2016-02-09 23:12:00 +00:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
2017-07-15 23:03:42 +00:00
|
|
|
// 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).
|
2014-06-18 23:36:48 +00:00
|
|
|
|
2019-05-30 21:47:29 +00:00
|
|
|
#include "table/plain/plain_table_key_coding.h"
|
2014-06-18 23:36:48 +00:00
|
|
|
|
2015-09-16 23:57:43 +00:00
|
|
|
#include <algorithm>
|
|
|
|
#include <string>
|
2022-10-25 18:50:38 +00:00
|
|
|
|
2014-06-18 23:36:48 +00:00
|
|
|
#include "db/dbformat.h"
|
2019-09-16 17:31:27 +00:00
|
|
|
#include "file/writable_file_writer.h"
|
2019-05-30 21:47:29 +00:00
|
|
|
#include "table/plain/plain_table_factory.h"
|
2019-05-31 00:39:43 +00:00
|
|
|
#include "table/plain/plain_table_reader.h"
|
2014-06-18 23:36:48 +00:00
|
|
|
|
2020-02-20 20:07:53 +00:00
|
|
|
namespace ROCKSDB_NAMESPACE {
|
2014-06-18 23:36:48 +00:00
|
|
|
|
Avoid naming conflict of EntryType
Summary:
Fix build break on travis build:
$ OPT=-DTRAVIS V=1 make unity && make clean && OPT=-DTRAVIS V=1 make db_test && ./db_test
......
In file included from unity.cc:65:0:
./table/plain_table_key_coding.cc: In member function ‘rocksdb::Status rocksdb::PlainTableKeyDecoder::NextPrefixEncodingKey(const char*, const char*, rocksdb::ParsedInternalKey*, rocksdb::Slice*, size_t*, bool*)’:
./table/plain_table_key_coding.cc:224:3: error: reference to ‘EntryType’ is ambiguous
EntryType entry_type;
^
In file included from ./db/table_properties_collector.h:9:0,
from ./db/builder.h:11,
from ./db/builder.cc:10,
from unity.cc:1:
./include/rocksdb/table_properties.h:81:6: note: candidates are: enum rocksdb::EntryType
enum EntryType {
^
In file included from unity.cc:65:0:
./table/plain_table_key_coding.cc:16:6: note: enum rocksdb::{anonymous}::EntryType
enum EntryType : unsigned char {
^
./table/plain_table_key_coding.cc:231:51: error: ‘entry_type’ was not declared in this scope
const char* pos = DecodeSize(key_ptr, limit, &entry_type, &size);
^
make: *** [unity.o] Error 1
Test Plan:
OPT=-DTRAVIS V=1 make unity
And make sure it doesn't break anymore.
Reviewers: yhchiang, kradhakrishnan, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D36549
2015-04-06 18:33:01 +00:00
|
|
|
enum PlainTableEntryType : unsigned char {
|
2014-06-18 23:36:48 +00:00
|
|
|
kFullKey = 0,
|
|
|
|
kPrefixFromPreviousKey = 1,
|
|
|
|
kKeySuffix = 2,
|
|
|
|
};
|
|
|
|
|
2015-09-16 23:57:43 +00:00
|
|
|
namespace {
|
|
|
|
|
2014-06-18 23:36:48 +00:00
|
|
|
// Control byte:
|
|
|
|
// First two bits indicate type of entry
|
|
|
|
// Other bytes are inlined sizes. If all bits are 1 (0x03F), overflow bytes
|
|
|
|
// are used. key_size-0x3F will be encoded as a variint32 after this bytes.
|
|
|
|
|
|
|
|
const unsigned char kSizeInlineLimit = 0x3F;
|
|
|
|
|
|
|
|
// Return 0 for error
|
Avoid naming conflict of EntryType
Summary:
Fix build break on travis build:
$ OPT=-DTRAVIS V=1 make unity && make clean && OPT=-DTRAVIS V=1 make db_test && ./db_test
......
In file included from unity.cc:65:0:
./table/plain_table_key_coding.cc: In member function ‘rocksdb::Status rocksdb::PlainTableKeyDecoder::NextPrefixEncodingKey(const char*, const char*, rocksdb::ParsedInternalKey*, rocksdb::Slice*, size_t*, bool*)’:
./table/plain_table_key_coding.cc:224:3: error: reference to ‘EntryType’ is ambiguous
EntryType entry_type;
^
In file included from ./db/table_properties_collector.h:9:0,
from ./db/builder.h:11,
from ./db/builder.cc:10,
from unity.cc:1:
./include/rocksdb/table_properties.h:81:6: note: candidates are: enum rocksdb::EntryType
enum EntryType {
^
In file included from unity.cc:65:0:
./table/plain_table_key_coding.cc:16:6: note: enum rocksdb::{anonymous}::EntryType
enum EntryType : unsigned char {
^
./table/plain_table_key_coding.cc:231:51: error: ‘entry_type’ was not declared in this scope
const char* pos = DecodeSize(key_ptr, limit, &entry_type, &size);
^
make: *** [unity.o] Error 1
Test Plan:
OPT=-DTRAVIS V=1 make unity
And make sure it doesn't break anymore.
