diff options
Diffstat (limited to 'src/mongo/db/s/analyze_shard_key_read_write_distribution.cpp')
| -rw-r--r-- | src/mongo/db/s/analyze_shard_key_read_write_distribution.cpp | 204 |
1 files changed, 74 insertions, 130 deletions
diff --git a/src/mongo/db/s/analyze_shard_key_read_write_distribution.cpp b/src/mongo/db/s/analyze_shard_key_read_write_distribution.cpp index 55db53ef6ce..70bbeef8aab 100644 --- a/src/mongo/db/s/analyze_shard_key_read_write_distribution.cpp +++ b/src/mongo/db/s/analyze_shard_key_read_write_distribution.cpp @@ -48,40 +48,6 @@ namespace mongo { namespace analyze_shard_key { -namespace { - -/** - * Returns true if the query that specifies the given collation against the collection with the - * given default collator has simple collation. - */ -bool hasSimpleCollation(const CollatorInterface* defaultCollator, const BSONObj& collation) { - if (collation.isEmpty()) { - return !defaultCollator; - } - return SimpleBSONObjComparator::kInstance.evaluate(collation == CollationSpec::kSimpleSpec); -} - -/** - * Returns true if the given shard key contains any collatable fields (ones that can be affected in - * comparison or sort order by collation). - */ -bool shardKeyHasCollatableType(const ShardKeyPattern& shardKeyPattern, const BSONObj& shardKey) { - for (const BSONElement& elt : shardKey) { - if (CollationIndexKey::isCollatableType(elt.type())) { - return true; - } - if (shardKeyPattern.isHashedPattern() && - shardKeyPattern.getHashedField().fieldNameStringData() == elt.fieldNameStringData()) { - // If the field is specified as "hashed" in the shard key pattern, then the hash value - // could have come from a collatable type. - return true; - } - } - return false; -} - -} // namespace - template <typename DistributionMetricsType, typename SampleSizeType> DistributionMetricsType DistributionMetricsCalculator<DistributionMetricsType, SampleSizeType>::_getMetrics() const { @@ -106,23 +72,13 @@ DistributionMetricsCalculator<DistributionMetricsType, SampleSizeType>::_getMetr } template <typename DistributionMetricsType, typename SampleSizeType> -BSONObj -DistributionMetricsCalculator<DistributionMetricsType, SampleSizeType>::_incrementMetricsForQuery( +QueryTargetingInfo +DistributionMetricsCalculator<DistributionMetricsType, SampleSizeType>::_getTargetingInfoForQuery( OperationContext* opCtx, - const BSONObj& primaryFilter, + const BSONObj& filter, const BSONObj& collation, - const BSONObj& secondaryFilter, - const boost::optional<LegacyRuntimeConstants>& runtimeConstants, - const boost::optional<BSONObj>& letParameters) { - auto filter = primaryFilter; - auto shardKey = uassertStatusOK(extractShardKeyFromBasicQuery( - opCtx, _targeter.getNS(), _getShardKeyPattern(), primaryFilter)); - if (shardKey.isEmpty() && !secondaryFilter.isEmpty()) { - shardKey = _getShardKeyPattern().extractShardKeyFromDoc(secondaryFilter); - filter = shardKey; - } - - // Increment metrics about range targeting. + const boost::optional<BSONObj>& letParameters, + const boost::optional<LegacyRuntimeConstants>& runtimeConstants) { auto&& cif = [&]() { if (collation.isEmpty()) { return std::unique_ptr<CollatorInterface>{}; @@ -131,48 +87,51 @@ DistributionMetricsCalculator<DistributionMetricsType, SampleSizeType>::_increme CollatorFactoryInterface::get(opCtx->getServiceContext())->makeFromBSON(collation)); }(); auto expCtx = make_intrusive<ExpressionContext>( - opCtx, std::move(cif), _getChunkManager().getNss(), runtimeConstants, letParameters); + opCtx, + std::move(cif), + _getChunkManager().getNss(), + runtimeConstants.value_or(Variables::generateRuntimeConstants(opCtx)), + letParameters); std::set<ShardId> shardIds; // This is not used. - std::set<ChunkRange> chunkRanges; - bool targetMinkeyToMaxKey = false; - getShardIdsForQuery(expCtx, - filter, - collation, - _getChunkManager(), - &shardIds, - &chunkRanges, - &targetMinkeyToMaxKey); - _incrementNumByRanges(chunkRanges); + QueryTargetingInfo info; + getShardIdsForQuery(expCtx, filter, collation, _getChunkManager(), &shardIds, &info); + + return info; +} - // Increment metrics about sharding targeting. - if (!shardKey.isEmpty()) { - // This query filters by shard key equality. If the query has a simple collation or the - // shard key doesn't contain a collatable field, then there is only one matching shard key - // value so the query is guaranteed to target only one shard. Otherwise, the number of - // shards that it targets depend on how the matching shard key values are distributed among - // shards. Given this, pessimistically classify it as targeting to multiple shards. - invariant(!targetMinkeyToMaxKey); - if (hasSimpleCollation(_getDefaultCollator(), collation) || - !shardKeyHasCollatableType(_getShardKeyPattern(), shardKey)) { +template <typename DistributionMetricsType, typename SampleSizeType> +void DistributionMetricsCalculator<DistributionMetricsType, SampleSizeType>:: + _incrementMetricsForQuery(const QueryTargetingInfo& info) { + // Increment metrics about range targeting. + _incrementNumByRanges(info.chunkRanges); + + // Increment metrics about shard targeting. + switch (info.desc) { + case QueryTargetingInfo::Description::kSingleKey: { _incrementNumSingleShard(); - invariant(chunkRanges.size() == 1U); - } else { + tassert(7531200, + "Found a point query that targets multiple chunks", + info.chunkRanges.size() == 1U); + break; + } + case QueryTargetingInfo::Description::kMultipleKeys: { + // This query targets a subset of the shard key space. Therefore, the number of shards + // that it targets depends on how the matching shard key ranges are distributed among + // shards. Given this, pessimistically classify it as targeting to multiple shards. _incrementNumMultiShard(); + break; + } + case QueryTargetingInfo::Description::kMinKeyToMaxKey: { + // This query targets the entire shard key space. Therefore, it always targets all + // shards and chunks. + _incrementNumScatterGather(); + invariant((int)info.chunkRanges.size() == _getChunkManager().numChunks()); + break; } - } else if (targetMinkeyToMaxKey) { - // This query targets the entire shard key space. Therefore, it always targets all - // shards and chunks. - _incrementNumScatterGather(); - invariant((int)chunkRanges.size() == _getChunkManager().numChunks()); - } else { - // This query targets a subset of the shard key space. Therefore, the number of shards - // that it targets depends on how the matching shard key ranges are distributed among - // shards. Given this, pessimistically classify it as targeting to multiple shards. - _incrementNumMultiShard(); + default: + MONGO_UNREACHABLE; } - - return shardKey; } ReadSampleSize ReadDistributionMetricsCalculator::_getSampleSize() const { @@ -210,7 +169,8 @@ void ReadDistributionMetricsCalculator::addQuery(OperationContext* opCtx, auto cmd = SampledReadCommand::parse(IDLParserContext("ReadDistributionMetricsCalculator"), doc.getCmd()); - _incrementMetricsForQuery(opCtx, cmd.getFilter(), cmd.getCollation()); + auto info = _getTargetingInfoForQuery(opCtx, cmd.getFilter(), cmd.getCollation()); + _incrementMetricsForQuery(info); } WriteSampleSize WriteDistributionMetricsCalculator::_getSampleSize() const { @@ -269,16 +229,18 @@ void WriteDistributionMetricsCalculator::_addUpdateQuery( OperationContext* opCtx, const write_ops::UpdateCommandRequest& cmd) { for (const auto& updateOp : cmd.getUpdates()) { _numUpdate++; - auto primaryFilter = updateOp.getQ(); auto collation = write_ops::collationOf(updateOp); - // If this is a non-upsert replacement update, the