From 3622cfa34aca061eaf0595c5fa711f2b34fd3a07 Mon Sep 17 00:00:00 2001 From: Hui Xiao Date: Mon, 24 Apr 2023 15:14:23 -0700 Subject: [PATCH] Add back io_uring stress test hack with DbStressFSWrapper for FS not supporting read async (#11404) Summary: **Context/Summary:** To better utilize `DbStressFSWrapper` for some assertion, https://github.com/facebook/rocksdb/pull/11288 removed an io_uring stress test hack for POSIX FS not supporting read async added in https://github.com/facebook/rocksdb/pull/11242 = It was removed based on the assumption that a later PR https://github.com/facebook/rocksdb/pull/11296 is sufficient to serve as an alternative workaround. But recent stress tests has shown the opposite, mostly because 11296 approach might be subjected to incompleteness when more `ReadOptions` are passed down as what https://github.com/facebook/rocksdb/pull/11288 has done. As a short-term solution to both work around POSIX FS constraint above and utilize `DbStressFSWrapper` for 11288 assertion, I proposed this PR. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11404 Test Plan: - Stress test ensures 11288's assertion is still effective in `DbStressFSWrapper` ``` ./db_stress --acquire_snapshot_one_in=10000 --adaptive_readahead=0 --allow_data_in_errors=True --async_io=1 --avoid_flush_during_recovery=1 --avoid_unnecessary_blocking_io=0 --backup_max_size=104857600 --backup_one_in=100000 --batch_protection_bytes_per_key=8 --block_size=16384 --bloom_bits=16 --bottommost_compression_type=disable --bytes_per_sync=0 --cache_index_and_filter_blocks=0 --cache_size=8388608 --cache_type=hyper_clock_cache --charge_compression_dictionary_building_buffer=0 --charge_file_metadata=1 --charge_filter_construction=1 --charge_table_reader=0 --checkpoint_one_in=1000000 --checksum_type=kxxHash64 --clear_column_family_one_in=0 --compact_files_one_in=1000000 --compact_range_one_in=1000000 --compaction_pri=1 --compaction_ttl=0 --compression_max_dict_buffer_bytes=32767 --compression_max_dict_bytes=16384 --compression_parallel_threads=1 --compression_type=lz4 --compression_use_zstd_dict_trainer=1 --compression_zstd_max_train_bytes=0 --continuous_verification_interval=0 --data_block_index_type=0 --db=$db --db_write_buffer_size=0 --delpercent=4 --delrangepercent=1 --destroy_db_initially=0 --detect_filter_construct_corruption=1 --disable_wal=0 --enable_compaction_filter=0 --enable_pipelined_write=1 --enable_thread_tracking=0 --expected_values_dir=$exp --fail_if_options_file_error=1 --fifo_allow_compaction=0 --file_checksum_impl=crc32c --flush_one_in=1000000 --format_version=4 --get_current_wal_file_one_in=0 --get_live_files_one_in=1000000 --get_property_one_in=1000000 --get_sorted_wal_files_one_in=0 --index_block_restart_interval=4 --index_type=0 --ingest_external_file_one_in=0 --initial_auto_readahead_size=16384 --iterpercent=10 --key_len_percent_dist=1,30,69 --kill_random_test=888887 --level_compaction_dynamic_level_bytes=0 --lock_wal_one_in=1000000 --log2_keys_per_lock=10 --long_running_snapshots=0 --manual_wal_flush_one_in=1000 --mark_for_compaction_one_file_in=10 --max_auto_readahead_size=16384 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_key_len=3 --max_manifest_file_size=1073741824 --max_write_batch_group_size_bytes=1048576 --max_write_buffer_number=3 --max_write_buffer_size_to_maintain=8388608 --memtable_prefix_bloom_size_ratio=0.1 --memtable_protection_bytes_per_key=4 --memtable_whole_key_filtering=0 --memtablerep=skip_list --min_write_buffer_number_to_merge=2 --mmap_read=0 --mock_direct_io=True --nooverwritepercent=1 --num_file_reads_for_auto_readahead=0 --open_files=100 --open_metadata_write_fault_one_in=0 --open_read_fault_one_in=0 --open_write_fault_one_in=0 --ops_per_thread=20000000 --optimize_filters_for_memory=0 --paranoid_file_checks=0 --partition_filters=0 --partition_pinning=0 --pause_background_one_in=1000000 --periodic_compaction_seconds=0 --prefix_size=1 --prefixpercent=5 --prepopulate_block_cache=1 --preserve_internal_time_seconds=36000 --progress_reports=0 --read_fault_one_in=32 --readahead_size=16384 --readpercent=45 --recycle_log_file_num=0 --reopen=20 --ribbon_starting_level=1 --secondary_cache_fault_one_in=32 --secondary_cache_uri=compressed_secondary_cache://capacity=8388608 --snapshot_hold_ops=100000 --sst_file_manager_bytes_per_sec=0 --sst_file_manager_bytes_per_truncate=0 --stats_dump_period_sec=10 --subcompactions=1 --sync=0 --sync_fault_injection=1 --target_file_size_base=2097152 --target_file_size_multiplier=2 --test_batches_snapshots=0 --top_level_index_pinning=2 --unpartitioned_pinning=3 --use_direct_io_for_flush_and_compaction=0 --use_direct_reads=1 --use_full_merge_v1=0 --use_get_entity=0 --use_merge=1 --use_multi_get_entity=0 --use_multiget=1 --use_put_entity_one_in=0 --user_timestamp_size=0 --value_size_mult=32 --verify_checksum=1 --verify_checksum_one_in=1000000 --verify_db_one_in=100000 --verify_sst_unique_id_in_manifest=1 --wal_bytes_per_sync=524288 --wal_compression=none --write_buffer_size=4194304 --write_dbid_to_manifest=1 --writepercent=35 ``` - Monitor future stress test to show `MultiGet error: Not implemented: ReadAsync` is gone Reviewed By: ltamasi Differential Revision: D45242280 Pulled By: hx235 fbshipit-source-id: 9823e3fbd4e9672efdd31478a2f2cbd68a98bdf5 --- db_stress_tool/db_stress_tool.cc | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/db_stress_tool/db_stress_tool.cc b/db_stress_tool/db_stress_tool.cc index c41c5051f..62e96af23 100644 --- a/db_stress_tool/db_stress_tool.cc +++ b/db_stress_tool/db_stress_tool.cc @@ -31,6 +31,7 @@ namespace ROCKSDB_NAMESPACE { namespace { static std::shared_ptr env_guard; static std::shared_ptr env_wrapper_guard; +static std::shared_ptr legacy_env_wrapper_guard; static std::shared_ptr dbsl_env_wrapper_guard; static std::shared_ptr fault_env_guard; @@ -99,6 +100,17 @@ int db_stress_tool(int argc, char** argv) { env_wrapper_guard = std::make_shared( raw_env, std::make_shared(raw_env->GetFileSystem())); + if (!env_opts && !FLAGS_use_io_uring) { + // If using the default Env (Posix), wrap DbStressEnvWrapper with the + // legacy EnvWrapper. This is a workaround to prevent MultiGet and scans + // from failing when IO uring is disabled. The EnvWrapper + // has a default implementation of ReadAsync that redirects to Read. + legacy_env_wrapper_guard = std::make_shared(raw_env); + env_wrapper_guard = std::make_shared( + legacy_env_wrapper_guard, + std::make_shared( + legacy_env_wrapper_guard->GetFileSystem())); + } db_stress_env = env_wrapper_guard.get(); FLAGS_rep_factory = StringToRepFactory(FLAGS_memtablerep.c_str());