rocksdb/table/table_properties.cc
Hui Xiao 49623f9c8e Account memory of big memory users in BlockBasedTable in global memory limit (#9748)
Summary:
**Context:**
Through heap profiling, we discovered that `BlockBasedTableReader` objects can accumulate and lead to high memory usage (e.g, `max_open_file = -1`). These memories are currently not saved, not tracked, not constrained and not cache evict-able. As a first step to improve this, similar to https://github.com/facebook/rocksdb/pull/8428,  this PR is to track an estimate of `BlockBasedTableReader` object's memory in block cache and fail future creation if the memory usage exceeds the available space of cache at the time of creation.

**Summary:**
- Approximate big memory users  (`BlockBasedTable::Rep` and `TableProperties` )' memory usage in addition to the existing estimated ones (filter block/index block/un-compression dictionary)
- Charge all of these memory usages to block cache on `BlockBasedTable::Open()` and release them on `~BlockBasedTable()` as there is no memory usage fluctuation of concern in between
- Refactor on CacheReservationManager (and its call-sites) to add concurrent support for BlockBasedTable  used in this PR.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/9748

Test Plan:
- New unit tests
- db bench: `OpenDb` : **-0.52% in ms**
  - Setup `./db_bench -benchmarks=fillseq -db=/dev/shm/testdb -disable_auto_compactions=1 -write_buffer_size=1048576`
  - Repeated run with pre-change w/o feature and post-change with feature, benchmark `OpenDb`:  `./db_bench -benchmarks=readrandom -use_existing_db=1 -db=/dev/shm/testdb -reserve_table_reader_memory=true (remove this when running w/o feature) -file_opening_threads=3 -open_files=-1 -report_open_timing=true| egrep 'OpenDb:'`

#-run | (feature-off) avg milliseconds | std milliseconds | (feature-on) avg milliseconds | std milliseconds | change (%)
-- | -- | -- | -- | -- | --
10 | 11.4018 | 5.95173 | 9.47788 | 1.57538 | -16.87382694
20 | 9.23746 | 0.841053 | 9.32377 | 1.14074 | 0.9343477536
40 | 9.0876 | 0.671129 | 9.35053 | 1.11713 | 2.893283155
80 | 9.72514 | 2.28459 | 9.52013 | 1.0894 | -2.108041632
160 | 9.74677 | 0.991234 | 9.84743 | 1.73396 | 1.032752389
320 | 10.7297 | 5.11555 | 10.547 | 1.97692 | **-1.70275031**
640 | 11.7092 | 2.36565 | 11.7869 | 2.69377 | **0.6635807741**

-  db bench on write with cost to cache in WriteBufferManager (just in case this PR's CRM refactoring accidentally slows down anything in WBM) : `fillseq` : **+0.54% in micros/op**
`./db_bench -benchmarks=fillseq -db=/dev/shm/testdb -disable_auto_compactions=1 -cost_write_buffer_to_cache=true -write_buffer_size=10000000000 | egrep 'fillseq'`

#-run | (pre-PR) avg micros/op | std micros/op | (post-PR)  avg micros/op | std micros/op | change (%)
-- | -- | -- | -- | -- | --
10 | 6.15 | 0.260187 | 6.289 | 0.371192 | 2.260162602
20 | 7.28025 | 0.465402 | 7.37255 | 0.451256 | 1.267813605
40 | 7.06312 | 0.490654 | 7.13803 | 0.478676 | **1.060579461**
80 | 7.14035 | 0.972831 | 7.14196 | 0.92971 | **0.02254791432**

-  filter bench: `bloom filter`: **-0.78% in ms/key**
    - ` ./filter_bench -impl=2 -quick -reserve_table_builder_memory=true | grep 'Build avg'`

#-run | (pre-PR) avg ns/key | std ns/key | (post-PR)  ns/key | std ns/key | change (%)
-- | -- | -- | -- | -- | --
10 | 26.4369 | 0.442182 | 26.3273 | 0.422919 | **-0.4145720565**
20 | 26.4451 | 0.592787 | 26.1419 | 0.62451 | **-1.1465262**

- Crash test `python3 tools/db_crashtest.py blackbox --reserve_table_reader_memory=1 --cache_size=1` killed as normal

Reviewed By: ajkr

Differential Revision: D35136549

Pulled By: hx235

fbshipit-source-id: 146978858d0f900f43f4eb09bfd3e83195e3be28
2022-04-06 10:33:00 -07:00

