Fix estimation of batch storage size when no blocks exist

If there's no blocks in database (m_height == 0):
  Don't assign incorrect block range to check.
  Skip average block size check.

Test:

Run blockchain_converter with an existing source blockchain.bin and
a non-existent LMDB destination database.

The converter creates a BlockchainLMDB instance with zero height, due to
not being initialized with a genesis block, normally done by
Blockchain::init(). While different than the behavior of bitmonerod,
blockchain_import, and blockchain_export, the initialization hasn't been
strictly necessary.

The db batch size estimation normally uses an average block size, or a
default minimum block size, whichever is greater. In this case, as
there's no existing blocks to check for an average block size, the
default should be used.
This commit is contained in:
warptangent 2015-08-04 14:59:42 -07:00
parent 9c3715e46a
commit 525bf5811f
No known key found for this signature in database
GPG key ID: 0E490BEBFBE4E92D

View file

@ -28,6 +28,7 @@
#include "db_lmdb.h"
#include <boost/filesystem.hpp>
#include <boost/format.hpp>
#include <memory> // std::unique_ptr
#include <cstring> // memcpy
#include <random>
@ -262,6 +263,7 @@ void mdb_txn_safe::allow_new_txns()
void BlockchainLMDB::do_resize(uint64_t increase_size)
{
LOG_PRINT_L3("BlockchainLMDB::" << __func__);
CRITICAL_REGION_LOCAL(m_synchronization_lock);
const uint64_t add_size = 1LL << 30;
@ -327,6 +329,7 @@ void BlockchainLMDB::do_resize(uint64_t increase_size)
// threshold_size is used for batch transactions
bool BlockchainLMDB::need_resize(uint64_t threshold_size) const
{
LOG_PRINT_L3("BlockchainLMDB::" << __func__);
#if defined(ENABLE_AUTO_RESIZE)
MDB_envinfo mei;
@ -346,7 +349,8 @@ bool BlockchainLMDB::need_resize(uint64_t threshold_size) const
LOG_PRINT_L1("Space used: " << size_used);
LOG_PRINT_L1("Space remaining: " << mei.me_mapsize - size_used);
LOG_PRINT_L1("Size threshold: " << threshold_size);
LOG_PRINT_L1("Percent used: " << (double)size_used/mei.me_mapsize << " Percent threshold: " << RESIZE_PERCENT);
float resize_percent_old = RESIZE_PERCENT;
LOG_PRINT_L1(boost::format("Percent used: %.04f Percent threshold: %.04f") % ((double)size_used/mei.me_mapsize) % resize_percent_old);
if (threshold_size > 0)
{
@ -376,7 +380,8 @@ bool BlockchainLMDB::need_resize(uint64_t threshold_size) const
void BlockchainLMDB::check_and_resize_for_batch(uint64_t batch_num_blocks)
{
LOG_PRINT_L1("[batch] checking DB size");
LOG_PRINT_L3("BlockchainLMDB::" << __func__);
LOG_PRINT_L1("[" << __func__ << "] " << "checking DB size");
const uint64_t min_increase_size = 128 * (1 << 20);
uint64_t threshold_size = 0;
uint64_t increase_size = 0;
@ -407,6 +412,7 @@ void BlockchainLMDB::check_and_resize_for_batch(uint64_t batch_num_blocks)
uint64_t BlockchainLMDB::get_estimated_batch_size(uint64_t batch_num_blocks) const
{
LOG_PRINT_L3("BlockchainLMDB::" << __func__);
uint64_t threshold_size = 0;
// batch size estimate * batch safety factor = final size estimate
@ -419,12 +425,22 @@ uint64_t BlockchainLMDB::get_estimated_batch_size(uint64_t batch_num_blocks) con
// For resizing purposes, allow for at least 4k average block size.
uint64_t min_block_size = 4 * 1024;
uint64_t block_stop = m_height - 1;
uint64_t block_stop = 0;
if (m_height > 1)
block_stop = m_height - 1;
uint64_t block_start = 0;
if (block_stop >= num_prev_blocks)
block_start = block_stop - num_prev_blocks + 1;
uint32_t num_blocks_used = 0;
uint64_t total_block_size = 0;
LOG_PRINT_L1("[" << __func__ << "] " << "m_height: " << m_height << " block_start: " << block_start << " block_stop: " << block_stop);
size_t avg_block_size = 0;
if (m_height == 0)
{
LOG_PRINT_L1("No existing blocks to check for average block size");
}
else
{
for (uint64_t block_num = block_start; block_num <= block_stop; ++block_num)
{
uint32_t block_size = get_block_size(block_num);
@ -433,8 +449,9 @@ uint64_t BlockchainLMDB::get_estimated_batch_size(uint64_t batch_num_blocks) con
// some blocks were to be skipped for being outliers.
++num_blocks_used;
}
size_t avg_block_size = total_block_size / num_blocks_used;
avg_block_size = total_block_size / num_blocks_used;
LOG_PRINT_L1("average block size across recent " << num_blocks_used << " blocks: " << avg_block_size);
}
if (avg_block_size < min_block_size)
avg_block_size = min_block_size;
LOG_PRINT_L1("estimated average block size for batch: " << avg_block_size);