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/util/rate_limiter_test.cc

434 lines
18 KiB

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
//
// 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.
fix rate limiter to avoid starvation Summary: The current implementation of rate limiter has the possibility to introduce resource starvation when change its limit. This diff aims to fix this problem by consuming request bytes partially. Test Plan: ``` ./rate_limiter_test [==========] Running 4 tests from 1 test case. [----------] Global test environment set-up. [----------] 4 tests from RateLimiterTest [ RUN ] RateLimiterTest.OverflowRate [ OK ] RateLimiterTest.OverflowRate (0 ms) [ RUN ] RateLimiterTest.StartStop [ OK ] RateLimiterTest.StartStop (0 ms) [ RUN ] RateLimiterTest.Rate request size [1 - 1023], limit 10 KB/sec, actual rate: 10.355712 KB/sec, elapsed 2.00 seconds request size [1 - 1023], limit 20 KB/sec, actual rate: 19.136564 KB/sec, elapsed 2.00 seconds request size [1 - 2047], limit 20 KB/sec, actual rate: 20.783976 KB/sec, elapsed 2.10 seconds request size [1 - 2047], limit 40 KB/sec, actual rate: 39.308144 KB/sec, elapsed 2.10 seconds request size [1 - 4095], limit 40 KB/sec, actual rate: 40.318349 KB/sec, elapsed 2.20 seconds request size [1 - 4095], limit 80 KB/sec, actual rate: 79.667396 KB/sec, elapsed 2.20 seconds request size [1 - 8191], limit 80 KB/sec, actual rate: 81.807158 KB/sec, elapsed 2.30 seconds request size [1 - 8191], limit 160 KB/sec, actual rate: 160.659761 KB/sec, elapsed 2.20 seconds request size [1 - 16383], limit 160 KB/sec, actual rate: 160.700990 KB/sec, elapsed 3.00 seconds request size [1 - 16383], limit 320 KB/sec, actual rate: 317.639481 KB/sec, elapsed 2.50 seconds [ OK ] RateLimiterTest.Rate (22618 ms) [ RUN ] RateLimiterTest.LimitChangeTest [COMPLETE] request size 10 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 10 KB, new limit 5KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 10KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 80KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 160KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 320KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 80KB/sec, refill period 1000 ms [ OK ] RateLimiterTest.LimitChangeTest (5002 ms) [----------] 4 tests from RateLimiterTest (27620 ms total) [----------] Global test environment tear-down [==========] 4 tests from 1 test case ran. (27621 ms total) [ PASSED ] 4 tests. ``` Reviewers: sdong, IslamAbdelRahman, yiwu, andrewkr Reviewed By: andrewkr Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D60207
9 years ago
#include "util/rate_limiter.h"
#include <chrono>
#include <cinttypes>
Implement superior user & mid IO priority level in GenericRateLimiter (#8595) Summary: Context: An extra IO_USER priority in rate limiter allows users to optionally charge WAL writes / SST reads to rate limiter at this priority level, which then has higher priority than IO_HIGH and IO_LOW. With an extra IO_USER priority, it allows users to better specify the relative urgency/importance among different requests in rate limiter. As a consequence, IO resource management can better prioritize and limit resource based on user's need. The IO_USER is implemented as superior priority in GenericRateLimiter, in the sense that its request queue will always be iterated first without being constrained to fairness. The reason is that the notion of fairness is only meaningful in helping lower priorities in background IO (i.e, IO_HIGH/MID/LOW) to gain some fair chance to run so that it does not block foreground IO (i.e, the ones that are charged at the level of IO_USER). As we can see, the ultimate goal here is to not blocking foreground IO at IO_USER level, which justifies the superiority of IO_USER. Similar benefits exist for IO_MID priority. - Rewrote the logic of deciding the order of iterating request queues of high/low priorities to include the extra user/mid priority w/o affecting the existing behavior (see PR's [comment](https://github.com/facebook/rocksdb/pull/8595/files#r678749331)) - Included the request queue of user-pri/mid-pri in the code path of next-leader-candidate signaling and GenericRateLimiter's destructor - Included the extra user/mid-pri in bookkeeping data structures: total_bytes_through_ and total_requests_ - Re-written the previous impl of explicitly iterating priorities with a loop from Env::IO_LOW to Env::IO_TOTAL Pull Request resolved: https://github.com/facebook/rocksdb/pull/8595 Test Plan: - passed existing rate_limiter_test.cc - passed added unit tests in rate_limiter_test.cc - run performance test to verify performance with only high/low requests is not affected by this change - Set-up command: `TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=fillrandom --duration=5 --compression_type=none --num=100000000 --disable_auto_compactions=true --write_buffer_size=1048576 --writable_file_max_buffer_size=65536 --target_file_size_base=1048576 --max_bytes_for_level_base=4194304 --level0_slowdown_writes_trigger=$(((1 << 31) - 1)) --level0_stop_writes_trigger=$(((1 << 31) - 1))` - Test command: `TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=overwrite --use_existing_db=true --disable_wal=true --duration=30 --compression_type=none --num=100000000 --write_buffer_size=1048576 --writable_file_max_buffer_size=65536 --target_file_size_base=1048576 --max_bytes_for_level_base=4194304 --level0_slowdown_writes_trigger=$(((1 << 31) - 1)) --level0_stop_writes_trigger=$(((1 << 31) - 1)) --statistics=true --rate_limiter_bytes_per_sec=1048576 --rate_limiter_refill_period_us=1000 --threads=32 |& grep -E '(flush|compact)\.write\.bytes'` - Before (on branch upstream/master): `rocksdb.compact.write.bytes COUNT : 4014162` `rocksdb.flush.write.bytes COUNT : 26715832` rocksdb.flush.write.bytes/rocksdb.compact.write.bytes ~= 6.66 - After (on branch rate_limiter_user_pri): `rocksdb.compact.write.bytes COUNT : 3807822` `rocksdb.flush.write.bytes COUNT : 26098659` rocksdb.flush.write.bytes/rocksdb.compact.write.bytes ~= 6.85 Reviewed By: ajkr Differential Revision: D30577783 Pulled By: hx235 fbshipit-source-id: 0881f2705ffd13ecd331256bde7e8ec874a353f4
3 years ago
#include <cstdint>
#include <limits>
#include "db/db_test_util.h"
#include "port/port.h"
#include "rocksdb/system_clock.h"
#include "test_util/sync_point.h"
#include "test_util/testharness.h"
#include "util/random.h"
namespace ROCKSDB_NAMESPACE {
// TODO(yhchiang): the rate will not be accurate when we run test in parallel.
