rocksdb/db/internal_stats.h
Peter Dillinger 9f7801c5f1 Major Cache refactoring, CPU efficiency improvement (#10975)
Summary:
This is several refactorings bundled into one to avoid having to incrementally re-modify uses of Cache several times. Overall, there are breaking changes to Cache class, and it becomes more of low-level interface for implementing caches, especially block cache. New internal APIs make using Cache cleaner than before, and more insulated from block cache evolution. Hopefully, this is the last really big block cache refactoring, because of rather effectively decoupling the implementations from the uses. This change also removes the EXPERIMENTAL designation on the SecondaryCache support in Cache. It seems reasonably mature at this point but still subject to change/evolution (as I warn in the API docs for Cache).

The high-level motivation for this refactoring is to minimize code duplication / compounding complexity in adding SecondaryCache support to HyperClockCache (in a later PR). Other benefits listed below.

* static_cast lines of code +29 -35 (net removed 6)
* reinterpret_cast lines of code +6 -32 (net removed 26)

## cache.h and secondary_cache.h
* Always use CacheItemHelper with entries instead of just a Deleter. There are several motivations / justifications:
  * Simpler for implementations to deal with just one Insert and one Lookup.
  * Simpler and more efficient implementation because we don't have to track which entries are using helpers and which are using deleters
  * Gets rid of hack to classify cache entries by their deleter. Instead, the CacheItemHelper includes a CacheEntryRole. This simplifies a lot of code (cache_entry_roles.h almost eliminated). Fixes https://github.com/facebook/rocksdb/issues/9428.
  * Makes it trivial to adjust SecondaryCache behavior based on kind of block (e.g. don't re-compress filter blocks).
  * It is arguably less convenient for many direct users of Cache, but direct users of Cache are now rare with introduction of typed_cache.h (below).
  * I considered and rejected an alternative approach in which we reduce customizability by assuming each secondary cache compatible value starts with a Slice referencing the uncompressed block contents (already true or mostly true), but we apparently intend to stack secondary caches. Saving an entry from a compressed secondary to a lower tier requires custom handling offered by SaveToCallback, etc.
* Make CreateCallback part of the helper and introduce CreateContext to work with it (alternative to https://github.com/facebook/rocksdb/issues/10562). This cleans up the interface while still allowing context to be provided for loading/parsing values into primary cache. This model works for async lookup in BlockBasedTable reader (reader owns a CreateContext) under the assumption that it always waits on secondary cache operations to finish. (Otherwise, the CreateContext could be destroyed while async operation depending on it continues.) This likely contributes most to the observed performance improvement because it saves an std::function backed by a heap allocation.
* Use char* for serialized data, e.g. in SaveToCallback, where void* was confusingly used. (We use `char*` for serialized byte data all over RocksDB, with many advantages over `void*`. `memcpy` etc. are legacy APIs that should not be mimicked.)
* Add a type alias Cache::ObjectPtr = void*, so that we can better indicate the intent of the void* when it is to be the object associated with a Cache entry. Related: started (but did not complete) a refactoring to move away from "value" of a cache entry toward "object" or "obj". (It is confusing to call Cache a key-value store (like DB) when it is really storing arbitrary in-memory objects, not byte strings.)
* Remove unnecessary key param from DeleterFn. This is good for efficiency in HyperClockCache, which does not directly store the cache key in memory. (Alternative to https://github.com/facebook/rocksdb/issues/10774)
* Add allocator to Cache DeleterFn. This is a kind of future-proofing change in case we get more serious about using the Cache allocator for memory tracked by the Cache. Right now, only the uncompressed block contents are allocated using the allocator, and a pointer to that allocator is saved as part of the cached object so that the deleter can use it. (See CacheAllocationPtr.) If in the future we are able to "flatten out" our Cache objects some more, it would be good not to have to track the allocator as part of each object.
* Removes legacy `ApplyToAllCacheEntries` and changes `ApplyToAllEntries` signature for Deleter->CacheItemHelper change.

## typed_cache.h
Adds various "typed" interfaces to the Cache as internal APIs, so that most uses of Cache can use simple type safe code without casting and without explicit deleters, etc. Almost all of the non-test, non-glue code uses of Cache have been migrated. (Follow-up work: CompressedSecondaryCache deserves deeper attention to migrate.) This change expands RocksDB's internal usage of metaprogramming and SFINAE (https://en.cppreference.com/w/cpp/language/sfinae).

The existing usages of Cache are divided up at a high level into these new interfaces. See updated existing uses of Cache for examples of how these are used.
* PlaceholderCacheInterface - Used for making cache reservations, with entries that have a charge but no value.
* BasicTypedCacheInterface<TValue> - Used for primary cache storage of objects of type TValue, which can be cleaned up with std::default_delete<TValue>. The role is provided by TValue::kCacheEntryRole or given in an optional template parameter.
* FullTypedCacheInterface<TValue, TCreateContext> - Used for secondary cache compatible storage of objects of type TValue. In addition to BasicTypedCacheInterface constraints, we require TValue::ContentSlice() to return persistable data. This simplifies usage for the normal case of simple secondary cache compatibility (can give you a Slice to the data already in memory). In addition to TCreateContext performing the role of Cache::CreateContext, it is also expected to provide a factory function for creating TValue.
* For each of these, there's a "Shared" version (e.g. FullTypedSharedCacheInterface) that holds a shared_ptr to the Cache, rather than assuming external ownership by holding only a raw `Cache*`.

These interfaces introduce specific handle types for each interface instantiation, so that it's easy to see what kind of object is controlled by a handle. (Ultimately, this might not be worth the extra complexity, but it seems OK so far.)

Note: I attempted to make the cache 'charge' automatically inferred from the cache object type, such as by expecting an ApproximateMemoryUsage() function, but this is not so clean because there are cases where we need to compute the charge ahead of time and don't want to re-compute it.

## block_cache.h
This header is essentially the replacement for the old block_like_traits.h. It includes various things to support block cache access with typed_cache.h for block-based table.