337 lines
13 KiB
C++

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
#include "rocksdb/table_properties.h"
#include "port/malloc.h"
#include "port/port.h"
#include "rocksdb/env.h"
#include "rocksdb/unique_id.h"
#include "table/table_properties_internal.h"
#include "table/unique_id_impl.h"
#include "util/random.h"
#include "util/string_util.h"
namespace ROCKSDB_NAMESPACE {
const uint32_t TablePropertiesCollectorFactory::Context::kUnknownColumnFamily =
port::kMaxInt32;
namespace {
void AppendProperty(
std::string& props,
const std::string& key,
const std::string& value,
const std::string& prop_delim,
const std::string& kv_delim) {
props.append(key);
props.append(kv_delim);
props.append(value);
props.append(prop_delim);
}
template <class TValue>
void AppendProperty(
std::string& props,
const std::string& key,
const TValue& value,
const std::string& prop_delim,
const std::string& kv_delim) {
AppendProperty(
props, key, ToString(value), prop_delim, kv_delim
);
}
}
std::string TableProperties::ToString(
const std::string& prop_delim,
const std::string& kv_delim) const {
std::string result;
result.reserve(1024);
// Basic Info
AppendProperty(result, "# data blocks", num_data_blocks, prop_delim,
kv_delim);
AppendProperty(result, "# entries", num_entries, prop_delim, kv_delim);
AppendProperty(result, "# deletions", num_deletions, prop_delim, kv_delim);
AppendProperty(result, "# merge operands", num_merge_operands, prop_delim,
kv_delim);
AppendProperty(result, "# range deletions", num_range_deletions, prop_delim,
kv_delim);
AppendProperty(result, "raw key size", raw_key_size, prop_delim, kv_delim);
AppendProperty(result, "raw average key size",
num_entries != 0 ? 1.0 * raw_key_size / num_entries : 0.0,
prop_delim, kv_delim);
AppendProperty(result, "raw value size", raw_value_size, prop_delim,
kv_delim);
AppendProperty(result, "raw average value size",
num_entries != 0 ? 1.0 * raw_value_size / num_entries : 0.0,
prop_delim, kv_delim);
AppendProperty(result, "data block size", data_size, prop_delim, kv_delim);
char index_block_size_str[80];
snprintf(index_block_size_str, sizeof(index_block_size_str),
"index block size (user-key? %d, delta-value? %d)",
static_cast<int>(index_key_is_user_key),
static_cast<int>(index_value_is_delta_encoded));
AppendProperty(result, index_block_size_str, index_size, prop_delim,
kv_delim);
if (index_partitions != 0) {
AppendProperty(result, "# index partitions", index_partitions, prop_delim,
kv_delim);
AppendProperty(result, "top-level index size", top_level_index_size, prop_delim,
kv_delim);
}
AppendProperty(result, "filter block size", filter_size, prop_delim,
kv_delim);
AppendProperty(result, "# entries for filter", num_filter_entries, prop_delim,
kv_delim);
AppendProperty(result, "(estimated) table size",
data_size + index_size + filter_size, prop_delim, kv_delim);
AppendProperty(
result, "filter policy name",
filter_policy_name.empty() ? std::string("N/A") : filter_policy_name,
prop_delim, kv_delim);
AppendProperty(result, "prefix extractor name",
prefix_extractor_name.empty() ? std::string("N/A")
: prefix_extractor_name,
prop_delim, kv_delim);
AppendProperty(result, "column family ID",
column_family_id ==
ROCKSDB_NAMESPACE::TablePropertiesCollectorFactory::
Context::kUnknownColumnFamily
? std::string("N/A")
: ROCKSDB_NAMESPACE::ToString(column_family_id),
prop_delim, kv_delim);
AppendProperty(
result, "column family name",
column_family_name.empty() ? std::string("N/A") : column_family_name,
prop_delim, kv_delim);
AppendProperty(result, "comparator name",
comparator_name.empty() ? std::string("N/A") : comparator_name,
prop_delim, kv_delim);
AppendProperty(
result, "merge operator name",
merge_operator_name.empty() ? std::string("N/A") : merge_operator_name,
prop_delim, kv_delim);
AppendProperty(result, "property collectors names",
property_collectors_names.empty() ? std::string("N/A")
: property_collectors_names,
prop_delim, kv_delim);