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
10 years ago
class RateLimiterTest : public testing::Test {};
TEST_F(RateLimiterTest, OverflowRate) {
GenericRateLimiter limiter(port::kMaxInt64, 1000, 10,
RateLimiter::Mode::kWritesOnly,
SystemClock::Default(), false /* auto_tuned */);
ASSERT_GT(limiter.GetSingleBurstBytes(), 1000000000ll);
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
10 years ago
TEST_F(RateLimiterTest, StartStop) {
std::unique_ptr<RateLimiter> limiter(NewGenericRateLimiter(100, 100, 10));
}
Implement superior user & mid IO priority level in GenericRateLimiter (#8595) Summary: Context: An extra IO_USER priority in rate limiter allows users to optionally charge WAL writes / SST reads to rate limiter at this priority level, which then has higher priority than IO_HIGH and IO_LOW. With an extra IO_USER priority, it allows users to better specify the relative urgency/importance among different requests in rate limiter. As a consequence, IO resource management can better prioritize and limit resource based on user's need. The IO_USER is implemented as superior priority in GenericRateLimiter, in the sense that its request queue will always be iterated first without being constrained to fairness. The reason is that the notion of fairness is only meaningful in helping lower priorities in background IO (i.e, IO_HIGH/MID/LOW) to gain some fair chance to run so that it does not block foreground IO (i.e, the ones that are charged at the level of IO_USER). As we can see, the ultimate goal here is to not blocking foreground IO at IO_USER level, which justifies the superiority of IO_USER. Similar benefits exist for IO_MID priority. - Rewrote the logic of deciding the order of iterating request queues of high/low priorities to include the extra user/mid priority w/o affecting the existing behavior (see PR's [comment](https://github.com/facebook/rocksdb/pull/8595/files#r678749331)) - Included the request queue of user-pri/mid-pri in the code path of next-leader-candidate signaling and GenericRateLimiter's destructor - Included the extra user/mid-pri in bookkeeping data structures: total_bytes_through_ and total_requests_ - Re-written the previous impl of explicitly iterating priorities with a loop from Env::IO_LOW to Env::IO_TOTAL Pull Request resolved: https://github.com/facebook/rocksdb/pull/8595 Test Plan: - passed existing rate_limiter_test.cc - passed added unit tests in rate_limiter_test.cc - run performance test to verify performance with only high/low requests is not affected by this change - Set-up command: `TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=fillrandom --duration=5 --compression_type=none --num=100000000 --disable_auto_compactions=true --write_buffer_size=1048576 --writable_file_max_buffer_size=65536 --target_file_size_base=1048576 --max_bytes_for_level_base=4194304 --level0_slowdown_writes_trigger=$(((1 << 31) - 1)) --level0_stop_writes_trigger=$(((1 << 31) - 1))` - Test command: `TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=overwrite --use_existing_db=true --disable_wal=true --duration=30 --compression_type=none --num=100000000 --write_buffer_size=1048576 --writable_file_max_buffer_size=65536 --target_file_size_base=1048576 --max_bytes_for_level_base=4194304 --level0_slowdown_writes_trigger=$(((1 << 31) - 1)) --level0_stop_writes_trigger=$(((1 << 31) - 1)) --statistics=true --rate_limiter_bytes_per_sec=1048576 --rate_limiter_refill_period_us=1000 --threads=32 |& grep -E '(flush|compact)\.write\.bytes'` - Before (on branch upstream/master): `rocksdb.compact.write.bytes COUNT : 4014162` `rocksdb.flush.write.bytes COUNT : 26715832` rocksdb.flush.write.bytes/rocksdb.compact.write.bytes ~= 6.66 - After (on branch rate_limiter_user_pri): `rocksdb.compact.write.bytes COUNT : 3807822` `rocksdb.flush.write.bytes COUNT : 26098659` rocksdb.flush.write.bytes/rocksdb.compact.write.bytes ~= 6.85 Reviewed By: ajkr Differential Revision: D30577783 Pulled By: hx235 fbshipit-source-id: 0881f2705ffd13ecd331256bde7e8ec874a353f4
3 years ago
TEST_F(RateLimiterTest, GetTotalBytesThrough) {
std::unique_ptr<RateLimiter> limiter(NewGenericRateLimiter(
20 /* rate_bytes_per_sec */, 1000 * 1000 /* refill_period_us */,
10 /* fairness */));
for (int i = Env::IO_LOW; i <= Env::IO_TOTAL; ++i) {
ASSERT_EQ(limiter->GetTotalBytesThrough(static_cast<Env::IOPriority>(i)),
0);
}
std::int64_t request_byte = 10;
std::int64_t request_byte_sum = 0;
for (int i = Env::IO_LOW; i < Env::IO_TOTAL; ++i) {
limiter->Request(request_byte, static_cast<Env::IOPriority>(i),
nullptr /* stats */, RateLimiter::OpType::kWrite);
request_byte_sum += request_byte;
}
for (int i = Env::IO_LOW; i < Env::IO_TOTAL; ++i) {
EXPECT_EQ(limiter->GetTotalBytesThrough(static_cast<Env::IOPriority>(i)),
request_byte)
<< "Failed to track total_bytes_through_ correctly when IOPriority = "
<< static_cast<Env::IOPriority>(i);
}
EXPECT_EQ(limiter->GetTotalBytesThrough(Env::IO_TOTAL), request_byte_sum)
<< "Failed to track total_bytes_through_ correctly when IOPriority = "
"Env::IO_TOTAL";
}
TEST_F(RateLimiterTest, GetTotalRequests) {
std::unique_ptr<RateLimiter> limiter(NewGenericRateLimiter(
20 /* rate_bytes_per_sec */, 1000 * 1000 /* refill_period_us */,
10 /* fairness */));
for (int i = Env::IO_LOW; i <= Env::IO_TOTAL; ++i) {
ASSERT_EQ(limiter->GetTotalRequests(static_cast<Env::IOPriority>(i)), 0);
}
std::int64_t total_requests_sum = 0;
for (int i = Env::IO_LOW; i < Env::IO_TOTAL; ++i) {
limiter->Request(10, static_cast<Env::IOPriority>(i), nullptr /* stats */,
RateLimiter::OpType::kWrite);
total_requests_sum += 1;
}
for (int i = Env::IO_LOW; i < Env::IO_TOTAL; ++i) {
EXPECT_EQ(limiter->GetTotalRequests(static_cast<Env::IOPriority>(i)), 1)
<< "Failed to track total_requests_ correctly when IOPriority = "
<< static_cast<Env::IOPriority>(i);
}
EXPECT_EQ(limiter->GetTotalRequests(Env::IO_TOTAL), total_requests_sum)
<< "Failed to track total_requests_ correctly when IOPriority = "
"Env::IO_TOTAL";
}
TEST_F(RateLimiterTest, GetTotalPendingRequests) {
std::unique_ptr<RateLimiter> limiter(
NewGenericRateLimiter(20 /* rate_bytes_per_sec */));
for (int i = Env::IO_LOW; i <= Env::IO_TOTAL; ++i) {
ASSERT_EQ(limiter->GetTotalPendingRequests(static_cast<Env::IOPriority>(i)),
0);
}
// This is a variable for making sure the following callback is called
// and the assertions in it are indeed excuted
bool nonzero_pending_requests_verified_ = false;
SyncPoint::GetInstance()->SetCallBack(
"GenericRateLimiter::Request:PostEnqueueRequest", [&](void* arg) {
port::Mutex* request_mutex = (port::Mutex*)arg;
// We temporarily unlock the mutex so that the following
// GetTotalPendingRequests() can acquire it
request_mutex->Unlock();
EXPECT_EQ(limiter->GetTotalPendingRequests(Env::IO_USER), 1);
EXPECT_EQ(limiter->GetTotalPendingRequests(Env::IO_HIGH), 0);
EXPECT_EQ(limiter->GetTotalPendingRequests(Env::IO_MID), 0);
EXPECT_EQ(limiter->GetTotalPendingRequests(Env::IO_LOW), 0);
EXPECT_EQ(limiter->GetTotalPendingRequests(Env::IO_TOTAL), 1);
// We lock the mutex again so that the request thread can resume running
// with the mutex locked
request_mutex->Lock();
nonzero_pending_requests_verified_ = true;
});
SyncPoint::GetInstance()->EnableProcessing();
limiter->Request(20, Env::IO_USER, nullptr /* stats */,
RateLimiter::OpType::kWrite);
ASSERT_EQ(nonzero_pending_requests_verified_, true);
EXPECT_EQ(limiter->GetTotalPendingRequests(Env::IO_USER), 0);
EXPECT_EQ(limiter->GetTotalPendingRequests(Env::IO_HIGH), 0);
EXPECT_EQ(limiter->GetTotalPendingRequests(Env::IO_MID), 0);
EXPECT_EQ(limiter->GetTotalPendingRequests(Env::IO_LOW), 0);
EXPECT_EQ(limiter->GetTotalPendingRequests(Env::IO_TOTAL), 0);