## block_based_table_reader.cc
Before this change, accessing the block cache here was an awkward mix of static polymorphism (template TBlocklike) and switch-case on a dynamic BlockType value. This change mostly unifies on static polymorphism, relying on minor hacks in block_cache.h to distinguish variants of Block. We still check BlockType in some places (especially for stats, which could be improved in follow-up work) but at least the BlockType is a static constant from the template parameter. (No more awkward partial redundancy between static and dynamic info.) This likely contributes to the overall performance improvement, but hasn't been tested in isolation.

The other key source of simplification here is a more unified system of creating block cache objects: for directly populating from primary cache and for promotion from secondary cache. Both use BlockCreateContext, for context and for factory functions.

## block_based_table_builder.cc, cache_dump_load_impl.cc
Before this change, warming caches was super ugly code. Both of these source files had switch statements to basically transition from the dynamic BlockType world to the static TBlocklike world. None of that mess is needed anymore as there's a new, untyped WarmInCache function that handles all the details just as promotion from SecondaryCache would. (Fixes `TODO akanksha: Dedup below code` in block_based_table_builder.cc.)

## Everything else
Mostly just updating Cache users to use new typed APIs when reasonably possible, or changed Cache APIs when not.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/10975

Test Plan:
tests updated

Performance test setup similar to https://github.com/facebook/rocksdb/issues/10626 (by cache size, LRUCache when not "hyper" for HyperClockCache):

34MB 1thread base.hyper -> kops/s: 0.745 io_bytes/op: 2.52504e+06 miss_ratio: 0.140906 max_rss_mb: 76.4844
34MB 1thread new.hyper -> kops/s: 0.751 io_bytes/op: 2.5123e+06 miss_ratio: 0.140161 max_rss_mb: 79.3594
34MB 1thread base -> kops/s: 0.254 io_bytes/op: 1.36073e+07 miss_ratio: 0.918818 max_rss_mb: 45.9297
34MB 1thread new -> kops/s: 0.252 io_bytes/op: 1.36157e+07 miss_ratio: 0.918999 max_rss_mb: 44.1523
34MB 32thread base.hyper -> kops/s: 7.272 io_bytes/op: 2.88323e+06 miss_ratio: 0.162532 max_rss_mb: 516.602
34MB 32thread new.hyper -> kops/s: 7.214 io_bytes/op: 2.99046e+06 miss_ratio: 0.168818 max_rss_mb: 518.293
34MB 32thread base -> kops/s: 3.528 io_bytes/op: 1.35722e+07 miss_ratio: 0.914691 max_rss_mb: 264.926
34MB 32thread new -> kops/s: 3.604 io_bytes/op: 1.35744e+07 miss_ratio: 0.915054 max_rss_mb: 264.488
233MB 1thread base.hyper -> kops/s: 53.909 io_bytes/op: 2552.35 miss_ratio: 0.0440566 max_rss_mb: 241.984
233MB 1thread new.hyper -> kops/s: 62.792 io_bytes/op: 2549.79 miss_ratio: 0.044043 max_rss_mb: 241.922
233MB 1thread base -> kops/s: 1.197 io_bytes/op: 2.75173e+06 miss_ratio: 0.103093 max_rss_mb: 241.559
233MB 1thread new -> kops/s: 1.199 io_bytes/op: 2.73723e+06 miss_ratio: 0.10305 max_rss_mb: 240.93
233MB 32thread base.hyper -> kops/s: 1298.69 io_bytes/op: 2539.12 miss_ratio: 0.0440307 max_rss_mb: 371.418
233MB 32thread new.hyper -> kops/s: 1421.35 io_bytes/op: 2538.75 miss_ratio: 0.0440307 max_rss_mb: 347.273
233MB 32thread base -> kops/s: 9.693 io_bytes/op: 2.77304e+06 miss_ratio: 0.103745 max_rss_mb: 569.691
233MB 32thread new -> kops/s: 9.75 io_bytes/op: 2.77559e+06 miss_ratio: 0.103798 max_rss_mb: 552.82
1597MB 1thread base.hyper -> kops/s: 58.607 io_bytes/op: 1449.14 miss_ratio: 0.0249324 max_rss_mb: 1583.55
1597MB 1thread new.hyper -> kops/s: 69.6 io_bytes/op: 1434.89 miss_ratio: 0.0247167 max_rss_mb: 1584.02
1597MB 1thread base -> kops/s: 60.478 io_bytes/op: 1421.28 miss_ratio: 0.024452 max_rss_mb: 1589.45
1597MB 1thread new -> kops/s: 63.973 io_bytes/op: 1416.07 miss_ratio: 0.0243766 max_rss_mb: 1589.24
1597MB 32thread base.hyper -> kops/s: 1436.2 io_bytes/op: 1357.93 miss_ratio: 0.0235353 max_rss_mb: 1692.92
1597MB 32thread new.hyper -> kops/s: 1605.03 io_bytes/op: 1358.04 miss_ratio: 0.023538 max_rss_mb: 1702.78
1597MB 32thread base -> kops/s: 280.059 io_bytes/op: 1350.34 miss_ratio: 0.023289 max_rss_mb: 1675.36
1597MB 32thread new -> kops/s: 283.125 io_bytes/op: 1351.05 miss_ratio: 0.0232797 max_rss_mb: 1703.83

Almost uniformly improving over base revision, especially for hot paths with HyperClockCache, up to 12% higher throughput seen (1597MB, 32thread, hyper). The improvement for that is likely coming from much simplified code for providing context for secondary cache promotion (CreateCallback/CreateContext), and possibly from less branching in block_based_table_reader. And likely a small improvement from not reconstituting key for DeleterFn.

