@nxtedition/rocksdb 5.2.21 → 5.2.26

package/deps/rocksdb/rocksdb/include/rocksdb/table.h

@@ -44,11 +44,15 @@ class WritableFileWriter;
 struct ConfigOptions;
 struct EnvOptions;
 
+// Types of checksums to use for checking integrity of logical blocks within
+// files. All checksums currently use 32 bits of checking power (1 in 4B
+// chance of failing to detect random corruption).
 enum ChecksumType : char {
   kNoChecksum = 0x0,
   kCRC32c = 0x1,
   kxxHash = 0x2,
   kxxHash64 = 0x3,
+  kXXH3 = 0x4,  // Supported since RocksDB 6.27
 };
 
 // `PinningTier` is used to specify which tier of block-based tables should
@@ -117,9 +121,10 @@ struct BlockBasedTableOptions {
   // caching as they should now apply to range tombstone and compression
   // dictionary meta-blocks, in addition to index and filter meta-blocks.
   //
-  // Indicating if we'd put index/filter blocks to the block cache.
-  // If not specified, each "table reader" object will pre-load index/filter
-  // block during table initialization.
+  // Whether to put index/filter blocks in the block cache. When false,
+  // each "table reader" object will pre-load index/filter blocks during
+  // table initialization. Index and filter partition blocks always use
+  // block cache regardless of this option.
   bool cache_index_and_filter_blocks = false;
 
   // If cache_index_and_filter_blocks is enabled, cache index and filter
@@ -190,6 +195,8 @@ struct BlockBasedTableOptions {
     kHashSearch = 0x01,
 
     // A two-level index implementation. Both levels are binary search indexes.
+    // Second level index blocks ("partitions") use block cache even when
+    // cache_index_and_filter_blocks=false.
     kTwoLevelIndexSearch = 0x02,
 
     // Like kBinarySearch, but index also contains first key of each block.
@@ -220,9 +227,8 @@ struct BlockBasedTableOptions {
   // kDataBlockBinaryAndHash.
   double data_block_hash_table_util_ratio = 0.75;
 
-  // This option is now deprecated. No matter what value it is set to,
-  // it will behave as if hash_index_allow_collision=true.
-  bool hash_index_allow_collision = true;
+  // Option hash_index_allow_collision is now deleted.
+  // It will behave as if hash_index_allow_collision=true.
 
   // Use the specified checksum type. Newly created table files will be
   // protected with this checksum type. Old table files will still be readable,
@@ -252,7 +258,7 @@ struct BlockBasedTableOptions {
   // block size specified here corresponds to uncompressed data.  The
   // actual size of the unit read from disk may be smaller if
   // compression is enabled.  This parameter can be changed dynamically.
-  size_t block_size = 4 * 1024;
+  uint64_t block_size = 4 * 1024;
 
   // This is used to close a block before it reaches the configured
   // 'block_size'. If the percentage of free space in the current block is less
@@ -281,20 +287,63 @@ struct BlockBasedTableOptions {
   // separately
   uint64_t metadata_block_size = 4096;
 
+  // If true, a dynamically updating charge to block cache, loosely based
+  // on the actual memory usage of table building, will occur to account
+  // the memory, if block cache available.
+  //
+  // Charged memory usage includes:
+  // 1. Bloom Filter (format_version >= 5) and Ribbon Filter construction
+  // 2. More to come...
+  //
+  // Note:
+  // 1. Bloom Filter (format_version >= 5) and Ribbon Filter construction
+  //
+  // If additional temporary memory of Ribbon Filter uses up too much memory
+  // relative to the avaible space left in the block cache
+  // at some point (i.e, causing a cache full under
+  // LRUCacheOptions::strict_capacity_limit = true), construction will fall back
+  // to Bloom Filter.
+  //
+  // Default: false
+  bool reserve_table_builder_memory = false;
+
+  // If true, a dynamically updating charge to block cache, loosely based
+  // on the actual memory usage of table reader, will occur to account
+  // the memory, if block cache available.
+  //
+  // Charged memory usage includes:
+  // 1. Table properties
+  // 2. Index block/Filter block/Uncompression dictionary if stored in table
+  // reader (i.e, BlockBasedTableOptions::cache_index_and_filter_blocks ==
+  // false)
+  // 3. Some internal data structures
+  // 4. More to come...
+  //
+  // Note:
+  // If creation of a table reader uses up too much memory
+  // relative to the avaible space left in the block cache
+  // at some point (i.e, causing a cache full under
+  // LRUCacheOptions::strict_capacity_limit = true), such creation will fail
+  // with Status::MemoryLimit().
+  //
+  // Default: false
+  bool reserve_table_reader_memory = false;
+
   // Note: currently this option requires kTwoLevelIndexSearch to be set as
   // well.
   // TODO(myabandeh): remove the note above once the limitation is lifted
   // Use partitioned full filters for each SST file. This option is
-  // incompatible with block-based filters.
+  // incompatible with block-based filters. Filter partition blocks use
+  // block cache even when cache_index_and_filter_blocks=false.
   bool partition_filters = false;
 