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearCallBack(
"GenericRateLimiter::Request:PostEnqueueRequest");
}
TEST_F(RateLimiterTest, Modes) {
for (auto mode : {RateLimiter::Mode::kWritesOnly,
RateLimiter::Mode::kReadsOnly, RateLimiter::Mode::kAllIo}) {
GenericRateLimiter limiter(2000 /* rate_bytes_per_sec */,
1000 * 1000 /* refill_period_us */,
10 /* fairness */, mode, SystemClock::Default(),
false /* auto_tuned */);
limiter.Request(1000 /* bytes */, Env::IO_HIGH, nullptr /* stats */,
RateLimiter::OpType::kRead);
if (mode == RateLimiter::Mode::kWritesOnly) {
ASSERT_EQ(0, limiter.GetTotalBytesThrough(Env::IO_HIGH));
} else {
ASSERT_EQ(1000, limiter.GetTotalBytesThrough(Env::IO_HIGH));
}
limiter.Request(1000 /* bytes */, Env::IO_HIGH, nullptr /* stats */,
RateLimiter::OpType::kWrite);
if (mode == RateLimiter::Mode::kAllIo) {
ASSERT_EQ(2000, limiter.GetTotalBytesThrough(Env::IO_HIGH));
} else {
ASSERT_EQ(1000, limiter.GetTotalBytesThrough(Env::IO_HIGH));
}
}
}
Implement superior user & mid IO priority level in GenericRateLimiter (#8595) Summary: Context: An extra IO_USER priority in rate limiter allows users to optionally charge WAL writes / SST reads to rate limiter at this priority level, which then has higher priority than IO_HIGH and IO_LOW. With an extra IO_USER priority, it allows users to better specify the relative urgency/importance among different requests in rate limiter. As a consequence, IO resource management can better prioritize and limit resource based on user's need. The IO_USER is implemented as superior priority in GenericRateLimiter, in the sense that its request queue will always be iterated first without being constrained to fairness. The reason is that the notion of fairness is only meaningful in helping lower priorities in background IO (i.e, IO_HIGH/MID/LOW) to gain some fair chance to run so that it does not block foreground IO (i.e, the ones that are charged at the level of IO_USER). As we can see, the ultimate goal here is to not blocking foreground IO at IO_USER level, which justifies the superiority of IO_USER. Similar benefits exist for IO_MID priority. - Rewrote the logic of deciding the order of iterating request queues of high/low priorities to include the extra user/mid priority w/o affecting the existing behavior (see PR's [comment](https://github.com/facebook/rocksdb/pull/8595/files#r678749331)) - Included the request queue of user-pri/mid-pri in the code path of next-leader-candidate signaling and GenericRateLimiter's destructor - Included the extra user/mid-pri in bookkeeping data structures: total_bytes_through_ and total_requests_ - Re-written the previous impl of explicitly iterating priorities with a loop from Env::IO_LOW to Env::IO_TOTAL Pull Request resolved: https://github.com/facebook/rocksdb/pull/8595 Test Plan: - passed existing rate_limiter_test.cc - passed added unit tests in rate_limiter_test.cc - run performance test to verify performance with only high/low requests is not affected by this change - Set-up command: `TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=fillrandom --duration=5 --compression_type=none --num=100000000 --disable_auto_compactions=true --write_buffer_size=1048576 --writable_file_max_buffer_size=65536 --target_file_size_base=1048576 --max_bytes_for_level_base=4194304 --level0_slowdown_writes_trigger=$(((1 << 31) - 1)) --level0_stop_writes_trigger=$(((1 << 31) - 1))` - Test command: `TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=overwrite --use_existing_db=true --disable_wal=true --duration=30 --compression_type=none --num=100000000 --write_buffer_size=1048576 --writable_file_max_buffer_size=65536 --target_file_size_base=1048576 --max_bytes_for_level_base=4194304 --level0_slowdown_writes_trigger=$(((1 << 31) - 1)) --level0_stop_writes_trigger=$(((1 << 31) - 1)) --statistics=true --rate_limiter_bytes_per_sec=1048576 --rate_limiter_refill_period_us=1000 --threads=32 |& grep -E '(flush|compact)\.write\.bytes'` - Before (on branch upstream/master): `rocksdb.compact.write.bytes COUNT : 4014162` `rocksdb.flush.write.bytes COUNT : 26715832` rocksdb.flush.write.bytes/rocksdb.compact.write.bytes ~= 6.66 - After (on branch rate_limiter_user_pri): `rocksdb.compact.write.bytes COUNT : 3807822` `rocksdb.flush.write.bytes COUNT : 26098659` rocksdb.flush.write.bytes/rocksdb.compact.write.bytes ~= 6.85 Reviewed By: ajkr Differential Revision: D30577783 Pulled By: hx235 fbshipit-source-id: 0881f2705ffd13ecd331256bde7e8ec874a353f4
3 years ago
TEST_F(RateLimiterTest, GeneratePriorityIterationOrder) {
std::unique_ptr<RateLimiter> limiter(NewGenericRateLimiter(
20 /* rate_bytes_per_sec */, 1000 * 1000 /* refill_period_us */,
10 /* fairness */));
bool possible_random_one_in_fairness_results_for_high_mid_pri[4][2] = {
{false, false}, {false, true}, {true, false}, {true, true}};
std::vector<Env::IOPriority> possible_priority_iteration_orders[4] = {
{Env::IO_USER, Env::IO_HIGH, Env::IO_MID, Env::IO_LOW},
{Env::IO_USER, Env::IO_HIGH, Env::IO_LOW, Env::IO_MID},
{Env::IO_USER, Env::IO_MID, Env::IO_LOW, Env::IO_HIGH},
{Env::IO_USER, Env::IO_LOW, Env::IO_MID, Env::IO_HIGH}};
for (int i = 0; i < 4; ++i) {
SyncPoint::GetInstance()->SetCallBack(
"GenericRateLimiter::GeneratePriorityIterationOrder::"
"PostRandomOneInFairnessForHighPri",
[&](void* arg) {
bool* high_pri_iterated_after_mid_low_pri = (bool*)arg;
*high_pri_iterated_after_mid_low_pri =
possible_random_one_in_fairness_results_for_high_mid_pri[i][0];
});
SyncPoint::GetInstance()->SetCallBack(
"GenericRateLimiter::GeneratePriorityIterationOrder::"
"PostRandomOneInFairnessForMidPri",
[&](void* arg) {
bool* mid_pri_itereated_after_low_pri = (bool*)arg;
*mid_pri_itereated_after_low_pri =
possible_random_one_in_fairness_results_for_high_mid_pri[i][1];
});
SyncPoint::GetInstance()->SetCallBack(
"GenericRateLimiter::GeneratePriorityIterationOrder::"
"PreReturnPriIterationOrder",
[&](void* arg) {
std::vector<Env::IOPriority>* pri_iteration_order =
(std::vector<Env::IOPriority>*)arg;
EXPECT_EQ(*pri_iteration_order, possible_priority_iteration_orders[i])
<< "Failed to generate priority iteration order correctly when "
"high_pri_iterated_after_mid_low_pri = "
<< possible_random_one_in_fairness_results_for_high_mid_pri[i][0]
<< ", mid_pri_itereated_after_low_pri = "
<< possible_random_one_in_fairness_results_for_high_mid_pri[i][1]
<< std::endl;
});
SyncPoint::GetInstance()->EnableProcessing();
limiter->Request(20 /* request max bytes to drain so that refill and order
generation will be triggered every time
GenericRateLimiter::Request() is called */
,
Env::IO_USER, nullptr /* stats */,
RateLimiter::OpType::kWrite);
}
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearCallBack(
"GenericRateLimiter::GeneratePriorityIterationOrder::"
"PreReturnPriIterationOrder");
SyncPoint::GetInstance()->ClearCallBack(
"GenericRateLimiter::GeneratePriorityIterationOrder::"
"PostRandomOneInFairnessForMidPri");
SyncPoint::GetInstance()->ClearCallBack(
"GenericRateLimiter::GeneratePriorityIterationOrder::"
"PostRandomOneInFairnessForHighPri");
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
10 years ago
TEST_F(RateLimiterTest, Rate) {
auto* env = Env::Default();
struct Arg {
Arg(int32_t _target_rate, int _burst)
: limiter(NewGenericRateLimiter(_target_rate /* rate_bytes_per_sec */,
100 * 1000 /* refill_period_us */,
10 /* fairness */)),
request_size(_target_rate /
10 /* refill period here is 1/10 second */),
burst(_burst) {}
std::unique_ptr<RateLimiter> limiter;
int32_t request_size;
int burst;
};
auto writer = [](void* p) {
const auto& thread_clock = SystemClock::Default();
auto* arg = static_cast<Arg*>(p);
// Test for 2 seconds