Reviewed By: anand1976

Differential Revision: D42417818

Pulled By: pdillinger

fbshipit-source-id: f86bfdd584dce27c028b151ba56818ad14f7a432
2023-01-11 14:20:40 -08:00

997 lines
36 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).
//
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
//
#pragma once
#include <map>
#include <memory>
#include <string>
#include <vector>
#include "cache/cache_entry_roles.h"
#include "db/version_set.h"
#include "rocksdb/system_clock.h"
#include "util/hash_containers.h"
namespace ROCKSDB_NAMESPACE {
template <class Stats>
class CacheEntryStatsCollector;
class DBImpl;
class MemTableList;
// Config for retrieving a property's value.
struct DBPropertyInfo {
bool need_out_of_mutex;
// gcc had an internal error for initializing union of pointer-to-member-
// functions. Workaround is to populate exactly one of the following function
// pointers with a non-nullptr value.
// @param value Value-result argument for storing the property's string value
// @param suffix Argument portion of the property. For example, suffix would
// be "5" for the property "rocksdb.num-files-at-level5". So far, only
// certain string properties take an argument.
bool (InternalStats::*handle_string)(std::string* value, Slice suffix);
// @param value Value-result argument for storing the property's uint64 value
// @param db Many of the int properties rely on DBImpl methods.
// @param version Version is needed in case the property is retrieved without
// holding db mutex, which is only supported for int properties.
bool (InternalStats::*handle_int)(uint64_t* value, DBImpl* db,
Version* version);
// @param props Map of general properties to populate
// @param suffix Argument portion of the property. (see handle_string)
bool (InternalStats::*handle_map)(std::map<std::string, std::string>* props,
Slice suffix);
// handle the string type properties rely on DBImpl methods
// @param value Value-result argument for storing the property's string value
bool (DBImpl::*handle_string_dbimpl)(std::string* value);
};
extern const DBPropertyInfo* GetPropertyInfo(const Slice& property);
#ifndef ROCKSDB_LITE
#undef SCORE
enum class LevelStatType {
INVALID = 0,
NUM_FILES,
COMPACTED_FILES,
SIZE_BYTES,
SCORE,
READ_GB,
RN_GB,
RNP1_GB,
WRITE_GB,
W_NEW_GB,
MOVED_GB,
WRITE_AMP,
READ_MBPS,
WRITE_MBPS,
COMP_SEC,
COMP_CPU_SEC,
COMP_COUNT,
AVG_SEC,
KEY_IN,
KEY_DROP,
R_BLOB_GB,
W_BLOB_GB,
TOTAL // total number of types
};
struct LevelStat {
// This what will be L?.property_name in the flat map returned to the user
std::string property_name;
// This will be what we will print in the header in the cli
std::string header_name;
};
struct DBStatInfo {
// This what will be property_name in the flat map returned to the user
std::string property_name;
};
class InternalStats {
public:
static const std::map<LevelStatType, LevelStat> compaction_level_stats;
enum InternalCFStatsType {
L0_FILE_COUNT_LIMIT_SLOWDOWNS,
LOCKED_L0_FILE_COUNT_LIMIT_SLOWDOWNS,
MEMTABLE_LIMIT_STOPS,
MEMTABLE_LIMIT_SLOWDOWNS,
L0_FILE_COUNT_LIMIT_STOPS,
LOCKED_L0_FILE_COUNT_LIMIT_STOPS,
PENDING_COMPACTION_BYTES_LIMIT_SLOWDOWNS,
PENDING_COMPACTION_BYTES_LIMIT_STOPS,
WRITE_STALLS_ENUM_MAX,
BYTES_FLUSHED,
BYTES_INGESTED_ADD_FILE,
INGESTED_NUM_FILES_TOTAL,
INGESTED_LEVEL0_NUM_FILES_TOTAL,
INGESTED_NUM_KEYS_TOTAL,
INTERNAL_CF_STATS_ENUM_MAX,
};
enum InternalDBStatsType {
kIntStatsWalFileBytes,
kIntStatsWalFileSynced,
kIntStatsBytesWritten,
kIntStatsNumKeysWritten,
kIntStatsWriteDoneByOther,
kIntStatsWriteDoneBySelf,
kIntStatsWriteWithWal,
kIntStatsWriteStallMicros,
kIntStatsNumMax,
};
static const std::map<InternalDBStatsType, DBStatInfo> db_stats_type_to_info;
InternalStats(int num_levels, SystemClock* clock, ColumnFamilyData* cfd);
// Per level compaction stats
struct CompactionOutputsStats {
uint64_t num_output_records = 0;
uint64_t bytes_written = 0;
uint64_t bytes_written_blob = 0;
uint64_t num_output_files = 0;
uint64_t num_output_files_blob = 0;
void Add(const CompactionOutputsStats& stats) {
this->num_output_records += stats.num_output_records;
this->bytes_written += stats.bytes_written;
this->bytes_written_blob += stats.bytes_written_blob;
this->num_output_files += stats.num_output_files;
this->num_output_files_blob += stats.num_output_files_blob;
}
};
// Per level compaction stats. comp_stats_[level] stores the stats for
// compactions that produced data for the specified "level".
struct CompactionStats {
uint64_t micros;
uint64_t cpu_micros;
// The number of bytes read from all non-output levels (table files)
uint64_t bytes_read_non_output_levels;
// The number of bytes read from the compaction output level (table files)
uint64_t bytes_read_output_level;
// The number of bytes read from blob files
uint64_t bytes_read_blob;
// Total number of bytes written to table files during compaction
uint64_t bytes_written;
// Total number of bytes written to blob files during compaction
uint64_t bytes_written_blob;
// Total number of bytes moved to the output level (table files)
uint64_t bytes_moved;
// The number of compaction input files in all non-output levels (table
// files)
int num_input_files_in_non_output_levels;
// The number of compaction input files in the output level (table files)
int num_input_files_in_output_level;
// The number of compaction output files (table files)