-  // EXPERIMENTAL Option to generate Bloom filters that minimize memory
+  // Option to generate Bloom/Ribbon filters that minimize memory
   // internal fragmentation.
   //
   // When false, malloc_usable_size is not available, or format_version < 5,
   // filters are generated without regard to internal fragmentation when
   // loaded into memory (historical behavior). When true (and
-  // malloc_usable_size is available and format_version >= 5), then Bloom
+  // malloc_usable_size is available and format_version >= 5), then
   // filters are generated to "round up" and "round down" their sizes to
   // minimize internal fragmentation when loaded into memory, assuming the
   // reading DB has the same memory allocation characteristics as the
@@ -313,7 +362,8 @@ struct BlockBasedTableOptions {
   // NOTE: Because some memory counted by block cache might be unmapped pages
   // within internal fragmentation, this option can increase observed RSS
   // memory usage. With cache_index_and_filter_blocks=true, this option makes
-  // the block cache better at using space it is allowed.
+  // the block cache better at using space it is allowed. (These issues
+  // should not arise with partitioned filters.)
   //
   // NOTE: Do not set to true if you do not trust malloc_usable_size. With
   // this option, RocksDB might access an allocated memory object beyond its
@@ -337,6 +387,20 @@ struct BlockBasedTableOptions {
   // This must generally be true for gets to be efficient.
   bool whole_key_filtering = true;
 
+  // If true, detect corruption during Bloom Filter (format_version >= 5)
+  // and Ribbon Filter construction.
+  //
+  // This is an extra check that is only
+  // useful in detecting software bugs or CPU+memory malfunction.
+  // Turning on this feature increases filter construction time by 30%.
+  //
+  // This parameter can be changed dynamically by
+  // DB::SetOptions({{"block_based_table_factory",
+  //                  "{detect_filter_construct_corruption=true;}"}});
+  //
+  // TODO: optimize this performance
+  bool detect_filter_construct_corruption = false;
+
   // Verify that decompressing the compressed block gives back the input. This
   // is a verification mode that we use to detect bugs in compression
   // algorithms.
@@ -365,10 +429,9 @@ struct BlockBasedTableOptions {
   // Default: 0 (disabled)
   uint32_t read_amp_bytes_per_bit = 0;
 
-  // We currently have five versions:
-  // 0 -- This version is currently written out by all RocksDB's versions by
-  // default.  Can be read by really old RocksDB's. Doesn't support changing
-  // checksum (default is CRC32).
+  // We currently have these versions:
+  // 0 -- This version can be read by really old RocksDB's. Doesn't support
+  // changing checksum type (default is CRC32).
   // 1 -- Can be read by RocksDB's versions since 3.0. Supports non-default
   // checksum, like xxHash. It is written by RocksDB when
   // BlockBasedTableOptions::checksum is something other than kCRC32c. (version
@@ -391,7 +454,7 @@ struct BlockBasedTableOptions {
   // 5 -- Can be read by RocksDB's versions since 6.6.0. Full and partitioned
   // filters use a generally faster and more accurate Bloom filter
   // implementation, with a different schema.
-  uint32_t format_version = 4;
+  uint32_t format_version = 5;
 
   // Store index blocks on disk in compressed format. Changing this option to
   // false  will avoid the overhead of decompression if index blocks are evicted
@@ -435,6 +498,85 @@ struct BlockBasedTableOptions {
 
