You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
rocksdb/buckifier/buckify_rocksdb.py

341 lines
12 KiB

#!/usr/bin/env python3
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
from __future__ import absolute_import, division, print_function, unicode_literals
try:
from builtins import str
except ImportError:
from __builtin__ import str
import fnmatch
import json
import os
import sys
from targets_builder import TARGETSBuilder
from util import ColorString
# This script generates TARGETS file for Buck.
# Buck is a build tool specifying dependencies among different build targets.
# User can pass extra dependencies as a JSON object via command line, and this
# script can include these dependencies in the generate TARGETS file.
# Usage:
# $python3 buckifier/buckify_rocksdb.py
# (This generates a TARGET file without user-specified dependency for unit
# tests.)
# $python3 buckifier/buckify_rocksdb.py \
# '{"fake": {
# "extra_deps": [":test_dep", "//fakes/module:mock1"],
# "extra_compiler_flags": ["-DROCKSDB_LITE", "-Os"]
# }
# }'
# (Generated TARGETS file has test_dep and mock1 as dependencies for RocksDB
# unit tests, and will use the extra_compiler_flags to compile the unit test
# source.)
# tests to export as libraries for inclusion in other projects
_EXPORTED_TEST_LIBS = ["env_basic_test"]
# Parse src.mk files as a Dictionary of
# VAR_NAME => list of files
def parse_src_mk(repo_path):
src_mk = repo_path + "/src.mk"
src_files = {}
for line in open(src_mk):
line = line.strip()
if len(line) == 0 or line[0] == "#":
continue
if "=" in line:
current_src = line.split("=")[0].strip()
src_files[current_src] = []
elif ".c" in line:
src_path = line.split("\\")[0].strip()
src_files[current_src].append(src_path)
return src_files
# get all .cc / .c files
def get_cc_files(repo_path):
cc_files = []
for root, _dirnames, filenames in os.walk(
repo_path
): # noqa: B007 T25377293 Grandfathered in
root = root[(len(repo_path) + 1) :]
if "java" in root:
# Skip java
continue
for filename in fnmatch.filter(filenames, "*.cc"):
cc_files.append(os.path.join(root, filename))
for filename in fnmatch.filter(filenames, "*.c"):
cc_files.append(os.path.join(root, filename))
return cc_files
# Get non_parallel tests from Makefile
def get_non_parallel_tests(repo_path):
Makefile = repo_path + "/Makefile"
s = set({})
found_non_parallel_tests = False
for line in open(Makefile):
line = line.strip()
if line.startswith("NON_PARALLEL_TEST ="):
found_non_parallel_tests = True
elif found_non_parallel_tests:
if line.endswith("\\"):
# remove the trailing \
line = line[:-1]
line = line.strip()
s.add(line)
else:
# we consumed all the non_parallel tests
break
return s
# Parse extra dependencies passed by user from command line
def get_dependencies():
deps_map = {"": {"extra_deps": [], "extra_compiler_flags": []}}
if len(sys.argv) < 2:
return deps_map
def encode_dict(data):
rv = {}
for k, v in data.items():
if isinstance(v, dict):
v = encode_dict(v)
rv[k] = v
return rv
extra_deps = json.loads(sys.argv[1], object_hook=encode_dict)
for target_alias, deps in extra_deps.items():
deps_map[target_alias] = deps
return deps_map
# Prepare TARGETS file for buck
def generate_targets(repo_path, deps_map):
print(ColorString.info("Generating TARGETS"))
# parsed src.mk file
src_mk = parse_src_mk(repo_path)
# get all .cc files
cc_files = get_cc_files(repo_path)
# get non_parallel tests from Makefile
non_parallel_tests = get_non_parallel_tests(repo_path)
if src_mk is None or cc_files is None or non_parallel_tests is None:
return False
extra_argv = ""
if len(sys.argv) >= 2:
# Heuristically quote and canonicalize whitespace for inclusion
# in how the file was generated.
extra_argv = " '{0}'".format(" ".join(sys.argv[1].split()))
TARGETS = TARGETSBuilder("%s/TARGETS" % repo_path, extra_argv)
# rocksdb_lib
TARGETS.add_library(
"rocksdb_lib",
src_mk["LIB_SOURCES"] +
# always add range_tree, it's only excluded on ppc64, which we don't use internally
src_mk["RANGE_TREE_SOURCES"] + src_mk["TOOL_LIB_SOURCES"],
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2 years ago
deps=[
"//folly/container:f14_hash",
"//folly/experimental/coro:blocking_wait",
"//folly/experimental/coro:collect",
"//folly/experimental/coro:coroutine",
Use optimized folly DistributedMutex in LRUCache when available (#10179) Summary: folly DistributedMutex is faster than standard mutexes though imposes some static obligations on usage. See https://github.com/facebook/folly/blob/main/folly/synchronization/DistributedMutex.h for details. Here we use this alternative for our Cache implementations (especially LRUCache) for better locking performance, when RocksDB is compiled with folly. Also added information about which distributed mutex implementation is being used to cache_bench output and to DB LOG. Intended follow-up: * Use DMutex in more places, perhaps improving API to support non-scoped locking * Fix linking with fbcode compiler (needs ROCKSDB_NO_FBCODE=1 currently) Credit: Thanks Siying for reminding me about this line of work that was previously left unfinished. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10179 Test Plan: for correctness, existing tests. CircleCI config updated. Also Meta-internal buck build updated. For performance, ran simultaneous before & after cache_bench. Out of three comparison runs, the middle improvement to ops/sec was +21%: Baseline: USE_CLANG=1 DEBUG_LEVEL=0 make -j24 cache_bench (fbcode compiler) ``` Complete in 20.201 s; Rough parallel ops/sec = 1584062 Thread ops/sec = 107176 Operation latency (ns): Count: 32000000 Average: 9257.9421 StdDev: 122412.04 Min: 134 Median: 3623.0493 Max: 56918500 Percentiles: P50: 3623.05 P75: 10288.02 P99: 30219.35 P99.9: 683522.04 P99.99: 7302791.63 ``` New: (add USE_FOLLY=1) ``` Complete in 16.674 s; Rough parallel ops/sec = 1919135 (+21%) Thread ops/sec = 135487 Operation latency (ns): Count: 32000000 Average: 7304.9294 StdDev: 108530.28 Min: 132 Median: 3777.6012 Max: 91030902 Percentiles: P50: 3777.60 P75: 10169.89 P99: 24504.51 P99.9: 59721.59 P99.99: 1861151.83 ``` Reviewed By: anand1976 Differential Revision: D37182983 Pulled By: pdillinger fbshipit-source-id: a17eb05f25b832b6a2c1356f5c657e831a5af8d1
2 years ago
"//folly/experimental/coro:task",
"//folly/synchronization:distributed_mutex",
],
)
# rocksdb_whole_archive_lib
TARGETS.add_library(
"rocksdb_whole_archive_lib",
src_mk["LIB_SOURCES"] +
# always add range_tree, it's only excluded on ppc64, which we don't use internally
src_mk["RANGE_TREE_SOURCES"] + src_mk["TOOL_LIB_SOURCES"],
Multi file concurrency in MultiGet using coroutines and async IO (#9968) Summary: This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code. A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest. TODO: 1. Figure out how to build it in CircleCI (requires some dependencies to be installed) 2. Do some stress testing with coroutines enabled No regression in synchronous MultiGet between this branch and main - ``` ./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics ``` Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)``` Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)``` More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file. 1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) - No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)``` Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)``` 2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file - No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)``` Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)``` 3. Single thread CPU bound workload with ~2 key overlap/file - No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)``` Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)``` 4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file - No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ``` Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968 Reviewed By: akankshamahajan15 Differential Revision: D36348563 Pulled By: anand1976 fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