int num_output_files;
// The number of compaction output files (blob files)
int num_output_files_blob;
// Total incoming entries during compaction between levels N and N+1
uint64_t num_input_records;
// Accumulated diff number of entries
// (num input entries - num output entries) for compaction levels N and N+1
uint64_t num_dropped_records;
// Total output entries from compaction
uint64_t num_output_records;
// Number of compactions done
int count;
// Number of compactions done per CompactionReason
int counts[static_cast<int>(CompactionReason::kNumOfReasons)]{};
explicit CompactionStats()
: micros(0),
cpu_micros(0),
bytes_read_non_output_levels(0),
bytes_read_output_level(0),
bytes_read_blob(0),
bytes_written(0),
bytes_written_blob(0),
bytes_moved(0),
num_input_files_in_non_output_levels(0),
num_input_files_in_output_level(0),
num_output_files(0),
num_output_files_blob(0),
num_input_records(0),
num_dropped_records(0),
num_output_records(0),
count(0) {
int num_of_reasons = static_cast<int>(CompactionReason::kNumOfReasons);
for (int i = 0; i < num_of_reasons; i++) {
counts[i] = 0;
}
}
explicit CompactionStats(CompactionReason reason, int c)
: micros(0),
cpu_micros(0),
bytes_read_non_output_levels(0),
bytes_read_output_level(0),
bytes_read_blob(0),
bytes_written(0),
bytes_written_blob(0),
bytes_moved(0),
num_input_files_in_non_output_levels(0),
num_input_files_in_output_level(0),
num_output_files(0),
num_output_files_blob(0),
num_input_records(0),
num_dropped_records(0),
num_output_records(0),
count(c) {
int num_of_reasons = static_cast<int>(CompactionReason::kNumOfReasons);
for (int i = 0; i < num_of_reasons; i++) {
counts[i] = 0;
}
int r = static_cast<int>(reason);
if (r >= 0 && r < num_of_reasons) {
counts[r] = c;
} else {
count = 0;
}
}
CompactionStats(const CompactionStats& c)
: micros(c.micros),
cpu_micros(c.cpu_micros),
bytes_read_non_output_levels(c.bytes_read_non_output_levels),
bytes_read_output_level(c.bytes_read_output_level),
bytes_read_blob(c.bytes_read_blob),
bytes_written(c.bytes_written),
bytes_written_blob(c.bytes_written_blob),
bytes_moved(c.bytes_moved),
num_input_files_in_non_output_levels(
c.num_input_files_in_non_output_levels),
num_input_files_in_output_level(c.num_input_files_in_output_level),
num_output_files(c.num_output_files),
num_output_files_blob(c.num_output_files_blob),
num_input_records(c.num_input_records),
num_dropped_records(c.num_dropped_records),
num_output_records(c.num_output_records),
count(c.count) {
int num_of_reasons = static_cast<int>(CompactionReason::kNumOfReasons);
for (int i = 0; i < num_of_reasons; i++) {
counts[i] = c.counts[i];
}
}
CompactionStats& operator=(const CompactionStats& c) {
micros = c.micros;
cpu_micros = c.cpu_micros;
bytes_read_non_output_levels = c.bytes_read_non_output_levels;
bytes_read_output_level = c.bytes_read_output_level;
bytes_read_blob = c.bytes_read_blob;
bytes_written = c.bytes_written;
bytes_written_blob = c.bytes_written_blob;
bytes_moved = c.bytes_moved;
num_input_files_in_non_output_levels =
c.num_input_files_in_non_output_levels;
num_input_files_in_output_level = c.num_input_files_in_output_level;
num_output_files = c.num_output_files;
num_output_files_blob = c.num_output_files_blob;
num_input_records = c.num_input_records;
num_dropped_records = c.num_dropped_records;
num_output_records = c.num_output_records;
count = c.count;
int num_of_reasons = static_cast<int>(CompactionReason::kNumOfReasons);
for (int i = 0; i < num_of_reasons; i++) {
counts[i] = c.counts[i];
}
return *this;
}
void Clear() {
this->micros = 0;
this->cpu_micros = 0;
this->bytes_read_non_output_levels = 0;
this->bytes_read_output_level = 0;
this->bytes_read_blob = 0;
this->bytes_written = 0;
this->bytes_written_blob = 0;
this->bytes_moved = 0;
this->num_input_files_in_non_output_levels = 0;
this->num_input_files_in_output_level = 0;
this->num_output_files = 0;
this->num_output_files_blob = 0;
this->num_input_records = 0;
this->num_dropped_records = 0;
this->num_output_records = 0;
this->count = 0;
int num_of_reasons = static_cast<int>(CompactionReason::kNumOfReasons);
for (int i = 0; i < num_of_reasons; i++) {
counts[i] = 0;
}
}
void Add(const CompactionStats& c) {
this->micros += c.micros;
this->cpu_micros += c.cpu_micros;
this->bytes_read_non_output_levels += c.bytes_read_non_output_levels;
this->bytes_read_output_level += c.bytes_read_output_level;
this->bytes_read_blob += c.bytes_read_blob;
this->bytes_written += c.bytes_written;
this->bytes_written_blob += c.bytes_written_blob;
this->bytes_moved += c.bytes_moved;
this->num_input_files_in_non_output_levels +=
c.num_input_files_in_non_output_levels;
this->num_input_files_in_output_level +=
c.num_input_files_in_output_level;
this->num_output_files += c.num_output_files;
this->num_output_files_blob += c.num_output_files_blob;
this->num_input_records += c.num_input_records;
this->num_dropped_records += c.num_dropped_records;
this->num_output_records += c.num_output_records;
this->count += c.count;
int num_of_reasons = static_cast<int>(CompactionReason::kNumOfReasons);
for (int i = 0; i < num_of_reasons; i++) {
counts[i] += c.counts[i];
}
}
void Add(const CompactionOutputsStats& stats) {
this->num_output_files += static_cast<int>(stats.num_output_files);
this->num_output_records += stats.num_output_records;
this->bytes_written += stats.bytes_written;