   IndexShorteningMode index_shortening =
       IndexShorteningMode::kShortenSeparators;
+
+  // RocksDB does auto-readahead for iterators on noticing more than two reads
+  // for a table file if user doesn't provide readahead_size. The readahead
+  // starts at BlockBasedTableOptions.initial_auto_readahead_size (default: 8KB)
+  // and doubles on every additional read upto max_auto_readahead_size and
+  // max_auto_readahead_size can be configured.
+  //
+  // Special Value: 0 - If max_auto_readahead_size is set 0 then it will disable
+  // the implicit auto prefetching.
+  // If max_auto_readahead_size provided is less
+  // than initial_auto_readahead_size, then RocksDB will sanitize the
+  // initial_auto_readahead_size and set it to max_auto_readahead_size.
+  //
+  // Value should be provided along with KB i.e. 256 * 1024 as it will prefetch
+  // the blocks.
+  //
+  // Found that 256 KB readahead size provides the best performance, based on
+  // experiments, for auto readahead. Experiment data is in PR #3282.
+  //
+  // This parameter can be changed dynamically by
+  // DB::SetOptions({{"block_based_table_factory",
+  //                  "{max_auto_readahead_size=0;}"}}));
+  //
+  // Changing the value dynamically will only affect files opened after the
+  // change.
+  //
+  // Default: 256 KB (256 * 1024).
+  size_t max_auto_readahead_size = 256 * 1024;
+
+  // If enabled, prepopulate warm/hot blocks (data, uncompressed dict, index and
+  // filter blocks) which are already in memory into block cache at the time of
+  // flush. On a flush, the block that is in memory (in memtables) get flushed
+  // to the device. If using Direct IO, additional IO is incurred to read this
+  // data back into memory again, which is avoided by enabling this option. This
+  // further helps if the workload exhibits high temporal locality, where most
+  // of the reads go to recently written data. This also helps in case of
+  // Distributed FileSystem.
+  //
+  // This parameter can be changed dynamically by
+  // DB::SetOptions({{"block_based_table_factory",
+  //                  "{prepopulate_block_cache=kFlushOnly;}"}}));
+  enum class PrepopulateBlockCache : char {
+    // Disable prepopulate block cache.
+    kDisable,
+    // Prepopulate blocks during flush only.
+    kFlushOnly,
+  };
+
+  PrepopulateBlockCache prepopulate_block_cache =
+      PrepopulateBlockCache::kDisable;
+
+  // RocksDB does auto-readahead for iterators on noticing more than two reads
+  // for a table file if user doesn't provide readahead_size. The readahead size
+  // starts at initial_auto_readahead_size and doubles on every additional read
+  // upto BlockBasedTableOptions.max_auto_readahead_size.
+  // max_auto_readahead_size can also be configured.
+  //
+  // Scenarios:
+  // - If initial_auto_readahead_size is set 0 then it will disabled the
+  //   implicit auto prefetching irrespective of max_auto_readahead_size.
+  // - If max_auto_readahead_size is set 0, it will disable the internal
+  //    prefetching irrespective of initial_auto_readahead_size.
+  // - If initial_auto_readahead_size > max_auto_readahead_size, then RocksDB
+  //   will sanitize the value of initial_auto_readahead_size to
+  //   max_auto_readahead_size and readahead_size will be
+  //   max_auto_readahead_size.
+  //
+  // Value should be provided along with KB i.e. 8 * 1024 as it will prefetch
+  // the blocks.
+  //
+  // This parameter can be changed dynamically by
+  // DB::SetOptions({{"block_based_table_factory",
+  //                  "{initial_auto_readahead_size=0;}"}}));
+  //
+  // Changing the value dynamically will only affect files opened after the
+  // change.
+  //
+  // Default: 8 KB (8 * 1024).
+  size_t initial_auto_readahead_size = 8 * 1024;
 };
 
 // Table Properties that are specific to block-based table properties.
@@ -528,7 +670,7 @@ struct PlainTableOptions {
 
   // @store_index_in_file: compute plain table index and bloom filter during
   //                       file building and store it in file. When reading
-  //                       file, index will be mmaped instead of recomputation.
+  //                       file, index will be mapped instead of recomputation.
   bool store_index_in_file = false;
 };
 
@@ -684,7 +826,7 @@ class TableFactory : public Customizable {
   // to use in this table.
   virtual TableBuilder* NewTableBuilder(
       const TableBuilderOptions& table_builder_options,
-      uint32_t column_family_id, WritableFileWriter* file) const = 0;
+      WritableFileWriter* file) const = 0;
 
   // Return is delete range supported
   virtual bool IsDeleteRangeSupported() const { return false; }

package/deps/rocksdb/rocksdb/include/rocksdb/table_properties.h

@@ -5,8 +5,12 @@
 #pragma once
 
 #include <stdint.h>
+
 #include <map>
+#include <memory>
 #include <string>
+
+#include "rocksdb/customizable.h"
 #include "rocksdb/status.h"
 #include "rocksdb/types.h"
 
@@ -26,13 +30,14 @@ namespace ROCKSDB_NAMESPACE {
 //      ++pos) {
 //   ...
 // }
-typedef std::map<std::string, std::string> UserCollectedProperties;
+using UserCollectedProperties = std::map<std::string, std::string>;
 
