rocksdb/table/table.cc
Abhishek Kona 391885c4e4 stat's collection in leveldb
Summary:
Prototype stat's collection. Diff is a good estimate of what
the final code will look like.
A few assumptions :
  * Used a global static instance of the statistics object. Plan to pass
  it to each internal function. Static allows metrics only at app
  level.
  * In the Ticker's do not do any locking. Depend on the mutex at each
   function of LevelDB. If we ever remove the mutex, we should change
   here too. The other option is use atomic objects anyways as there
   won't be any contention as they will be always acquired only by one
   thread.
  * The counters are dumb, increment through lifecycle. Plan to use ods
    etc to get last5min stat etc.

Test Plan:
made changes in db_bench
Ran ./db_bench --statistics=1 --num=10000 --cache_size=5000
This will print the cache hit/miss stats.

Reviewers: dhruba, heyongqiang

Differential Revision: https://reviews.facebook.net/D6441
2012-11-08 13:55:49 -08:00

294 lines
8.6 KiB
C++

// 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.
#include "leveldb/table.h"
#include "leveldb/cache.h"
#include "leveldb/comparator.h"
#include "leveldb/env.h"
#include "leveldb/filter_policy.h"
#include "leveldb/options.h"
#include "leveldb/statistics.h"
#include "table/block.h"
#include "table/filter_block.h"
#include "table/format.h"
#include "table/two_level_iterator.h"
#include "util/coding.h"
namespace leveldb {
struct Table::Rep {
~Rep() {
delete filter;
delete [] filter_data;
delete index_block;
}
Options options;
Status status;
RandomAccessFile* file;
uint64_t cache_id;
FilterBlockReader* filter;
const char* filter_data;
BlockHandle metaindex_handle; // Handle to metaindex_block: saved from footer
Block* index_block;
};
Status Table::Open(const Options& options,
RandomAccessFile* file,
uint64_t size,
Table** table) {
*table = NULL;
if (size < Footer::kEncodedLength) {
return Status::InvalidArgument("file is too short to be an sstable");
}
char footer_space[Footer::kEncodedLength];
Slice footer_input;
Status s = file->Read(size - Footer::kEncodedLength, Footer::kEncodedLength,
&footer_input, footer_space);
if (!s.ok()) return s;
Footer footer;
s = footer.DecodeFrom(&footer_input);
if (!s.ok()) return s;
// Read the index block
BlockContents contents;
Block* index_block = NULL;
if (s.ok()) {
s = ReadBlock(file, ReadOptions(), footer.index_handle(), &contents);
if (s.ok()) {
index_block = new Block(contents);
}
}
if (s.ok()) {
// We've successfully read the footer and the index block: we're
// ready to serve requests.
Rep* rep = new Table::Rep;
rep->options = options;
rep->file = file;
rep->metaindex_handle = footer.metaindex_handle();
rep->index_block = index_block;
rep->cache_id = (options.block_cache ? options.block_cache->NewId() : 0);
rep->filter_data = NULL;
rep->filter = NULL;
*table = new Table(rep);
(*table)->ReadMeta(footer);
} else {
if (index_block) delete index_block;
}
return s;
}
void Table::ReadMeta(const Footer& footer) {
if (rep_->options.filter_policy == NULL) {
return; // Do not need any metadata
}
// TODO(sanjay): Skip this if footer.metaindex_handle() size indicates
// it is an empty block.
ReadOptions opt;
BlockContents contents;
if (!ReadBlock(rep_->file, opt, footer.metaindex_handle(), &contents).ok()) {
// Do not propagate errors since meta info is not needed for operation
return;
}
Block* meta = new Block(contents);
Iterator* iter = meta->NewIterator(BytewiseComparator());
std::string key = "filter.";
key.append(rep_->options.filter_policy->Name());
iter->Seek(key);
if (iter->Valid() && iter->key() == Slice(key)) {
ReadFilter(iter->value());
}
delete iter;
delete meta;
}
void Table::ReadFilter(const Slice& filter_handle_value) {
Slice v = filter_handle_value;
BlockHandle filter_handle;
if (!filter_handle.DecodeFrom(&v).ok()) {
return;
}
// We might want to unify with ReadBlock() if we start
// requiring checksum verification in Table::Open.
ReadOptions opt;
BlockContents block;
if (!ReadBlock(rep_->file, opt, filter_handle, &block).ok()) {
return;
}
if (block.heap_allocated) {
rep_->filter_data = block.data.data(); // Will need to delete later
}
rep_->filter = new FilterBlockReader(rep_->options.filter_policy, block.data);
}
Table::~Table() {
delete rep_;
}
