& cmd, MessageTracker::UP tracker) const
{
tracker->setMetric(_env._metrics.joinBuckets);
if (!validateJoinCommand(cmd, *tracker)) {
return tracker;
}
document::Bucket destBucket = cmd.getBucket();
// To avoid a potential deadlock all operations locking multiple
// buckets must lock their buckets in the same order (sort order of
// bucket id, lowest countbits, lowest location first).
// Sort buckets to join in order to ensure we lock in correct order
std::sort(cmd.getSourceBuckets().begin(), cmd.getSourceBuckets().end());
{
// Create empty bucket for target.
StorBucketDatabase::WrappedEntry entry =
_env.getBucketDatabase(destBucket.getBucketSpace()).get(destBucket.getBucketId(), "join",
StorBucketDatabase::CREATE_IF_NONEXISTING);
entry.write();
}
document::Bucket firstBucket(destBucket.getBucketSpace(), cmd.getSourceBuckets()[0]);
document::Bucket secondBucket(destBucket.getBucketSpace(), cmd.getSourceBuckets()[1]);
PersistenceUtil::LockResult lock1(_env.lockAndGetDisk(firstBucket));
PersistenceUtil::LockResult lock2;
if (firstBucket != secondBucket) {
lock2 = _env.lockAndGetDisk(secondBucket);
}
spi::Result result = _spi.join(spi::Bucket(firstBucket), spi::Bucket(secondBucket), spi::Bucket(destBucket));
if (!tracker->checkForError(result)) {
return tracker;
}
uint64_t lastModified = 0;
for (uint32_t i = 0; i < cmd.getSourceBuckets().size(); i++) {
document::Bucket srcBucket(destBucket.getBucketSpace(), cmd.getSourceBuckets()[i]);
FileStorHandler::RemapInfo target(cmd.getBucket());
_env._fileStorHandler.remapQueueAfterJoin(FileStorHandler::RemapInfo(srcBucket), target);
// Remove source from bucket db.
StorBucketDatabase::WrappedEntry entry =
_env.getBucketDatabase(srcBucket.getBucketSpace()).get(srcBucket.getBucketId(), "join-remove-source");
if (entry.exists()) {
lastModified = std::max(lastModified, entry->info.getLastModified());
entry.remove();
}
}
{
StorBucketDatabase::WrappedEntry entry =
_env.getBucketDatabase(destBucket.getBucketSpace()).get(destBucket.getBucketId(), "join",
StorBucketDatabase::CREATE_IF_NONEXISTING);
if (entry->info.getLastModified() == 0) {
entry->info.setLastModified(std::max(lastModified, entry->info.getLastModified()));
}
entry.write();
}
return tracker;
}
bool
SplitJoinHandler::validateJoinCommand(const api::JoinBucketsCommand& cmd, MessageTracker& tracker)
{
if (cmd.getSourceBuckets().size() != 2) {
tracker.fail(api::ReturnCode::ILLEGAL_PARAMETERS,
"Join needs exactly two buckets to be joined together" + cmd.getBucketId().toString());
return false;
}
// Verify that source and target buckets look sane.
for (uint32_t i = 0; i < cmd.getSourceBuckets().size(); i++) {
if (cmd.getSourceBuckets()[i] == cmd.getBucketId()) {
tracker.fail(api::ReturnCode::ILLEGAL_PARAMETERS,
"Join had both source and target bucket " + cmd.getBucketId().toString());
return false;
}
if (!cmd.getBucketId().contains(cmd.getSourceBuckets()[i])) {
tracker.fail(api::ReturnCode::ILLEGAL_PARAMETERS,
"Source bucket " + cmd.getSourceBuckets()[i].toString()
+ " is not contained in target " + cmd.getBucketId().toString());
return false;
}
}
return true;
}
}
|