this->bytes_written_blob += stats.bytes_written_blob;
this->num_output_files_blob +=
static_cast<int>(stats.num_output_files_blob);
}
void Subtract(const CompactionStats& c) {
this->micros -= c.micros;
this->cpu_micros -= c.cpu_micros;
this->bytes_read_non_output_levels -= c.bytes_read_non_output_levels;
this->bytes_read_output_level -= c.bytes_read_output_level;
this->bytes_read_blob -= c.bytes_read_blob;
this->bytes_written -= c.bytes_written;
this->bytes_written_blob -= c.bytes_written_blob;
this->bytes_moved -= c.bytes_moved;
this->num_input_files_in_non_output_levels -=
c.num_input_files_in_non_output_levels;
this->num_input_files_in_output_level -=
c.num_input_files_in_output_level;
this->num_output_files -= c.num_output_files;
this->num_output_files_blob -= c.num_output_files_blob;
this->num_input_records -= c.num_input_records;
this->num_dropped_records -= c.num_dropped_records;
this->num_output_records -= c.num_output_records;
this->count -= c.count;
int num_of_reasons = static_cast<int>(CompactionReason::kNumOfReasons);
for (int i = 0; i < num_of_reasons; i++) {
counts[i] -= c.counts[i];
}
}
void ResetCompactionReason(CompactionReason reason) {
int num_of_reasons = static_cast<int>(CompactionReason::kNumOfReasons);
assert(count == 1); // only support update one compaction reason
for (int i = 0; i < num_of_reasons; i++) {
counts[i] = 0;
}
int r = static_cast<int>(reason);
assert(r >= 0 && r < num_of_reasons);
counts[r] = 1;
}
};
// Compaction stats, for per_key_placement compaction, it includes 2 levels
// stats: the last level and the penultimate level.
struct CompactionStatsFull {
// the stats for the target primary output level
CompactionStats stats;
// stats for penultimate level output if exist
bool has_penultimate_level_output = false;
CompactionStats penultimate_level_stats;
explicit CompactionStatsFull() : stats(), penultimate_level_stats() {}
explicit CompactionStatsFull(CompactionReason reason, int c)
: stats(reason, c), penultimate_level_stats(reason, c){};
uint64_t TotalBytesWritten() const {
uint64_t bytes_written = stats.bytes_written + stats.bytes_written_blob;
if (has_penultimate_level_output) {
bytes_written += penultimate_level_stats.bytes_written +
penultimate_level_stats.bytes_written_blob;
}
return bytes_written;
}
uint64_t DroppedRecords() {
uint64_t output_records = stats.num_output_records;
if (has_penultimate_level_output) {
output_records += penultimate_level_stats.num_output_records;
}
if (stats.num_input_records > output_records) {
return stats.num_input_records - output_records;
}
return 0;
}
void SetMicros(uint64_t val) {
stats.micros = val;
penultimate_level_stats.micros = val;
}
void AddCpuMicros(uint64_t val) {
stats.cpu_micros += val;
penultimate_level_stats.cpu_micros += val;
}
};
// For use with CacheEntryStatsCollector
struct CacheEntryRoleStats {
uint64_t cache_capacity = 0;
uint64_t cache_usage = 0;
size_t table_size = 0;
size_t occupancy = 0;
std::string cache_id;
std::array<uint64_t, kNumCacheEntryRoles> total_charges;
std::array<size_t, kNumCacheEntryRoles> entry_counts;
uint32_t collection_count = 0;
uint32_t copies_of_last_collection = 0;
uint64_t last_start_time_micros_ = 0;
uint64_t last_end_time_micros_ = 0;
void Clear() {
// Wipe everything except collection_count
uint32_t saved_collection_count = collection_count;
*this = CacheEntryRoleStats();
collection_count = saved_collection_count;
}
void BeginCollection(Cache*, SystemClock*, uint64_t start_time_micros);
std::function<void(const Slice& key, Cache::ObjectPtr value, size_t charge,
const Cache::CacheItemHelper* helper)>
GetEntryCallback();
void EndCollection(Cache*, SystemClock*, uint64_t end_time_micros);
void SkippedCollection();
std::string ToString(SystemClock* clock) const;
void ToMap(std::map<std::string, std::string>* values,
SystemClock* clock) const;
private:
uint64_t GetLastDurationMicros() const;
};
void Clear() {
for (int i = 0; i < kIntStatsNumMax; i++) {
db_stats_[i].store(0);
}
for (int i = 0; i < INTERNAL_CF_STATS_ENUM_MAX; i++) {
cf_stats_count_[i] = 0;
cf_stats_value_[i] = 0;
}
for (auto& comp_stat : comp_stats_) {
comp_stat.Clear();
}
per_key_placement_comp_stats_.Clear();
for (auto& h : file_read_latency_) {
h.Clear();
}
blob_file_read_latency_.Clear();
cf_stats_snapshot_.Clear();
db_stats_snapshot_.Clear();
bg_error_count_ = 0;
started_at_ = clock_->NowMicros();
has_cf_change_since_dump_ = true;
}
void AddCompactionStats(int level, Env::Priority thread_pri,
const CompactionStats& stats) {
comp_stats_[level].Add(stats);
comp_stats_by_pri_[thread_pri].Add(stats);
}
void AddCompactionStats(int level, Env::Priority thread_pri,
const CompactionStatsFull& comp_stats_full) {
AddCompactionStats(level, thread_pri, comp_stats_full.stats);
if (comp_stats_full.has_penultimate_level_output) {
per_key_placement_comp_stats_.Add(
comp_stats_full.penultimate_level_stats);
}
}
void IncBytesMoved(int level, uint64_t amount) {
comp_stats_[level].bytes_moved += amount;
}
void AddCFStats(InternalCFStatsType type, uint64_t value) {
has_cf_change_since_dump_ = true;
cf_stats_value_[type] += value;
++cf_stats_count_[type];
}
void AddDBStats(InternalDBStatsType type, uint64_t value,
bool concurrent = false) {
auto& v = db_stats_[type];
if (concurrent) {
v.fetch_add(value, std::memory_order_relaxed);