auto until = thread_clock->NowMicros() + 2 * 1000000;
Random r((uint32_t)(thread_clock->NowNanos() %
std::numeric_limits<uint32_t>::max()));
while (thread_clock->NowMicros() < until) {
Implement superior user & mid IO priority level in GenericRateLimiter (#8595) Summary: Context: An extra IO_USER priority in rate limiter allows users to optionally charge WAL writes / SST reads to rate limiter at this priority level, which then has higher priority than IO_HIGH and IO_LOW. With an extra IO_USER priority, it allows users to better specify the relative urgency/importance among different requests in rate limiter. As a consequence, IO resource management can better prioritize and limit resource based on user's need. The IO_USER is implemented as superior priority in GenericRateLimiter, in the sense that its request queue will always be iterated first without being constrained to fairness. The reason is that the notion of fairness is only meaningful in helping lower priorities in background IO (i.e, IO_HIGH/MID/LOW) to gain some fair chance to run so that it does not block foreground IO (i.e, the ones that are charged at the level of IO_USER). As we can see, the ultimate goal here is to not blocking foreground IO at IO_USER level, which justifies the superiority of IO_USER. Similar benefits exist for IO_MID priority. - Rewrote the logic of deciding the order of iterating request queues of high/low priorities to include the extra user/mid priority w/o affecting the existing behavior (see PR's [comment](https://github.com/facebook/rocksdb/pull/8595/files#r678749331)) - Included the request queue of user-pri/mid-pri in the code path of next-leader-candidate signaling and GenericRateLimiter's destructor - Included the extra user/mid-pri in bookkeeping data structures: total_bytes_through_ and total_requests_ - Re-written the previous impl of explicitly iterating priorities with a loop from Env::IO_LOW to Env::IO_TOTAL Pull Request resolved: https://github.com/facebook/rocksdb/pull/8595 Test Plan: - passed existing rate_limiter_test.cc - passed added unit tests in rate_limiter_test.cc - run performance test to verify performance with only high/low requests is not affected by this change - Set-up command: `TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=fillrandom --duration=5 --compression_type=none --num=100000000 --disable_auto_compactions=true --write_buffer_size=1048576 --writable_file_max_buffer_size=65536 --target_file_size_base=1048576 --max_bytes_for_level_base=4194304 --level0_slowdown_writes_trigger=$(((1 << 31) - 1)) --level0_stop_writes_trigger=$(((1 << 31) - 1))` - Test command: `TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=overwrite --use_existing_db=true --disable_wal=true --duration=30 --compression_type=none --num=100000000 --write_buffer_size=1048576 --writable_file_max_buffer_size=65536 --target_file_size_base=1048576 --max_bytes_for_level_base=4194304 --level0_slowdown_writes_trigger=$(((1 << 31) - 1)) --level0_stop_writes_trigger=$(((1 << 31) - 1)) --statistics=true --rate_limiter_bytes_per_sec=1048576 --rate_limiter_refill_period_us=1000 --threads=32 |& grep -E '(flush|compact)\.write\.bytes'` - Before (on branch upstream/master): `rocksdb.compact.write.bytes COUNT : 4014162` `rocksdb.flush.write.bytes COUNT : 26715832` rocksdb.flush.write.bytes/rocksdb.compact.write.bytes ~= 6.66 - After (on branch rate_limiter_user_pri): `rocksdb.compact.write.bytes COUNT : 3807822` `rocksdb.flush.write.bytes COUNT : 26098659` rocksdb.flush.write.bytes/rocksdb.compact.write.bytes ~= 6.85 Reviewed By: ajkr Differential Revision: D30577783 Pulled By: hx235 fbshipit-source-id: 0881f2705ffd13ecd331256bde7e8ec874a353f4
3 years ago
for (int i = 0; i < static_cast<int>(r.Skewed(arg->burst * 2) + 1); ++i) {
arg->limiter->Request(r.Uniform(arg->request_size - 1) + 1,
Env::IO_USER, nullptr /* stats */,
RateLimiter::OpType::kWrite);
}
for (int i = 0; i < static_cast<int>(r.Skewed(arg->burst) + 1); ++i) {
arg->limiter->Request(r.Uniform(arg->request_size - 1) + 1,
Env::IO_HIGH, nullptr /* stats */,
RateLimiter::OpType::kWrite);
}
Implement superior user & mid IO priority level in GenericRateLimiter (#8595) Summary: Context: An extra IO_USER priority in rate limiter allows users to optionally charge WAL writes / SST reads to rate limiter at this priority level, which then has higher priority than IO_HIGH and IO_LOW. With an extra IO_USER priority, it allows users to better specify the relative urgency/importance among different requests in rate limiter. As a consequence, IO resource management can better prioritize and limit resource based on user's need. The IO_USER is implemented as superior priority in GenericRateLimiter, in the sense that its request queue will always be iterated first without being constrained to fairness. The reason is that the notion of fairness is only meaningful in helping lower priorities in background IO (i.e, IO_HIGH/MID/LOW) to gain some fair chance to run so that it does not block foreground IO (i.e, the ones that are charged at the level of IO_USER). As we can see, the ultimate goal here is to not blocking foreground IO at IO_USER level, which justifies the superiority of IO_USER. Similar benefits exist for IO_MID priority. - Rewrote the logic of deciding the order of iterating request queues of high/low priorities to include the extra user/mid priority w/o affecting the existing behavior (see PR's [comment](https://github.com/facebook/rocksdb/pull/8595/files#r678749331)) - Included the request queue of user-pri/mid-pri in the code path of next-leader-candidate signaling and GenericRateLimiter's destructor - Included the extra user/mid-pri in bookkeeping data structures: total_bytes_through_ and total_requests_ - Re-written the previous impl of explicitly iterating priorities with a loop from Env::IO_LOW to Env::IO_TOTAL Pull Request resolved: https://github.com/facebook/rocksdb/pull/8595 Test Plan: - passed existing rate_limiter_test.cc - passed added unit tests in rate_limiter_test.cc - run performance test to verify performance with only high/low requests is not affected by this change - Set-up command: `TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=fillrandom --duration=5 --compression_type=none --num=100000000 --disable_auto_compactions=true --write_buffer_size=1048576 --writable_file_max_buffer_size=65536 --target_file_size_base=1048576 --max_bytes_for_level_base=4194304 --level0_slowdown_writes_trigger=$(((1 << 31) - 1)) --level0_stop_writes_trigger=$(((1 << 31) - 1))` - Test command: `TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=overwrite --use_existing_db=true --disable_wal=true --duration=30 --compression_type=none --num=100000000 --write_buffer_size=1048576 --writable_file_max_buffer_size=65536 --target_file_size_base=1048576 --max_bytes_for_level_base=4194304 --level0_slowdown_writes_trigger=$(((1 << 31) - 1)) --level0_stop_writes_trigger=$(((1 << 31) - 1)) --statistics=true --rate_limiter_bytes_per_sec=1048576 --rate_limiter_refill_period_us=1000 --threads=32 |& grep -E '(flush|compact)\.write\.bytes'` - Before (on branch upstream/master): `rocksdb.compact.write.bytes COUNT : 4014162` `rocksdb.flush.write.bytes COUNT : 26715832` rocksdb.flush.write.bytes/rocksdb.compact.write.bytes ~= 6.66 - After (on branch rate_limiter_user_pri): `rocksdb.compact.write.bytes COUNT : 3807822` `rocksdb.flush.write.bytes COUNT : 26098659` rocksdb.flush.write.bytes/rocksdb.compact.write.bytes ~= 6.85 Reviewed By: ajkr Differential Revision: D30577783 Pulled By: hx235 fbshipit-source-id: 0881f2705ffd13ecd331256bde7e8ec874a353f4
3 years ago
for (int i = 0; i < static_cast<int>(r.Skewed(arg->burst / 2 + 1) + 1);