2 years ago
deps=[
"//folly/container:f14_hash",
"//folly/experimental/coro:blocking_wait",
"//folly/experimental/coro:collect",
"//folly/experimental/coro:coroutine",
Use optimized folly DistributedMutex in LRUCache when available (#10179) Summary: folly DistributedMutex is faster than standard mutexes though imposes some static obligations on usage. See https://github.com/facebook/folly/blob/main/folly/synchronization/DistributedMutex.h for details. Here we use this alternative for our Cache implementations (especially LRUCache) for better locking performance, when RocksDB is compiled with folly. Also added information about which distributed mutex implementation is being used to cache_bench output and to DB LOG. Intended follow-up: * Use DMutex in more places, perhaps improving API to support non-scoped locking * Fix linking with fbcode compiler (needs ROCKSDB_NO_FBCODE=1 currently) Credit: Thanks Siying for reminding me about this line of work that was previously left unfinished. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10179 Test Plan: for correctness, existing tests. CircleCI config updated. Also Meta-internal buck build updated. For performance, ran simultaneous before & after cache_bench. Out of three comparison runs, the middle improvement to ops/sec was +21%: Baseline: USE_CLANG=1 DEBUG_LEVEL=0 make -j24 cache_bench (fbcode compiler) ``` Complete in 20.201 s; Rough parallel ops/sec = 1584062 Thread ops/sec = 107176 Operation latency (ns): Count: 32000000 Average: 9257.9421 StdDev: 122412.04 Min: 134 Median: 3623.0493 Max: 56918500 Percentiles: P50: 3623.05 P75: 10288.02 P99: 30219.35 P99.9: 683522.04 P99.99: 7302791.63 ``` New: (add USE_FOLLY=1) ``` Complete in 16.674 s; Rough parallel ops/sec = 1919135 (+21%) Thread ops/sec = 135487 Operation latency (ns): Count: 32000000 Average: 7304.9294 StdDev: 108530.28 Min: 132 Median: 3777.6012 Max: 91030902 Percentiles: P50: 3777.60 P75: 10169.89 P99: 24504.51 P99.9: 59721.59 P99.99: 1861151.83 ``` Reviewed By: anand1976 Differential Revision: D37182983 Pulled By: pdillinger fbshipit-source-id: a17eb05f25b832b6a2c1356f5c657e831a5af8d1
2 years ago
"//folly/experimental/coro:task",
"//folly/synchronization:distributed_mutex",
],
headers=None,
extra_external_deps="",
link_whole=True,
)
# rocksdb_test_lib
TARGETS.add_library(
"rocksdb_test_lib",
src_mk.get("MOCK_LIB_SOURCES", [])
+ src_mk.get("TEST_LIB_SOURCES", [])
+ src_mk.get("EXP_LIB_SOURCES", [])
+ src_mk.get("ANALYZER_LIB_SOURCES", []),
[":rocksdb_lib"],
extra_test_libs=True,
)
# rocksdb_tools_lib
TARGETS.add_library(
"rocksdb_tools_lib",
src_mk.get("BENCH_LIB_SOURCES", [])
+ src_mk.get("ANALYZER_LIB_SOURCES", [])
+ ["test_util/testutil.cc"],
[":rocksdb_lib"],
)
# rocksdb_cache_bench_tools_lib
TARGETS.add_library(
"rocksdb_cache_bench_tools_lib",
src_mk.get("CACHE_BENCH_LIB_SOURCES", []),
[":rocksdb_lib"],
)
# rocksdb_stress_lib
TARGETS.add_rocksdb_library(
"rocksdb_stress_lib",
src_mk.get("ANALYZER_LIB_SOURCES", [])
+ src_mk.get("STRESS_LIB_SOURCES", [])
+ ["test_util/testutil.cc"],
)
# db_stress binary
TARGETS.add_binary(
"db_stress", ["db_stress_tool/db_stress.cc"], [":rocksdb_stress_lib"]
)
# bench binaries
for src in src_mk.get("MICROBENCH_SOURCES", []):
name = src.rsplit("/", 1)[1].split(".")[0] if "/" in src else src.split(".")[0]
TARGETS.add_binary(name, [src], [], extra_bench_libs=True)
print("Extra dependencies:\n{0}".format(json.dumps(deps_map)))