AppendProperty(
result, "SST file compression algo",
compression_name.empty() ? std::string("N/A") : compression_name,
prop_delim, kv_delim);
AppendProperty(
result, "SST file compression options",
compression_options.empty() ? std::string("N/A") : compression_options,
prop_delim, kv_delim);
AppendProperty(result, "creation time", creation_time, prop_delim, kv_delim);
AppendProperty(result, "time stamp of earliest key", oldest_key_time,
prop_delim, kv_delim);
AppendProperty(result, "file creation time", file_creation_time, prop_delim,
kv_delim);
AppendProperty(result, "slow compression estimated data size",
slow_compression_estimated_data_size, prop_delim, kv_delim);
AppendProperty(result, "fast compression estimated data size",
fast_compression_estimated_data_size, prop_delim, kv_delim);
// DB identity and DB session ID
AppendProperty(result, "DB identity", db_id, prop_delim, kv_delim);
AppendProperty(result, "DB session identity", db_session_id, prop_delim,
kv_delim);
AppendProperty(result, "DB host id", db_host_id, prop_delim, kv_delim);
AppendProperty(result, "original file number", orig_file_number, prop_delim,
kv_delim);
// Unique ID, when available
std::string id;
Status s = GetUniqueIdFromTableProperties(*this, &id);
AppendProperty(result, "unique ID",
s.ok() ? UniqueIdToHumanString(id) : "N/A", prop_delim,
kv_delim);
return result;
}
void TableProperties::Add(const TableProperties& tp) {
data_size += tp.data_size;
index_size += tp.index_size;
index_partitions += tp.index_partitions;
top_level_index_size += tp.top_level_index_size;
index_key_is_user_key += tp.index_key_is_user_key;
index_value_is_delta_encoded += tp.index_value_is_delta_encoded;
filter_size += tp.filter_size;
raw_key_size += tp.raw_key_size;
raw_value_size += tp.raw_value_size;
num_data_blocks += tp.num_data_blocks;
num_entries += tp.num_entries;
num_filter_entries += tp.num_filter_entries;
num_deletions += tp.num_deletions;
num_merge_operands += tp.num_merge_operands;
num_range_deletions += tp.num_range_deletions;
slow_compression_estimated_data_size +=
tp.slow_compression_estimated_data_size;
fast_compression_estimated_data_size +=
tp.fast_compression_estimated_data_size;
}
std::map<std::string, uint64_t>
TableProperties::GetAggregatablePropertiesAsMap() const {
std::map<std::string, uint64_t> rv;
rv["data_size"] = data_size;
rv["index_size"] = index_size;
rv["index_partitions"] = index_partitions;
rv["top_level_index_size"] = top_level_index_size;
rv["filter_size"] = filter_size;
rv["raw_key_size"] = raw_key_size;
rv["raw_value_size"] = raw_value_size;
rv["num_data_blocks"] = num_data_blocks;
rv["num_entries"] = num_entries;
rv["num_filter_entries"] = num_filter_entries;
rv["num_deletions"] = num_deletions;
rv["num_merge_operands"] = num_merge_operands;
rv["num_range_deletions"] = num_range_deletions;
rv["slow_compression_estimated_data_size"] =
slow_compression_estimated_data_size;
rv["fast_compression_estimated_data_size"] =
fast_compression_estimated_data_size;
return rv;
}
// WARNING: manual update to this function is needed
// whenever a new string property is added to TableProperties
// to reduce approximation error.
//
// TODO: eliminate the need of manually updating this function
// for new string properties
std::size_t TableProperties::ApproximateMemoryUsage() const {
std::size_t usage = 0;
#ifdef ROCKSDB_MALLOC_USABLE_SIZE
usage += malloc_usable_size((void*)this);
#else
usage += sizeof(*this);
#endif // ROCKSDB_MALLOC_USABLE_SIZE
std::size_t string_props_mem_usage =
db_id.size() + db_session_id.size() + db_host_id.size() +
column_family_name.size() + filter_policy_name.size() +
comparator_name.size() + merge_operator_name.size() +
prefix_extractor_name.size() + property_collectors_names.size() +
compression_name.size() + compression_options.size();
usage += string_props_mem_usage;
for (auto iter = user_collected_properties.begin();
iter != user_collected_properties.end(); ++iter) {