 // table properties' human-readable names in the property block.
 struct TablePropertiesNames {
   static const std::string kDbId;
   static const std::string kDbSessionId;
   static const std::string kDbHostId;
+  static const std::string kOriginalFileNumber;
   static const std::string kDataSize;
   static const std::string kIndexSize;
   static const std::string kIndexPartitions;
@@ -44,6 +49,7 @@ struct TablePropertiesNames {
   static const std::string kRawValueSize;
   static const std::string kNumDataBlocks;
   static const std::string kNumEntries;
+  static const std::string kNumFilterEntries;
   static const std::string kDeletedKeys;
   static const std::string kMergeOperands;
   static const std::string kNumRangeDeletions;
@@ -61,18 +67,23 @@ struct TablePropertiesNames {
   static const std::string kCreationTime;
   static const std::string kOldestKeyTime;
   static const std::string kFileCreationTime;
+  static const std::string kSlowCompressionEstimatedDataSize;
+  static const std::string kFastCompressionEstimatedDataSize;
 };
 
-extern const std::string kPropertiesBlock;
-extern const std::string kCompressionDictBlock;
-extern const std::string kRangeDelBlock;
-
 // `TablePropertiesCollector` provides the mechanism for users to collect
 // their own properties that they are interested in. This class is essentially
 // a collection of callback functions that will be invoked during table
 // building. It is constructed with TablePropertiesCollectorFactory. The methods
 // don't need to be thread-safe, as we will create exactly one
-// TablePropertiesCollector object per table and then call it sequentially
+// TablePropertiesCollector object per table and then call it sequentially.
+//
+// Statuses from these callbacks are currently logged when not OK, but
+// otherwise ignored by RocksDB.
+//
+// Exceptions MUST NOT propagate out of overridden functions into RocksDB,
+// because RocksDB is not exception-safe. This could cause undefined behavior
+// including data loss, unreported corruption, deadlocks, and more.
 class TablePropertiesCollector {
  public:
   virtual ~TablePropertiesCollector() {}
@@ -99,9 +110,9 @@ class TablePropertiesCollector {
   }
 
   // Called after each new block is cut
-  virtual void BlockAdd(uint64_t /* blockRawBytes */,
-                        uint64_t /* blockCompressedBytesFast */,
-                        uint64_t /* blockCompressedBytesSlow */) {
+  virtual void BlockAdd(uint64_t /* block_raw_bytes */,
+                        uint64_t /* block_compressed_bytes_fast */,
+                        uint64_t /* block_compressed_bytes_slow */) {
     // Nothing to do here. Callback registers can override.
     return;
   }
@@ -125,20 +136,33 @@ class TablePropertiesCollector {
 
 // Constructs TablePropertiesCollector. Internals create a new
 // TablePropertiesCollector for each new table
-class TablePropertiesCollectorFactory {
+//
+// Exceptions MUST NOT propagate out of overridden functions into RocksDB,
+// because RocksDB is not exception-safe. This could cause undefined behavior
+// including data loss, unreported corruption, deadlocks, and more.
+class TablePropertiesCollectorFactory : public Customizable {
  public:
   struct Context {
     uint32_t column_family_id;
+    // The level at creating the SST file (i.e, table), of which the
+    // properties are being collected.
+    int level_at_creation = kUnknownLevelAtCreation;
     static const uint32_t kUnknownColumnFamily;
+    static const int kUnknownLevelAtCreation = -1;
   };
 
-  virtual ~TablePropertiesCollectorFactory() {}
+  ~TablePropertiesCollectorFactory() override {}
+  static const char* Type() { return "TablePropertiesCollectorFactory"; }
+  static Status CreateFromString(
+      const ConfigOptions& options, const std::string& value,
+      std::shared_ptr<TablePropertiesCollectorFactory>* result);
+
   // has to be thread-safe
   virtual TablePropertiesCollector* CreateTablePropertiesCollector(
       TablePropertiesCollectorFactory::Context context) = 0;
 
   // The name of the properties collector can be used for debugging purpose.
-  virtual const char* Name() const = 0;
+  const char* Name() const override = 0;
 