} else {
v.store(v.load(std::memory_order_relaxed) + value,
std::memory_order_relaxed);
}
}
uint64_t GetDBStats(InternalDBStatsType type) {
return db_stats_[type].load(std::memory_order_relaxed);
}
HistogramImpl* GetFileReadHist(int level) {
return &file_read_latency_[level];
}
HistogramImpl* GetBlobFileReadHist() { return &blob_file_read_latency_; }
uint64_t GetBackgroundErrorCount() const { return bg_error_count_; }
uint64_t BumpAndGetBackgroundErrorCount() { return ++bg_error_count_; }
bool GetStringProperty(const DBPropertyInfo& property_info,
const Slice& property, std::string* value);
bool GetMapProperty(const DBPropertyInfo& property_info,
const Slice& property,
std::map<std::string, std::string>* value);
bool GetIntProperty(const DBPropertyInfo& property_info, uint64_t* value,
DBImpl* db);
bool GetIntPropertyOutOfMutex(const DBPropertyInfo& property_info,
Version* version, uint64_t* value);
// Unless there is a recent enough collection of the stats, collect and
// saved new cache entry stats. If `foreground`, require data to be more
// recent to skip re-collection.
//
// This should only be called while NOT holding the DB mutex.
void CollectCacheEntryStats(bool foreground);
const uint64_t* TEST_GetCFStatsValue() const { return cf_stats_value_; }
const std::vector<CompactionStats>& TEST_GetCompactionStats() const {
return comp_stats_;
}
const CompactionStats& TEST_GetPerKeyPlacementCompactionStats() const {
return per_key_placement_comp_stats_;
}
void TEST_GetCacheEntryRoleStats(CacheEntryRoleStats* stats, bool foreground);
// Store a mapping from the user-facing DB::Properties string to our
// DBPropertyInfo struct used internally for retrieving properties.
static const UnorderedMap<std::string, DBPropertyInfo> ppt_name_to_info;
static const std::string kPeriodicCFStats;
private:
void DumpDBMapStats(std::map<std::string, std::string>* db_stats);
void DumpDBStats(std::string* value);
void DumpCFMapStats(std::map<std::string, std::string>* cf_stats);
void DumpCFMapStats(
const VersionStorageInfo* vstorage,
std::map<int, std::map<LevelStatType, double>>* level_stats,
CompactionStats* compaction_stats_sum);
void DumpCFMapStatsByPriority(
std::map<int, std::map<LevelStatType, double>>* priorities_stats);
void DumpCFMapStatsIOStalls(std::map<std::string, std::string>* cf_stats);
void DumpCFStats(std::string* value);
// if is_periodic = true, it is an internal call by RocksDB periodically to
// dump the status.
void DumpCFStatsNoFileHistogram(bool is_periodic, std::string* value);
// if is_periodic = true, it is an internal call by RocksDB periodically to
// dump the status.
void DumpCFFileHistogram(std::string* value);
Cache* GetBlockCacheForStats();
Cache* GetBlobCacheForStats();
// Per-DB stats
std::atomic<uint64_t> db_stats_[kIntStatsNumMax];
// Per-ColumnFamily stats
uint64_t cf_stats_value_[INTERNAL_CF_STATS_ENUM_MAX];
uint64_t cf_stats_count_[INTERNAL_CF_STATS_ENUM_MAX];
// Initialize/reference the collector in constructor so that we don't need
// additional synchronization in InternalStats, relying on synchronization
// in CacheEntryStatsCollector::GetStats. This collector is pinned in cache
// (through a shared_ptr) so that it does not get immediately ejected from
// a full cache, which would force a re-scan on the next GetStats.
std::shared_ptr<CacheEntryStatsCollector<CacheEntryRoleStats>>
cache_entry_stats_collector_;
// Per-ColumnFamily/level compaction stats
std::vector<CompactionStats> comp_stats_;
std::vector<CompactionStats> comp_stats_by_pri_;
CompactionStats per_key_placement_comp_stats_;
std::vector<HistogramImpl> file_read_latency_;
HistogramImpl blob_file_read_latency_;
bool has_cf_change_since_dump_;
// How many periods of no change since the last time stats are dumped for
// a periodic dump.
int no_cf_change_period_since_dump_ = 0;
uint64_t last_histogram_num = std::numeric_limits<uint64_t>::max();
static const int kMaxNoChangePeriodSinceDump;
// Used to compute per-interval statistics
struct CFStatsSnapshot {
// ColumnFamily-level stats
CompactionStats comp_stats;
uint64_t ingest_bytes_flush; // Bytes written to L0 (Flush)
uint64_t stall_count; // Stall count
// Stats from compaction jobs - bytes written, bytes read, duration.
uint64_t compact_bytes_write;
uint64_t compact_bytes_read;
uint64_t compact_micros;
double seconds_up;
// AddFile specific stats
uint64_t ingest_bytes_addfile; // Total Bytes ingested
uint64_t ingest_files_addfile; // Total number of files ingested
uint64_t ingest_l0_files_addfile; // Total number of files ingested to L0
uint64_t ingest_keys_addfile; // Total number of keys ingested
CFStatsSnapshot()
: ingest_bytes_flush(0),
stall_count(0),
compact_bytes_write(0),
compact_bytes_read(0),
compact_micros(0),
seconds_up(0),
ingest_bytes_addfile(0),
ingest_files_addfile(0),
ingest_l0_files_addfile(0),
ingest_keys_addfile(0) {}
void Clear() {
comp_stats.Clear();
ingest_bytes_flush = 0;
stall_count = 0;
compact_bytes_write = 0;
compact_bytes_read = 0;
compact_micros = 0;
seconds_up = 0;
ingest_bytes_addfile = 0;
ingest_files_addfile = 0;
ingest_l0_files_addfile = 0;
ingest_keys_addfile = 0;
}
} cf_stats_snapshot_;
struct DBStatsSnapshot {
// DB-level stats
uint64_t ingest_bytes; // Bytes written by user
uint64_t wal_bytes; // Bytes written to WAL
uint64_t wal_synced; // Number of times WAL is synced