usage += (iter->first.size() + iter->second.size());
}
return usage;
}
const std::string TablePropertiesNames::kDbId = "rocksdb.creating.db.identity";
const std::string TablePropertiesNames::kDbSessionId =
"rocksdb.creating.session.identity";
const std::string TablePropertiesNames::kDbHostId =
"rocksdb.creating.host.identity";
const std::string TablePropertiesNames::kOriginalFileNumber =
"rocksdb.original.file.number";
const std::string TablePropertiesNames::kDataSize =
"rocksdb.data.size";
const std::string TablePropertiesNames::kIndexSize =
"rocksdb.index.size";
const std::string TablePropertiesNames::kIndexPartitions =
"rocksdb.index.partitions";
const std::string TablePropertiesNames::kTopLevelIndexSize =
"rocksdb.top-level.index.size";
const std::string TablePropertiesNames::kIndexKeyIsUserKey =
"rocksdb.index.key.is.user.key";
const std::string TablePropertiesNames::kIndexValueIsDeltaEncoded =
"rocksdb.index.value.is.delta.encoded";
const std::string TablePropertiesNames::kFilterSize =
"rocksdb.filter.size";
const std::string TablePropertiesNames::kRawKeySize =
"rocksdb.raw.key.size";
const std::string TablePropertiesNames::kRawValueSize =
"rocksdb.raw.value.size";
const std::string TablePropertiesNames::kNumDataBlocks =
"rocksdb.num.data.blocks";
const std::string TablePropertiesNames::kNumEntries =
"rocksdb.num.entries";
const std::string TablePropertiesNames::kNumFilterEntries =
"rocksdb.num.filter_entries";
const std::string TablePropertiesNames::kDeletedKeys = "rocksdb.deleted.keys";
const std::string TablePropertiesNames::kMergeOperands =
"rocksdb.merge.operands";
const std::string TablePropertiesNames::kNumRangeDeletions =
"rocksdb.num.range-deletions";
const std::string TablePropertiesNames::kFilterPolicy =
"rocksdb.filter.policy";
const std::string TablePropertiesNames::kFormatVersion =
"rocksdb.format.version";
const std::string TablePropertiesNames::kFixedKeyLen =
"rocksdb.fixed.key.length";
const std::string TablePropertiesNames::kColumnFamilyId =
"rocksdb.column.family.id";
const std::string TablePropertiesNames::kColumnFamilyName =
"rocksdb.column.family.name";
const std::string TablePropertiesNames::kComparator = "rocksdb.comparator";
const std::string TablePropertiesNames::kMergeOperator =
"rocksdb.merge.operator";
const std::string TablePropertiesNames::kPrefixExtractorName =
"rocksdb.prefix.extractor.name";
const std::string TablePropertiesNames::kPropertyCollectors =
"rocksdb.property.collectors";
const std::string TablePropertiesNames::kCompression = "rocksdb.compression";
const std::string TablePropertiesNames::kCompressionOptions =
"rocksdb.compression_options";
const std::string TablePropertiesNames::kCreationTime = "rocksdb.creation.time";
const std::string TablePropertiesNames::kOldestKeyTime =
"rocksdb.oldest.key.time";
const std::string TablePropertiesNames::kFileCreationTime =
"rocksdb.file.creation.time";
const std::string TablePropertiesNames::kSlowCompressionEstimatedDataSize =
"rocksdb.sample_for_compression.slow.data.size";
const std::string TablePropertiesNames::kFastCompressionEstimatedDataSize =
"rocksdb.sample_for_compression.fast.data.size";
#ifndef NDEBUG
void TEST_SetRandomTableProperties(TableProperties* props) {
Random* r = Random::GetTLSInstance();
// For now, TableProperties is composed of a number of uint64_t followed by
// a number of std::string, followed by some extras starting with
// user_collected_properties.
uint64_t* pu = &props->orig_file_number;
assert(static_cast<void*>(pu) == static_cast<void*>(props));
std::string* ps = &props->db_id;
const uint64_t* const pu_end = reinterpret_cast<const uint64_t*>(ps);
const std::string* const ps_end =
reinterpret_cast<const std::string*>(&props->user_collected_properties);
for (; pu < pu_end; ++pu) {
*pu = r->Next64();
}
assert(static_cast<void*>(pu) == static_cast<void*>(ps));
for (; ps < ps_end; ++ps) {
*ps = r->RandomBinaryString(13);
}
}
#endif
} // namespace ROCKSDB_NAMESPACE