++i) {
arg->limiter->Request(r.Uniform(arg->request_size - 1) + 1, Env::IO_MID,
nullptr /* stats */, RateLimiter::OpType::kWrite);
}
arg->limiter->Request(r.Uniform(arg->request_size - 1) + 1, Env::IO_LOW,
nullptr /* stats */, RateLimiter::OpType::kWrite);
}
};
int samples = 0;
int samples_at_minimum = 0;
for (int i = 1; i <= 16; i *= 2) {
int32_t target = i * 1024 * 10;
Arg arg(target, i / 4 + 1);
int64_t old_total_bytes_through = 0;
for (int iter = 1; iter <= 2; ++iter) {
// second iteration changes the target dynamically
if (iter == 2) {
target *= 2;
arg.limiter->SetBytesPerSecond(target);
}
auto start = env->NowMicros();
for (int t = 0; t < i; ++t) {
env->StartThread(writer, &arg);
}
env->WaitForJoin();
auto elapsed = env->NowMicros() - start;
double rate =
(arg.limiter->GetTotalBytesThrough() - old_total_bytes_through) *
1000000.0 / elapsed;
old_total_bytes_through = arg.limiter->GetTotalBytesThrough();
fprintf(stderr,
"request size [1 - %" PRIi32 "], limit %" PRIi32
" KB/sec, actual rate: %lf KB/sec, elapsed %.2lf seconds\n",
arg.request_size - 1, target / 1024, rate / 1024,
elapsed / 1000000.0);
++samples;
if (rate / target >= 0.80) {
++samples_at_minimum;
}
ASSERT_LE(rate / target, 1.25);
}
}
// This can fail in heavily loaded CI environments
bool skip_minimum_rate_check =
#if (defined(TRAVIS) || defined(CIRCLECI)) && defined(OS_MACOSX)
true;
#else
getenv("SANDCASTLE");
#endif
if (skip_minimum_rate_check) {
fprintf(stderr, "Skipped minimum rate check (%d / %d passed)\n",
samples_at_minimum, samples);
} else {
ASSERT_EQ(samples_at_minimum, samples);
}
}
fix rate limiter to avoid starvation Summary: The current implementation of rate limiter has the possibility to introduce resource starvation when change its limit. This diff aims to fix this problem by consuming request bytes partially. Test Plan: ``` ./rate_limiter_test [==========] Running 4 tests from 1 test case. [----------] Global test environment set-up. [----------] 4 tests from RateLimiterTest [ RUN ] RateLimiterTest.OverflowRate [ OK ] RateLimiterTest.OverflowRate (0 ms) [ RUN ] RateLimiterTest.StartStop [ OK ] RateLimiterTest.StartStop (0 ms) [ RUN ] RateLimiterTest.Rate request size [1 - 1023], limit 10 KB/sec, actual rate: 10.355712 KB/sec, elapsed 2.00 seconds request size [1 - 1023], limit 20 KB/sec, actual rate: 19.136564 KB/sec, elapsed 2.00 seconds request size [1 - 2047], limit 20 KB/sec, actual rate: 20.783976 KB/sec, elapsed 2.10 seconds request size [1 - 2047], limit 40 KB/sec, actual rate: 39.308144 KB/sec, elapsed 2.10 seconds request size [1 - 4095], limit 40 KB/sec, actual rate: 40.318349 KB/sec, elapsed 2.20 seconds request size [1 - 4095], limit 80 KB/sec, actual rate: 79.667396 KB/sec, elapsed 2.20 seconds request size [1 - 8191], limit 80 KB/sec, actual rate: 81.807158 KB/sec, elapsed 2.30 seconds request size [1 - 8191], limit 160 KB/sec, actual rate: 160.659761 KB/sec, elapsed 2.20 seconds request size [1 - 16383], limit 160 KB/sec, actual rate: 160.700990 KB/sec, elapsed 3.00 seconds request size [1 - 16383], limit 320 KB/sec, actual rate: 317.639481 KB/sec, elapsed 2.50 seconds [ OK ] RateLimiterTest.Rate (22618 ms) [ RUN ] RateLimiterTest.LimitChangeTest [COMPLETE] request size 10 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 10 KB, new limit 5KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 10KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 80KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 160KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 320KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 80KB/sec, refill period 1000 ms [ OK ] RateLimiterTest.LimitChangeTest (5002 ms) [----------] 4 tests from RateLimiterTest (27620 ms total) [----------] Global test environment tear-down [==========] 4 tests from 1 test case ran. (27621 ms total) [ PASSED ] 4 tests. ``` Reviewers: sdong, IslamAbdelRahman, yiwu, andrewkr Reviewed By: andrewkr Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D60207
9 years ago
TEST_F(RateLimiterTest, LimitChangeTest) {
// starvation test when limit changes to a smaller value
int64_t refill_period = 1000 * 1000;
auto* env = Env::Default();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
fix rate limiter to avoid starvation Summary: The current implementation of rate limiter has the possibility to introduce resource starvation when change its limit. This diff aims to fix this problem by consuming request bytes partially. Test Plan: ``` ./rate_limiter_test [==========] Running 4 tests from 1 test case. [----------] Global test environment set-up. [----------] 4 tests from RateLimiterTest [ RUN ] RateLimiterTest.OverflowRate [ OK ] RateLimiterTest.OverflowRate (0 ms) [ RUN ] RateLimiterTest.StartStop [ OK ] RateLimiterTest.StartStop (0 ms) [ RUN ] RateLimiterTest.Rate request size [1 - 1023], limit 10 KB/sec, actual rate: 10.355712 KB/sec, elapsed 2.00 seconds request size [1 - 1023], limit 20 KB/sec, actual rate: 19.136564 KB/sec, elapsed 2.00 seconds request size [1 - 2047], limit 20 KB/sec, actual rate: 20.783976 KB/sec, elapsed 2.10 seconds request size [1 - 2047], limit 40 KB/sec, actual rate: 39.308144 KB/sec, elapsed 2.10 seconds request size [1 - 4095], limit 40 KB/sec, actual rate: 40.318349 KB/sec, elapsed 2.20 seconds request size [1 - 4095], limit 80 KB/sec, actual rate: 79.667396 KB/sec, elapsed 2.20 seconds request size [1 - 8191], limit 80 KB/sec, actual rate: 81.807158 KB/sec, elapsed 2.30 seconds request size [1 - 8191], limit 160 KB/sec, actual rate: 160.659761 KB/sec, elapsed 2.20 seconds request size [1 - 16383], limit 160 KB/sec, actual rate: 160.700990 KB/sec, elapsed 3.00 seconds request size [1 - 16383], limit 320 KB/sec, actual rate: 317.639481 KB/sec, elapsed 2.50 seconds [ OK ] RateLimiterTest.Rate (22618 ms) [ RUN ] RateLimiterTest.LimitChangeTest [COMPLETE] request size 10 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 10 KB, new limit 5KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 10KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 80KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 160KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 320KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 80KB/sec, refill period 1000 ms [ OK ] RateLimiterTest.LimitChangeTest (5002 ms) [----------] 4 tests from RateLimiterTest (27620 ms total) [----------] Global test environment tear-down [==========] 4 tests from 1 test case ran. (27621 ms total) [ PASSED ] 4 tests. ``` Reviewers: sdong, IslamAbdelRahman, yiwu, andrewkr Reviewed By: andrewkr Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D60207
9 years ago
struct Arg {
Arg(int32_t _request_size, Env::IOPriority _pri,
std::shared_ptr<RateLimiter> _limiter)
: request_size(_request_size), pri(_pri), limiter(_limiter) {}
int32_t request_size;
Env::IOPriority pri;
std::shared_ptr<RateLimiter> limiter;
};
auto writer = [](void* p) {
auto* arg = static_cast<Arg*>(p);
arg->limiter->Request(arg->request_size, arg->pri, nullptr /* stats */,
RateLimiter::OpType::kWrite);