uint64_t write_with_wal; // Number of writes that request WAL
// These count the number of writes processed by the calling thread or
// another thread.
uint64_t write_other;
uint64_t write_self;
// Total number of keys written. write_self and write_other measure number
// of write requests written, Each of the write request can contain updates
// to multiple keys. num_keys_written is total number of keys updated by all
// those writes.
uint64_t num_keys_written;
// Total time writes delayed by stalls.
uint64_t write_stall_micros;
double seconds_up;
DBStatsSnapshot()
: ingest_bytes(0),
wal_bytes(0),
wal_synced(0),
write_with_wal(0),
write_other(0),
write_self(0),
num_keys_written(0),
write_stall_micros(0),
seconds_up(0) {}
void Clear() {
ingest_bytes = 0;
wal_bytes = 0;
wal_synced = 0;
write_with_wal = 0;
write_other = 0;
write_self = 0;
num_keys_written = 0;
write_stall_micros = 0;
seconds_up = 0;
}
} db_stats_snapshot_;
// Handler functions for getting property values. They use "value" as a value-
// result argument, and return true upon successfully setting "value".
bool HandleNumFilesAtLevel(std::string* value, Slice suffix);
bool HandleCompressionRatioAtLevelPrefix(std::string* value, Slice suffix);
bool HandleLevelStats(std::string* value, Slice suffix);
bool HandleStats(std::string* value, Slice suffix);
bool HandleCFMapStats(std::map<std::string, std::string>* compaction_stats,
Slice suffix);
bool HandleCFStats(std::string* value, Slice suffix);
bool HandleCFStatsNoFileHistogram(std::string* value, Slice suffix);
bool HandleCFFileHistogram(std::string* value, Slice suffix);
bool HandleCFStatsPeriodic(std::string* value, Slice suffix);
bool HandleDBMapStats(std::map<std::string, std::string>* compaction_stats,
Slice suffix);
bool HandleDBStats(std::string* value, Slice suffix);
bool HandleSsTables(std::string* value, Slice suffix);
bool HandleAggregatedTableProperties(std::string* value, Slice suffix);
bool HandleAggregatedTablePropertiesAtLevel(std::string* value, Slice suffix);
bool HandleAggregatedTablePropertiesMap(
std::map<std::string, std::string>* values, Slice suffix);
bool HandleAggregatedTablePropertiesAtLevelMap(
std::map<std::string, std::string>* values, Slice suffix);
bool HandleNumImmutableMemTable(uint64_t* value, DBImpl* db,
Version* version);
bool HandleNumImmutableMemTableFlushed(uint64_t* value, DBImpl* db,
Version* version);
bool HandleMemTableFlushPending(uint64_t* value, DBImpl* db,
Version* version);
bool HandleNumRunningFlushes(uint64_t* value, DBImpl* db, Version* version);
bool HandleCompactionPending(uint64_t* value, DBImpl* db, Version* version);
bool HandleNumRunningCompactions(uint64_t* value, DBImpl* db,
Version* version);
bool HandleBackgroundErrors(uint64_t* value, DBImpl* db, Version* version);
bool HandleCurSizeActiveMemTable(uint64_t* value, DBImpl* db,
Version* version);
bool HandleCurSizeAllMemTables(uint64_t* value, DBImpl* db, Version* version);
bool HandleSizeAllMemTables(uint64_t* value, DBImpl* db, Version* version);
bool HandleNumEntriesActiveMemTable(uint64_t* value, DBImpl* db,
Version* version);
bool HandleNumEntriesImmMemTables(uint64_t* value, DBImpl* db,
Version* version);
bool HandleNumDeletesActiveMemTable(uint64_t* value, DBImpl* db,
Version* version);
bool HandleNumDeletesImmMemTables(uint64_t* value, DBImpl* db,
Version* version);
bool HandleEstimateNumKeys(uint64_t* value, DBImpl* db, Version* version);
bool HandleNumSnapshots(uint64_t* value, DBImpl* db, Version* version);
bool HandleOldestSnapshotTime(uint64_t* value, DBImpl* db, Version* version);
bool HandleOldestSnapshotSequence(uint64_t* value, DBImpl* db,
Version* version);
bool HandleNumLiveVersions(uint64_t* value, DBImpl* db, Version* version);
bool HandleCurrentSuperVersionNumber(uint64_t* value, DBImpl* db,
Version* version);
bool HandleIsFileDeletionsEnabled(uint64_t* value, DBImpl* db,
Version* version);
bool HandleBaseLevel(uint64_t* value, DBImpl* db, Version* version);
bool HandleTotalSstFilesSize(uint64_t* value, DBImpl* db, Version* version);
bool HandleLiveSstFilesSize(uint64_t* value, DBImpl* db, Version* version);
bool HandleEstimatePendingCompactionBytes(uint64_t* value, DBImpl* db,
Version* version);
bool HandleEstimateTableReadersMem(uint64_t* value, DBImpl* db,
Version* version);
bool HandleEstimateLiveDataSize(uint64_t* value, DBImpl* db,
Version* version);
bool HandleMinLogNumberToKeep(uint64_t* value, DBImpl* db, Version* version);
bool HandleMinObsoleteSstNumberToKeep(uint64_t* value, DBImpl* db,
Version* version);
bool HandleActualDelayedWriteRate(uint64_t* value, DBImpl* db,
Version* version);
bool HandleIsWriteStopped(uint64_t* value, DBImpl* db, Version* version);
bool HandleEstimateOldestKeyTime(uint64_t* value, DBImpl* db,
Version* version);
bool HandleBlockCacheCapacity(uint64_t* value, DBImpl* db, Version* version);
bool HandleBlockCacheUsage(uint64_t* value, DBImpl* db, Version* version);
bool HandleBlockCachePinnedUsage(uint64_t* value, DBImpl* db,
Version* version);
bool HandleBlockCacheEntryStatsInternal(std::string* value, bool fast);
bool HandleBlockCacheEntryStatsMapInternal(
std::map<std::string, std::string>* values, bool fast);
bool HandleBlockCacheEntryStats(std::string* value, Slice suffix);
bool HandleBlockCacheEntryStatsMap(std::map<std::string, std::string>* values,