static void DeleteBlock(void* arg, void* ignored) {
delete reinterpret_cast<Block*>(arg);
}
static void DeleteCachedBlock(const Slice& key, void* value) {
Block* block = reinterpret_cast<Block*>(value);
delete block;
}
static void ReleaseBlock(void* arg, void* h) {
Cache* cache = reinterpret_cast<Cache*>(arg);
Cache::Handle* handle = reinterpret_cast<Cache::Handle*>(h);
cache->Release(handle);
}
// Convert an index iterator value (i.e., an encoded BlockHandle)
// into an iterator over the contents of the corresponding block.
Iterator* Table::BlockReader(void* arg,
const ReadOptions& options,
const Slice& index_value,
bool* didIO) {
Table* table = reinterpret_cast<Table*>(arg);
Cache* block_cache = table->rep_->options.block_cache;
Statistics* const statistics = table->rep_->options.statistics;
Block* block = NULL;
Cache::Handle* cache_handle = NULL;
BlockHandle handle;
Slice input = index_value;
Status s = handle.DecodeFrom(&input);
// We intentionally allow extra stuff in index_value so that we
// can add more features in the future.
if (s.ok()) {
BlockContents contents;
if (block_cache != NULL) {
char cache_key_buffer[16];
EncodeFixed64(cache_key_buffer, table->rep_->cache_id);
EncodeFixed64(cache_key_buffer+8, handle.offset());
Slice key(cache_key_buffer, sizeof(cache_key_buffer));
cache_handle = block_cache->Lookup(key);
if (cache_handle != NULL) {
block = reinterpret_cast<Block*>(block_cache->Value(cache_handle));
RecordTick(statistics, BLOCK_CACHE_HIT);
} else {
s = ReadBlock(table->rep_->file, options, handle, &contents);
if (s.ok()) {
block = new Block(contents);
if (contents.cachable && options.fill_cache) {
cache_handle = block_cache->Insert(
key, block, block->size(), &DeleteCachedBlock);
}
}
if (didIO != NULL) {
*didIO = true; // we did some io from storage
}
RecordTick(statistics, BLOCK_CACHE_MISS);
}
} else {
s = ReadBlock(table->rep_->file, options, handle, &contents);
if (s.ok()) {
block = new Block(contents);
}
}
}
Iterator* iter;
if (block != NULL) {
iter = block->NewIterator(table->rep_->options.comparator);
if (cache_handle == NULL) {
iter->RegisterCleanup(&DeleteBlock, block, NULL);
} else {
iter->RegisterCleanup(&ReleaseBlock, block_cache, cache_handle);
}
} else {
iter = NewErrorIterator(s);
}
return iter;
}
Iterator* Table::BlockReader(void* arg,
const ReadOptions& options,
const Slice& index_value) {
return BlockReader(arg, options, index_value, NULL);
}
Iterator* Table::NewIterator(const ReadOptions& options) const {
return NewTwoLevelIterator(
rep_->index_block->NewIterator(rep_->options.comparator),
&Table::BlockReader, const_cast<Table*>(this), options);
}
Status Table::InternalGet(const ReadOptions& options, const Slice& k,
void* arg,
void (*saver)(void*, const Slice&, const Slice&, bool)) {
Status s;
Iterator* iiter = rep_->index_block->NewIterator(rep_->options.comparator);
iiter->Seek(k);
if (iiter->Valid()) {
Slice handle_value = iiter->value();
FilterBlockReader* filter = rep_->filter;
BlockHandle handle;
if (filter != NULL &&
handle.DecodeFrom(&handle_value).ok() &&
!filter->KeyMayMatch(handle.offset(), k)) {
// Not found
} else {
bool didIO = false;
Iterator* block_iter = BlockReader(this, options, iiter->value(),
&didIO);
block_iter->Seek(k);
if (block_iter->Valid()) {
(*saver)(arg, block_iter->key(), block_iter->value(), didIO);
}
s = block_iter->status();
delete block_iter;
}
}
if (s.ok()) {
s = iiter->status();
}
delete iiter;
return s;
}
uint64_t Table::ApproximateOffsetOf(const Slice& key) const {
Iterator* index_iter =
rep_->index_block->NewIterator(rep_->options.comparator);
index_iter->Seek(key);
uint64_t result;
if (index_iter->Valid()) {
BlockHandle handle;
Slice input = index_iter->value();
Status s = handle.DecodeFrom(&input);
if (s.ok()) {
result = handle.offset();
} else {
// Strange: we can't decode the block handle in the index block.
// We'll just return the offset of the metaindex block, which is
// close to the whole file size for this case.
result = rep_->metaindex_handle.offset();
}
} else {
// key is past the last key in the file. Approximate the offset
// by returning the offset of the metaindex block (which is
// right near the end of the file).
result = rep_->metaindex_handle.offset();
}
delete index_iter;
return result;
}
} // namespace leveldb