fix rate limiter to avoid starvation Summary: The current implementation of rate limiter has the possibility to introduce resource starvation when change its limit. This diff aims to fix this problem by consuming request bytes partially. Test Plan: ``` ./rate_limiter_test [==========] Running 4 tests from 1 test case. [----------] Global test environment set-up. [----------] 4 tests from RateLimiterTest [ RUN ] RateLimiterTest.OverflowRate [ OK ] RateLimiterTest.OverflowRate (0 ms) [ RUN ] RateLimiterTest.StartStop [ OK ] RateLimiterTest.StartStop (0 ms) [ RUN ] RateLimiterTest.Rate request size [1 - 1023], limit 10 KB/sec, actual rate: 10.355712 KB/sec, elapsed 2.00 seconds request size [1 - 1023], limit 20 KB/sec, actual rate: 19.136564 KB/sec, elapsed 2.00 seconds request size [1 - 2047], limit 20 KB/sec, actual rate: 20.783976 KB/sec, elapsed 2.10 seconds request size [1 - 2047], limit 40 KB/sec, actual rate: 39.308144 KB/sec, elapsed 2.10 seconds request size [1 - 4095], limit 40 KB/sec, actual rate: 40.318349 KB/sec, elapsed 2.20 seconds request size [1 - 4095], limit 80 KB/sec, actual rate: 79.667396 KB/sec, elapsed 2.20 seconds request size [1 - 8191], limit 80 KB/sec, actual rate: 81.807158 KB/sec, elapsed 2.30 seconds request size [1 - 8191], limit 160 KB/sec, actual rate: 160.659761 KB/sec, elapsed 2.20 seconds request size [1 - 16383], limit 160 KB/sec, actual rate: 160.700990 KB/sec, elapsed 3.00 seconds request size [1 - 16383], limit 320 KB/sec, actual rate: 317.639481 KB/sec, elapsed 2.50 seconds [ OK ] RateLimiterTest.Rate (22618 ms) [ RUN ] RateLimiterTest.LimitChangeTest [COMPLETE] request size 10 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 10 KB, new limit 5KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 10KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 80KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 160KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 320KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 80KB/sec, refill period 1000 ms [ OK ] RateLimiterTest.LimitChangeTest (5002 ms) [----------] 4 tests from RateLimiterTest (27620 ms total) [----------] Global test environment tear-down [==========] 4 tests from 1 test case ran. (27621 ms total) [ PASSED ] 4 tests. ``` Reviewers: sdong, IslamAbdelRahman, yiwu, andrewkr Reviewed By: andrewkr Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D60207
9 years ago
};
for (uint32_t i = 1; i <= 16; i <<= 1) {
int32_t target = i * 1024 * 10;
// refill per second
for (int iter = 0; iter < 2; iter++) {
std::shared_ptr<RateLimiter> limiter =
std::make_shared<GenericRateLimiter>(
target, refill_period, 10, RateLimiter::Mode::kWritesOnly,
SystemClock::Default(), false /* auto_tuned */);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
fix rate limiter to avoid starvation Summary: The current implementation of rate limiter has the possibility to introduce resource starvation when change its limit. This diff aims to fix this problem by consuming request bytes partially. Test Plan: ``` ./rate_limiter_test [==========] Running 4 tests from 1 test case. [----------] Global test environment set-up. [----------] 4 tests from RateLimiterTest [ RUN ] RateLimiterTest.OverflowRate [ OK ] RateLimiterTest.OverflowRate (0 ms) [ RUN ] RateLimiterTest.StartStop [ OK ] RateLimiterTest.StartStop (0 ms) [ RUN ] RateLimiterTest.Rate request size [1 - 1023], limit 10 KB/sec, actual rate: 10.355712 KB/sec, elapsed 2.00 seconds request size [1 - 1023], limit 20 KB/sec, actual rate: 19.136564 KB/sec, elapsed 2.00 seconds request size [1 - 2047], limit 20 KB/sec, actual rate: 20.783976 KB/sec, elapsed 2.10 seconds request size [1 - 2047], limit 40 KB/sec, actual rate: 39.308144 KB/sec, elapsed 2.10 seconds request size [1 - 4095], limit 40 KB/sec, actual rate: 40.318349 KB/sec, elapsed 2.20 seconds request size [1 - 4095], limit 80 KB/sec, actual rate: 79.667396 KB/sec, elapsed 2.20 seconds request size [1 - 8191], limit 80 KB/sec, actual rate: 81.807158 KB/sec, elapsed 2.30 seconds request size [1 - 8191], limit 160 KB/sec, actual rate: 160.659761 KB/sec, elapsed 2.20 seconds request size [1 - 16383], limit 160 KB/sec, actual rate: 160.700990 KB/sec, elapsed 3.00 seconds request size [1 - 16383], limit 320 KB/sec, actual rate: 317.639481 KB/sec, elapsed 2.50 seconds [ OK ] RateLimiterTest.Rate (22618 ms) [ RUN ] RateLimiterTest.LimitChangeTest [COMPLETE] request size 10 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 10 KB, new limit 5KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 10KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 80KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 160KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 320KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 80KB/sec, refill period 1000 ms [ OK ] RateLimiterTest.LimitChangeTest (5002 ms) [----------] 4 tests from RateLimiterTest (27620 ms total) [----------] Global test environment tear-down [==========] 4 tests from 1 test case ran. (27621 ms total) [ PASSED ] 4 tests. ``` Reviewers: sdong, IslamAbdelRahman, yiwu, andrewkr Reviewed By: andrewkr Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D60207
9 years ago
{{"GenericRateLimiter::Request",
"RateLimiterTest::LimitChangeTest:changeLimitStart"},
{"RateLimiterTest::LimitChangeTest:changeLimitEnd",
"GenericRateLimiter::RefillBytesAndGrantRequests"}});
fix rate limiter to avoid starvation Summary: The current implementation of rate limiter has the possibility to introduce resource starvation when change its limit. This diff aims to fix this problem by consuming request bytes partially. Test Plan: ``` ./rate_limiter_test [==========] Running 4 tests from 1 test case. [----------] Global test environment set-up. [----------] 4 tests from RateLimiterTest [ RUN ] RateLimiterTest.OverflowRate [ OK ] RateLimiterTest.OverflowRate (0 ms) [ RUN ] RateLimiterTest.StartStop [ OK ] RateLimiterTest.StartStop (0 ms) [ RUN ] RateLimiterTest.Rate request size [1 - 1023], limit 10 KB/sec, actual rate: 10.355712 KB/sec, elapsed 2.00 seconds request size [1 - 1023], limit 20 KB/sec, actual rate: 19.136564 KB/sec, elapsed 2.00 seconds request size [1 - 2047], limit 20 KB/sec, actual rate: 20.783976 KB/sec, elapsed 2.10 seconds request size [1 - 2047], limit 40 KB/sec, actual rate: 39.308144 KB/sec, elapsed 2.10 seconds request size [1 - 4095], limit 40 KB/sec, actual rate: 40.318349 KB/sec, elapsed 2.20 seconds request size [1 - 4095], limit 80 KB/sec, actual rate: 79.667396 KB/sec, elapsed 2.20 seconds request size [1 - 8191], limit 80 KB/sec, actual rate: 81.807158 KB/sec, elapsed 2.30 seconds request size [1 - 8191], limit 160 KB/sec, actual rate: 160.659761 KB/sec, elapsed 2.20 seconds request size [1 - 16383], limit 160 KB/sec, actual rate: 160.700990 KB/sec, elapsed 3.00 seconds request size [1 - 16383], limit 320 KB/sec, actual rate: 317.639481 KB/sec, elapsed 2.50 seconds [ OK ] RateLimiterTest.Rate (22618 ms) [ RUN ] RateLimiterTest.LimitChangeTest [COMPLETE] request size 10 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 10 KB, new limit 5KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 20 KB, new limit 10KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 80KB/sec, refill period 1000 ms [COMPLETE] request size 40 KB, new limit 20KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 160KB/sec, refill period 1000 ms [COMPLETE] request size 80 KB, new limit 40KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 320KB/sec, refill period 1000 ms [COMPLETE] request size 160 KB, new limit 80KB/sec, refill period 1000 ms [ OK ] RateLimiterTest.LimitChangeTest (5002 ms) [----------] 4 tests from RateLimiterTest (27620 ms total) [----------] Global test environment tear-down [==========] 4 tests from 1 test case ran. (27621 ms total) [ PASSED ] 4 tests. ``` Reviewers: sdong, IslamAbdelRahman, yiwu, andrewkr Reviewed By: andrewkr Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D60207
9 years ago
Arg arg(target, Env::IO_HIGH, limiter);
// The idea behind is to start a request first, then before it refills,
// update limit to a different value (2X/0.5X). No starvation should
// be guaranteed under any situation
// TODO(lightmark): more test cases are welcome.