replacement document can be used as a - // filter. - auto secondaryFilter = [&] { + auto info = _getTargetingInfoForQuery( + opCtx, updateOp.getQ(), collation, cmd.getLet(), cmd.getLegacyRuntimeConstants()); + + if (info.desc != QueryTargetingInfo::Description::kSingleKey) { + // If this is a non-upsert replacement update, the replacement document can be used as + // the filter. auto isReplacementUpdate = !updateOp.getUpsert() && updateOp.getU().type() == write_ops::UpdateModification::Type::kReplacement; auto isExactIdQuery = [&] { return CollectionRoutingInfoTargeter::isExactIdQuery( - opCtx, cmd.getNamespace(), primaryFilter, collation, _getChunkManager()); + opCtx, cmd.getNamespace(), updateOp.getQ(), collation, _getChunkManager()); }; // Currently, targeting by replacement document is only done when an updateOne without @@ -287,17 +249,13 @@ void WriteDistributionMetricsCalculator::_addUpdateQuery( (!feature_flags::gFeatureFlagUpdateOneWithoutShardKey.isEnabled( serverGlobalParams.featureCompatibility) || isExactIdQuery())) { - return updateOp.getU().getUpdateReplacement(); + auto filter = _getShardKeyPattern().extractShardKeyFromDoc( + updateOp.getU().getUpdateReplacement()); + info = _getTargetingInfoForQuery( + opCtx, filter, collation, cmd.getLet(), cmd.getLegacyRuntimeConstants()); } - return BSONObj(); - }(); - _incrementMetricsForQuery(opCtx, - primaryFilter, - secondaryFilter, - collation, - updateOp.getMulti(), - cmd.getLegacyRuntimeConstants(), - cmd.getLet()); + } + _incrementMetricsForQuery(info, updateOp.getMulti()); } } @@ -305,44 +263,30 @@ void WriteDistributionMetricsCalculator::_addDeleteQuery( OperationContext* opCtx, const write_ops::DeleteCommandRequest& cmd) { for (const auto& deleteOp : cmd.getDeletes()) { _numDelete++; - auto primaryFilter = deleteOp.getQ(); - auto secondaryFilter = BSONObj(); - _incrementMetricsForQuery(opCtx, - primaryFilter, - secondaryFilter, - write_ops::collationOf(deleteOp), - deleteOp.getMulti(), - cmd.getLegacyRuntimeConstants(), - cmd.getLet()); + auto info = _getTargetingInfoForQuery(opCtx, + deleteOp.getQ(), + write_ops::collationOf(deleteOp), + cmd.getLet(), + cmd.getLegacyRuntimeConstants()); + _incrementMetricsForQuery(info, deleteOp.getMulti()); } } void WriteDistributionMetricsCalculator::_addFindAndModifyQuery( OperationContext* opCtx, const write_ops::FindAndModifyCommandRequest& cmd) { _numFindAndModify++; - auto primaryFilter = cmd.getQuery(); - auto secondaryFilter = BSONObj(); - _incrementMetricsForQuery(opCtx, - primaryFilter, - secondaryFilter, - cmd.getCollation().value_or(BSONObj()), - false /* isMulti */, - cmd.getLegacyRuntimeConstants(), - cmd.getLet()); + auto info = _getTargetingInfoForQuery(opCtx, + cmd.getQuery(), + cmd.getCollation().value_or(BSONObj()), + cmd.getLet(), + cmd.getLegacyRuntimeConstants()); + _incrementMetricsForQuery(info, false /* isMulti */); } -void WriteDistributionMetricsCalculator::_incrementMetricsForQuery( - OperationContext* opCtx, - const BSONObj& primaryFilter, - const BSONObj& secondaryFilter, - const BSONObj& collation, - bool isMulti, - const boost::optional<LegacyRuntimeConstants>& runtimeConstants, - const boost::optional<BSONObj>& letParameters) { - auto shardKey = DistributionMetricsCalculator::_incrementMetricsForQuery( - opCtx, primaryFilter, collation, secondaryFilter, runtimeConstants, letParameters); - - if (shardKey.isEmpty()) { +void WriteDistributionMetricsCalculator::_incrementMetricsForQuery(const QueryTargetingInfo& info, + bool isMulti) { + DistributionMetricsCalculator::_incrementMetricsForQuery(info); + if (info.desc != QueryTargetingInfo::Description::kSingleKey) { // Increment metrics about writes without shard key. if (isMulti) { _incrementNumMultiWritesWithoutShardKey(); |