   // Can be overridden by sub-classes to return the Name, followed by
   // configuration info that will // be logged to the info log when the
@@ -150,6 +174,9 @@ class TablePropertiesCollectorFactory {
 // table.
 struct TableProperties {
  public:
+  // the file number at creation time, or 0 for unknown. When known,
+  // combining with db_session_id must uniquely identify an SST file.
+  uint64_t orig_file_number = 0;
   // the total size of all data blocks.
   uint64_t data_size = 0;
   // the size of index block.
@@ -173,6 +200,8 @@ struct TableProperties {
   uint64_t num_data_blocks = 0;
   // the number of entries in this table
   uint64_t num_entries = 0;
+  // the number of unique entries (keys or prefixes) added to filters
+  uint64_t num_filter_entries = 0;
   // the number of deletions in the table
   uint64_t num_deletions = 0;
   // the number of merge operands in the table
@@ -195,6 +224,18 @@ struct TableProperties {
   uint64_t oldest_key_time = 0;
   // Actual SST file creation time. 0 means unknown.
   uint64_t file_creation_time = 0;
+  // Estimated size of data blocks if compressed using a relatively slower
+  // compression algorithm (see `ColumnFamilyOptions::sample_for_compression`).
+  // 0 means unknown.
+  uint64_t slow_compression_estimated_data_size = 0;
+  // Estimated size of data blocks if compressed using a relatively faster
+  // compression algorithm (see `ColumnFamilyOptions::sample_for_compression`).
+  // 0 means unknown.
+  uint64_t fast_compression_estimated_data_size = 0;
+  // Offset of the value of the property "external sst file global seqno" in the
+  // file if the property exists.
+  // 0 means not exists.
+  uint64_t external_sst_file_global_seqno_offset = 0;
 
   // DB identity
   // db_id is an identifier generated the first time the DB is created
@@ -247,9 +288,6 @@ struct TableProperties {
   UserCollectedProperties user_collected_properties;
   UserCollectedProperties readable_properties;
 
-  // The offset of the value of each property in the file.
-  std::map<std::string, uint64_t> properties_offsets;
-
   // convert this object to a human readable form
   //   @prop_delim: delimiter for each property.
   std::string ToString(const std::string& prop_delim = "; ",
@@ -263,6 +301,10 @@ struct TableProperties {
   // between tables. Keys match field names in this class instead
   // of using full property names.
   std::map<std::string, uint64_t> GetAggregatablePropertiesAsMap() const;
+
+  // Return the approximated memory usage of this TableProperties object,
+  // including memory used by the string properties and UserCollectedProperties
+  std::size_t ApproximateMemoryUsage() const;
 };
 
 // Extra properties

package/deps/rocksdb/rocksdb/include/rocksdb/thread_status.h

@@ -13,13 +13,15 @@
 
 #pragma once
 
-#include <stdint.h>
 #include <cstddef>
+#include <cstdint>
 #include <map>
 #include <string>
 #include <utility>
 #include <vector>
 
+#include "rocksdb/rocksdb_namespace.h"
+
 #if !defined(ROCKSDB_LITE) && !defined(NROCKSDB_THREAD_STATUS) && \
     defined(ROCKSDB_SUPPORT_THREAD_LOCAL)
 #define ROCKSDB_USING_THREAD_STATUS

package/deps/rocksdb/rocksdb/include/rocksdb/trace_reader_writer.h

@@ -19,8 +19,7 @@ namespace ROCKSDB_NAMESPACE {
 // a time.
 class TraceWriter {
  public:
-  TraceWriter() {}
-  virtual ~TraceWriter() {}
+  virtual ~TraceWriter() = default;
 
   virtual Status Write(const Slice& data) = 0;
   virtual Status Close() = 0;
@@ -28,21 +27,26 @@ class TraceWriter {
 };
 
 // TraceReader allows reading RocksDB traces from any system, one operation at
-// a time. A RocksDB Replayer could depend on this to replay opertions.
+// a time. A RocksDB Replayer could depend on this to replay operations.
 class TraceReader {
  public:
-  TraceReader() {}
-  virtual ~TraceReader() {}
+  virtual ~TraceReader() = default;
 
   virtual Status Read(std::string* data) = 0;
   virtual Status Close() = 0;
+
+  // Seek back to the trace header. Replayer can call this method to restart
+  // replaying. Note this method may fail if the reader is already closed.
+  virtual Status Reset() = 0;
 };
 
-// Factory methods to read/write traces from/to a file.
+// Factory methods to write/read traces to/from a file.
+// The implementations may not be thread-safe.
 Status NewFileTraceWriter(Env* env, const EnvOptions& env_options,
                           const std::string& trace_filename,
                           std::unique_ptr<TraceWriter>* trace_writer);
 Status NewFileTraceReader(Env* env, const EnvOptions& env_options,
                           const std::string& trace_filename,
                           std::unique_ptr<TraceReader>* trace_reader);
+
 }  // namespace ROCKSDB_NAMESPACE