env->StartThread(writer, &arg);
int32_t new_limit = (target << 1) >> (iter << 1);
TEST_SYNC_POINT("RateLimiterTest::LimitChangeTest:changeLimitStart");
arg.limiter->SetBytesPerSecond(new_limit);
TEST_SYNC_POINT("RateLimiterTest::LimitChangeTest:changeLimitEnd");
env->WaitForJoin();
fprintf(stderr,
"[COMPLETE] request size %" PRIi32 " KB, new limit %" PRIi32
"KB/sec, refill period %" PRIi64 " ms\n",
target / 1024, new_limit / 1024, refill_period / 1000);
}
}
}
TEST_F(RateLimiterTest, AutoTuneIncreaseWhenFull) {
const std::chrono::seconds kTimePerRefill(1);
const int kRefillsPerTune = 100; // needs to match util/rate_limiter.cc
Fix+clean up handling of mock sleeps (#7101) Summary: We have a number of tests hanging on MacOS and windows due to mishandling of code for mock sleeps. In addition, the code was in terrible shape because the same variable (addon_time_) would sometimes refer to microseconds and sometimes to seconds. One test even assumed it was nanoseconds but was written to pass anyway. This has been cleaned up so that DB tests generally use a SpecialEnv function to mock sleep, for either some number of microseconds or seconds depending on the function called. But to call one of these, the test must first call SetMockSleep (precondition enforced with assertion), which also turns sleeps in RocksDB into mock sleeps. To also removes accounting for actual clock time, call SetTimeElapseOnlySleepOnReopen, which implies SetMockSleep (on DB re-open). This latter setting only works by applying on DB re-open, otherwise havoc can ensue if Env goes back in time with DB open. More specifics: Removed some unused test classes, and updated comments on the general problem. Fixed DBSSTTest.GetTotalSstFilesSize using a sync point callback instead of mock time. For this we have the only modification to production code, inserting a sync point callback in flush_job.cc, which is not a change to production behavior. Removed unnecessary resetting of mock times to 0 in many tests. RocksDB deals in relative time. Any behaviors relying on absolute date/time are likely a bug. (The above test DBSSTTest.GetTotalSstFilesSize was the only one clearly injecting a specific absolute time for actual testing convenience.) Just in case I misunderstood some test, I put this note in each replacement: // NOTE: Presumed unnecessary and removed: resetting mock time in env Strengthened some tests like MergeTestTime, MergeCompactionTimeTest, and FilterCompactionTimeTest in db_test.cc stats_history_test and blob_db_test are each their own beast, rather deeply dependent on MockTimeEnv. Each gets its own variant of a work-around for TimedWait in a mock time environment. (Reduces redundancy and inconsistency in stats_history_test.) Intended follow-up: Remove TimedWait from the public API of InstrumentedCondVar, and only make that accessible through Env by passing in an InstrumentedCondVar and a deadline. Then the Env implementations mocking time can fix this problem without using sync points. (Test infrastructure using sync points interferes with individual tests' control over sync points.) With that change, we can simplify/consolidate the scattered work-arounds. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7101 Test Plan: make check on Linux and MacOS Reviewed By: zhichao-cao Differential Revision: D23032815 Pulled By: pdillinger fbshipit-source-id: 7f33967ada8b83011fb54e8279365c008bd6610b
4 years ago
SpecialEnv special_env(Env::Default(), /*time_elapse_only_sleep*/ true);
auto stats = CreateDBStatistics();
std::unique_ptr<RateLimiter> rate_limiter(new GenericRateLimiter(
1000 /* rate_bytes_per_sec */,
std::chrono::microseconds(kTimePerRefill).count(), 10 /* fairness */,
RateLimiter::Mode::kWritesOnly, special_env.GetSystemClock(),
true /* auto_tuned */));
Simplify GenericRateLimiter algorithm (#8602) Summary: `GenericRateLimiter` slow path handles requests that cannot be satisfied immediately. Such requests enter a queue, and their thread stays in `Request()` until they are granted or the rate limiter is stopped. These threads are responsible for unblocking themselves. The work to do so is split into two main duties. (1) Waiting for the next refill time. (2) Refilling the bytes and granting requests. Prior to this PR, the slow path logic involved a leader election algorithm to pick one thread to perform (1) followed by (2). It elected the thread whose request was at the front of the highest priority non-empty queue since that request was most likely to be granted. This algorithm was efficient in terms of reducing intermediate wakeups, which is a thread waking up only to resume waiting after finding its request is not granted. However, the conceptual complexity of this algorithm was too high. It took me a long time to draw a timeline to understand how it works for just one edge case yet there were so many. This PR drops the leader election to reduce conceptual complexity. Now, the two duties can be performed by whichever thread acquires the lock first. The risk of this change is increasing the number of intermediate wakeups, however, we took steps to mitigate that. - `wait_until_refill_pending_` flag ensures only one thread performs (1). This\ prevents the thundering herd problem at the next refill time. The remaining\ threads wait on their condition variable with an unbounded duration -- thus we\ must remember to notify them to ensure forward progress. - (1) is typically done by a thread at the front of a queue. This is trivial\ when the queues are initially empty as the first choice that arrives must be\ the only entry in its queue. When queues are initially non-empty, we achieve\ this by having (2) notify a thread at the front of a queue (preferring higher\ priority) to perform the next duty. - We do not require any additional wakeup for (2). Typically it will just be\ done by the thread that finished (1). Combined, the second and third bullet points above suggest the refill/granting will typically be done by a request at the front of its queue. This is important because one wakeup is saved when a granted request happens to be in an already running thread. Note there are a few cases that still lead to intermediate wakeup, however. The first two are existing issues that also apply to the old algorithm, however, the third (including both subpoints) is new. - No request may be granted (only possible when rate limit dynamically\ decreases). - Requests from a different queue may be granted. - (2) may be run by a non-front request thread causing it to not be granted even\ if some requests in that same queue are granted. It can happen for a couple\ (unlikely) reasons. - A new request may sneak in and grab the lock at the refill time, before the\ thread finishing (1) can wake up and grab it. - A new request may sneak in and grab the lock and execute (1) before (2)'s\ chosen candidate can wake up and grab the lock. Then that non-front request\ thread performing (1) can carry over to perform (2). Pull Request resolved: https://github.com/facebook/rocksdb/pull/8602 Test Plan: - Use existing tests. The edge cases listed in the comment are all performance\ related; I could not really think of any related to correctness. The logic\ looks the same whether a thread wakes up/finishes its work early/on-time/late,\ or whether the thread is chosen vs. "steals" the work. - Verified write throughput and CPU overhead are basically the same with and\ without this change, even in a rate limiter heavy workload: Test command: ``` $ rm -rf /dev/shm/dbbench/ && TEST_TMPDIR=/dev/shm /usr/bin/time ./db_bench -benchmarks=fillrandom -num_multi_db=64 -num_low_pri_threads=64 -num_high_pri_threads=64 -write_buffer_size=262144 -target_file_size_base=262144 -max_bytes_for_level_base=1048576 -rate_limiter_bytes_per_sec=16777216 -key_size=24 -value_size=1000 -num=10000 -compression_type=none -rate_limiter_refill_period_us=1000 ``` Results before this PR: ``` fillrandom : 108.463 micros/op 9219 ops/sec; 9.0 MB/s 7.40user 8.84system 1:26.20elapsed 18%CPU (0avgtext+0avgdata 256140maxresident)k ``` Results after this PR: ``` fillrandom : 108.108 micros/op 9250 ops/sec; 9.0 MB/s 7.45user 8.23system 1:26.68elapsed 18%CPU (0avgtext+0avgdata 255688maxresident)k ``` Reviewed By: hx235 Differential Revision: D30048013 Pulled By: ajkr fbshipit-source-id: 6741bba9d9dfbccab359806d725105817fef818b
3 years ago
// Rate limiter uses `CondVar::TimedWait()`, which does not have access to the
// `Env` to advance its time according to the fake wait duration. The
// workaround is to install a callback that advance the `Env`'s mock time.