# Dictionary test executable name -> relative source file path
test_source_map = {}
# c_test.c is added through TARGETS.add_c_test(). If there
# are more than one .c test file, we need to extend
# TARGETS.add_c_test() to include other C tests too.
for test_src in src_mk.get("TEST_MAIN_SOURCES_C", []):
if test_src != "db/c_test.c":
print("Don't know how to deal with " + test_src)
return False
TARGETS.add_c_test()
try:
with open(f"{repo_path}/buckifier/bench.json") as json_file:
fast_fancy_bench_config_list = json.load(json_file)
for config_dict in fast_fancy_bench_config_list:
clean_benchmarks = {}
benchmarks = config_dict["benchmarks"]
for binary, benchmark_dict in benchmarks.items():
clean_benchmarks[binary] = {}
for benchmark, overloaded_metric_list in benchmark_dict.items():
clean_benchmarks[binary][benchmark] = []
for metric in overloaded_metric_list:
if not isinstance(metric, dict):
clean_benchmarks[binary][benchmark].append(metric)
TARGETS.add_fancy_bench_config(
config_dict["name"],
clean_benchmarks,
False,
config_dict["expected_runtime_one_iter"],
config_dict["sl_iterations"],
config_dict["regression_threshold"],
)
with open(f"{repo_path}/buckifier/bench-slow.json") as json_file:
slow_fancy_bench_config_list = json.load(json_file)
for config_dict in slow_fancy_bench_config_list:
clean_benchmarks = {}
benchmarks = config_dict["benchmarks"]
for binary, benchmark_dict in benchmarks.items():
clean_benchmarks[binary] = {}
for benchmark, overloaded_metric_list in benchmark_dict.items():
clean_benchmarks[binary][benchmark] = []
for metric in overloaded_metric_list:
if not isinstance(metric, dict):
clean_benchmarks[binary][benchmark].append(metric)
for config_dict in slow_fancy_bench_config_list:
TARGETS.add_fancy_bench_config(
config_dict["name"] + "_slow",
clean_benchmarks,
True,
config_dict["expected_runtime_one_iter"],
config_dict["sl_iterations"],
config_dict["regression_threshold"],
)
# it is better servicelab experiments break
# than rocksdb github ci
except Exception:
pass
TARGETS.add_test_header()
for test_src in src_mk.get("TEST_MAIN_SOURCES", []):
test = test_src.split(".c")[0].strip().split("/")[-1].strip()
test_source_map[test] = test_src
print("" + test + " " + test_src)
for target_alias, deps in deps_map.items():
for test, test_src in sorted(test_source_map.items()):
if len(test) == 0:
print(ColorString.warning("Failed to get test name for %s" % test_src))
continue
test_target_name = test if not target_alias else test + "_" + target_alias
if test in _EXPORTED_TEST_LIBS:
test_library = "%s_lib" % test_target_name
TARGETS.add_library(
test_library,
[test_src],
deps=[":rocksdb_test_lib"],
extra_test_libs=True,
)
TARGETS.register_test(
test_target_name,
test_src,
deps=json.dumps(deps["extra_deps"] + [":" + test_library]),
extra_compiler_flags=json.dumps(deps["extra_compiler_flags"]),
)
else:
TARGETS.register_test(
test_target_name,
test_src,
deps=json.dumps(deps["extra_deps"] + [":rocksdb_test_lib"]),
extra_compiler_flags=json.dumps(deps["extra_compiler_flags"]),
)
print(ColorString.info("Generated TARGETS Summary:"))
print(ColorString.info("- %d libs" % TARGETS.total_lib))
print(ColorString.info("- %d binarys" % TARGETS.total_bin))
print(ColorString.info("- %d tests" % TARGETS.total_test))
return True
def get_rocksdb_path():
# rocksdb = {script_dir}/..
script_dir = os.path.dirname(sys.argv[0])
script_dir = os.path.abspath(script_dir)
rocksdb_path = os.path.abspath(os.path.join(script_dir, "../"))
return rocksdb_path
def exit_with_error(msg):
print(ColorString.error(msg))
sys.exit(1)
def main():
deps_map = get_dependencies()
# Generate TARGETS file for buck
ok = generate_targets(get_rocksdb_path(), deps_map)
if not ok:
exit_with_error("Failed to generate TARGETS files")
if __name__ == "__main__":
main()