Slice suffix);
bool HandleFastBlockCacheEntryStats(std::string* value, Slice suffix);
bool HandleFastBlockCacheEntryStatsMap(
std::map<std::string, std::string>* values, Slice suffix);
bool HandleLiveSstFilesSizeAtTemperature(std::string* value, Slice suffix);
bool HandleNumBlobFiles(uint64_t* value, DBImpl* db, Version* version);
bool HandleBlobStats(std::string* value, Slice suffix);
bool HandleTotalBlobFileSize(uint64_t* value, DBImpl* db, Version* version);
bool HandleLiveBlobFileSize(uint64_t* value, DBImpl* db, Version* version);
bool HandleLiveBlobFileGarbageSize(uint64_t* value, DBImpl* db,
Version* version);
bool HandleBlobCacheCapacity(uint64_t* value, DBImpl* db, Version* version);
bool HandleBlobCacheUsage(uint64_t* value, DBImpl* db, Version* version);
bool HandleBlobCachePinnedUsage(uint64_t* value, DBImpl* db,
Version* version);
// Total number of background errors encountered. Every time a flush task
// or compaction task fails, this counter is incremented. The failure can
// be caused by any possible reason, including file system errors, out of
// resources, or input file corruption. Failing when retrying the same flush
// or compaction will cause the counter to increase too.
uint64_t bg_error_count_;
const int number_levels_;
SystemClock* clock_;
ColumnFamilyData* cfd_;
uint64_t started_at_;
};
#else
class InternalStats {
public:
enum InternalCFStatsType {
L0_FILE_COUNT_LIMIT_SLOWDOWNS,
LOCKED_L0_FILE_COUNT_LIMIT_SLOWDOWNS,
MEMTABLE_LIMIT_STOPS,
MEMTABLE_LIMIT_SLOWDOWNS,
L0_FILE_COUNT_LIMIT_STOPS,
LOCKED_L0_FILE_COUNT_LIMIT_STOPS,
PENDING_COMPACTION_BYTES_LIMIT_SLOWDOWNS,
PENDING_COMPACTION_BYTES_LIMIT_STOPS,
WRITE_STALLS_ENUM_MAX,
BYTES_FLUSHED,
BYTES_INGESTED_ADD_FILE,
INGESTED_NUM_FILES_TOTAL,
INGESTED_LEVEL0_NUM_FILES_TOTAL,
INGESTED_NUM_KEYS_TOTAL,
INTERNAL_CF_STATS_ENUM_MAX,
};
enum InternalDBStatsType {
kIntStatsWalFileBytes,
kIntStatsWalFileSynced,
kIntStatsBytesWritten,
kIntStatsNumKeysWritten,
kIntStatsWriteDoneByOther,
kIntStatsWriteDoneBySelf,
kIntStatsWriteWithWal,
kIntStatsWriteStallMicros,
kIntStatsNumMax,
};
InternalStats(int /*num_levels*/, SystemClock* /*clock*/,
ColumnFamilyData* /*cfd*/) {}
// Per level compaction stats
struct CompactionOutputsStats {
uint64_t num_output_records = 0;
uint64_t bytes_written = 0;
uint64_t bytes_written_blob = 0;
uint64_t num_output_files = 0;
uint64_t num_output_files_blob = 0;
void Add(const CompactionOutputsStats& stats) {
this->num_output_records += stats.num_output_records;
this->bytes_written += stats.bytes_written;
this->bytes_written_blob += stats.bytes_written_blob;
this->num_output_files += stats.num_output_files;
this->num_output_files_blob += stats.num_output_files_blob;
}
};
struct CompactionStats {
uint64_t micros;
uint64_t cpu_micros;
uint64_t bytes_read_non_output_levels;
uint64_t bytes_read_output_level;
uint64_t bytes_read_blob;
uint64_t bytes_written;
uint64_t bytes_written_blob;
uint64_t bytes_moved;
int num_input_files_in_non_output_levels;
int num_input_files_in_output_level;
int num_output_files;
int num_output_files_blob;
uint64_t num_input_records;
uint64_t num_dropped_records;
uint64_t num_output_records;
int count;
explicit CompactionStats() {}
explicit CompactionStats(CompactionReason /*reason*/, int /*c*/) {}
explicit CompactionStats(const CompactionStats& /*c*/) {}
void Add(const CompactionStats& /*c*/) {}
void Add(const CompactionOutputsStats& /*c*/) {}
void Subtract(const CompactionStats& /*c*/) {}
};
struct CompactionStatsFull {
// the stats for the target primary output level (per level stats)
CompactionStats stats;
// stats for output_to_penultimate_level level (per level stats)
bool has_penultimate_level_output = false;
CompactionStats penultimate_level_stats;
explicit CompactionStatsFull(){};
explicit CompactionStatsFull(CompactionReason /*reason*/, int /*c*/){};
uint64_t TotalBytesWritten() const { return 0; }
uint64_t DroppedRecords() { return 0; }
void SetMicros(uint64_t /*val*/){};
void AddCpuMicros(uint64_t /*val*/){};
};
void AddCompactionStats(int /*level*/, Env::Priority /*thread_pri*/,
const CompactionStats& /*stats*/) {}
void AddCompactionStats(int /*level*/, Env::Priority /*thread_pri*/,
const CompactionStatsFull& /*unmerged_stats*/) {}
void IncBytesMoved(int /*level*/, uint64_t /*amount*/) {}
void AddCFStats(InternalCFStatsType /*type*/, uint64_t /*value*/) {}
void AddDBStats(InternalDBStatsType /*type*/, uint64_t /*value*/,
bool /*concurrent */ = false) {}
HistogramImpl* GetFileReadHist(int /*level*/) { return nullptr; }
HistogramImpl* GetBlobFileReadHist() { return nullptr; }
uint64_t GetBackgroundErrorCount() const { return 0; }
uint64_t BumpAndGetBackgroundErrorCount() { return 0; }
bool GetStringProperty(const DBPropertyInfo& /*property_info*/,
const Slice& /*property*/, std::string* /*value*/) {
return false;
}
bool GetMapProperty(const DBPropertyInfo& /*property_info*/,
const Slice& /*property*/,
std::map<std::string, std::string>* /*value*/) {
return false;
}
bool GetIntProperty(const DBPropertyInfo& /*property_info*/,
uint64_t* /*value*/, DBImpl* /*db*/) const {
return false;
}
bool GetIntPropertyOutOfMutex(const DBPropertyInfo& /*property_info*/,
Version* /*version*/,
uint64_t* /*value*/) const {
return false;
}
};
#endif // !ROCKSDB_LITE
} // namespace ROCKSDB_NAMESPACE