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
Simplify GenericRateLimiter algorithm (#8602) Summary: `GenericRateLimiter` slow path handles requests that cannot be satisfied immediately. Such requests enter a queue, and their thread stays in `Request()` until they are granted or the rate limiter is stopped. These threads are responsible for unblocking themselves. The work to do so is split into two main duties. (1) Waiting for the next refill time. (2) Refilling the bytes and granting requests. Prior to this PR, the slow path logic involved a leader election algorithm to pick one thread to perform (1) followed by (2). It elected the thread whose request was at the front of the highest priority non-empty queue since that request was most likely to be granted. This algorithm was efficient in terms of reducing intermediate wakeups, which is a thread waking up only to resume waiting after finding its request is not granted. However, the conceptual complexity of this algorithm was too high. It took me a long time to draw a timeline to understand how it works for just one edge case yet there were so many. This PR drops the leader election to reduce conceptual complexity. Now, the two duties can be performed by whichever thread acquires the lock first. The risk of this change is increasing the number of intermediate wakeups, however, we took steps to mitigate that. - `wait_until_refill_pending_` flag ensures only one thread performs (1). This\ prevents the thundering herd problem at the next refill time. The remaining\ threads wait on their condition variable with an unbounded duration -- thus we\ must remember to notify them to ensure forward progress. - (1) is typically done by a thread at the front of a queue. This is trivial\ when the queues are initially empty as the first choice that arrives must be\ the only entry in its queue. When queues are initially non-empty, we achieve\ this by having (2) notify a thread at the front of a queue (preferring higher\ priority) to perform the next duty. - We do not require any additional wakeup for (2). Typically it will just be\ done by the thread that finished (1). Combined, the second and third bullet points above suggest the refill/granting will typically be done by a request at the front of its queue. This is important because one wakeup is saved when a granted request happens to be in an already running thread. Note there are a few cases that still lead to intermediate wakeup, however. The first two are existing issues that also apply to the old algorithm, however, the third (including both subpoints) is new. - No request may be granted (only possible when rate limit dynamically\ decreases). - Requests from a different queue may be granted. - (2) may be run by a non-front request thread causing it to not be granted even\ if some requests in that same queue are granted. It can happen for a couple\ (unlikely) reasons. - A new request may sneak in and grab the lock at the refill time, before the\ thread finishing (1) can wake up and grab it. - A new request may sneak in and grab the lock and execute (1) before (2)'s\ chosen candidate can wake up and grab the lock. Then that non-front request\ thread performing (1) can carry over to perform (2). Pull Request resolved: https://github.com/facebook/rocksdb/pull/8602 Test Plan: - Use existing tests. The edge cases listed in the comment are all performance\ related; I could not really think of any related to correctness. The logic\ looks the same whether a thread wakes up/finishes its work early/on-time/late,\ or whether the thread is chosen vs. "steals" the work. - Verified write throughput and CPU overhead are basically the same with and\ without this change, even in a rate limiter heavy workload: Test command: ``` $ rm -rf /dev/shm/dbbench/ && TEST_TMPDIR=/dev/shm /usr/bin/time ./db_bench -benchmarks=fillrandom -num_multi_db=64 -num_low_pri_threads=64 -num_high_pri_threads=64 -write_buffer_size=262144 -target_file_size_base=262144 -max_bytes_for_level_base=1048576 -rate_limiter_bytes_per_sec=16777216 -key_size=24 -value_size=1000 -num=10000 -compression_type=none -rate_limiter_refill_period_us=1000 ``` Results before this PR: ``` fillrandom : 108.463 micros/op 9219 ops/sec; 9.0 MB/s 7.40user 8.84system 1:26.20elapsed 18%CPU (0avgtext+0avgdata 256140maxresident)k ``` Results after this PR: ``` fillrandom : 108.108 micros/op 9250 ops/sec; 9.0 MB/s 7.45user 8.23system 1:26.68elapsed 18%CPU (0avgtext+0avgdata 255688maxresident)k ``` Reviewed By: hx235 Differential Revision: D30048013 Pulled By: ajkr fbshipit-source-id: 6741bba9d9dfbccab359806d725105817fef818b
3 years ago
"GenericRateLimiter::Request:PostTimedWait", [&](void* arg) {
int64_t time_waited_us = *static_cast<int64_t*>(arg);
special_env.SleepForMicroseconds(static_cast<int>(time_waited_us));
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
// verify rate limit increases after a sequence of periods where rate limiter
// is always drained
int64_t orig_bytes_per_sec = rate_limiter->GetSingleBurstBytes();
rate_limiter->Request(orig_bytes_per_sec, Env::IO_HIGH, stats.get(),
RateLimiter::OpType::kWrite);
while (std::chrono::microseconds(special_env.NowMicros()) <=
kRefillsPerTune * kTimePerRefill) {
rate_limiter->Request(orig_bytes_per_sec, Env::IO_HIGH, stats.get(),
RateLimiter::OpType::kWrite);
}
int64_t new_bytes_per_sec = rate_limiter->GetSingleBurstBytes();
ASSERT_GT(new_bytes_per_sec, orig_bytes_per_sec);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
// decreases after a sequence of periods where rate limiter is not drained
orig_bytes_per_sec = new_bytes_per_sec;
special_env.SleepForMicroseconds(static_cast<int>(
kRefillsPerTune * std::chrono::microseconds(kTimePerRefill).count()));
// make a request so tuner can be triggered
rate_limiter->Request(1 /* bytes */, Env::IO_HIGH, stats.get(),
RateLimiter::OpType::kWrite);
new_bytes_per_sec = rate_limiter->GetSingleBurstBytes();
ASSERT_LT(new_bytes_per_sec, orig_bytes_per_sec);
}
} // namespace ROCKSDB_NAMESPACE
int main(int argc, char** argv) {
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
10 years ago
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}