Reviewers: yhchiang, kradhakrishnan, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D36549
2015-04-06 18:33:01 +00:00
|
|
|
size_t EncodeSize(PlainTableEntryType type, uint32_t key_size,
|
|
|
|
char* out_buffer) {
|
2014-06-18 23:36:48 +00:00
|
|
|
out_buffer[0] = type << 6;
|
|
|
|
|
2014-09-08 03:10:17 +00:00
|
|
|
if (key_size < static_cast<uint32_t>(kSizeInlineLimit)) {
|
2014-06-18 23:36:48 +00:00
|
|
|
// size inlined
|
|
|
|
out_buffer[0] |= static_cast<char>(key_size);
|
|
|
|
return 1;
|
|
|
|
} else {
|
|
|
|
out_buffer[0] |= kSizeInlineLimit;
|
|
|
|
char* ptr = EncodeVarint32(out_buffer + 1, key_size - kSizeInlineLimit);
|
|
|
|
return ptr - out_buffer;
|
|
|
|
}
|
|
|
|
}
|
2015-09-16 23:57:43 +00:00
|
|
|
} // namespace
|
2014-06-18 23:36:48 +00:00
|
|
|
|
2015-09-16 23:57:43 +00:00
|
|
|
// Fill bytes_read with number of bytes read.
|
|
|
|
inline Status PlainTableKeyDecoder::DecodeSize(uint32_t start_offset,
|
|
|
|
PlainTableEntryType* entry_type,
|
|
|
|
uint32_t* key_size,
|
|
|
|
uint32_t* bytes_read) {
|
|
|
|
Slice next_byte_slice;
|
|
|
|
bool success = file_reader_.Read(start_offset, 1, &next_byte_slice);
|
|
|
|
if (!success) {
|
|
|
|
return file_reader_.status();
|
|
|
|
}
|
Avoid naming conflict of EntryType
Summary:
Fix build break on travis build:
$ OPT=-DTRAVIS V=1 make unity && make clean && OPT=-DTRAVIS V=1 make db_test && ./db_test
......
In file included from unity.cc:65:0:
./table/plain_table_key_coding.cc: In member function ‘rocksdb::Status rocksdb::PlainTableKeyDecoder::NextPrefixEncodingKey(const char*, const char*, rocksdb::ParsedInternalKey*, rocksdb::Slice*, size_t*, bool*)’:
./table/plain_table_key_coding.cc:224:3: error: reference to ‘EntryType’ is ambiguous
EntryType entry_type;
^
In file included from ./db/table_properties_collector.h:9:0,
from ./db/builder.h:11,
from ./db/builder.cc:10,
from unity.cc:1:
./include/rocksdb/table_properties.h:81:6: note: candidates are: enum rocksdb::EntryType
enum EntryType {
^
In file included from unity.cc:65:0:
./table/plain_table_key_coding.cc:16:6: note: enum rocksdb::{anonymous}::EntryType
enum EntryType : unsigned char {
^
./table/plain_table_key_coding.cc:231:51: error: ‘entry_type’ was not declared in this scope
const char* pos = DecodeSize(key_ptr, limit, &entry_type, &size);
^
make: *** [unity.o] Error 1
Test Plan:
OPT=-DTRAVIS V=1 make unity
And make sure it doesn't break anymore.
Reviewers: yhchiang, kradhakrishnan, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D36549
2015-04-06 18:33:01 +00:00
|
|
|
*entry_type = static_cast<PlainTableEntryType>(
|
2015-09-16 23:57:43 +00:00
|
|
|
(static_cast<unsigned char>(next_byte_slice[0]) & ~kSizeInlineLimit) >>
|
|
|
|
6);
|
|
|
|
char inline_key_size = next_byte_slice[0] & kSizeInlineLimit;
|
2014-06-18 23:36:48 +00:00
|
|
|
if (inline_key_size < kSizeInlineLimit) {
|
|
|
|
*key_size = inline_key_size;
|
2015-09-16 23:57:43 +00:00
|
|
|
*bytes_read = 1;
|
|
|
|
return Status::OK();
|
2014-06-18 23:36:48 +00:00
|
|
|
} else {
|
|
|
|
uint32_t extra_size;
|
2015-09-16 23:57:43 +00:00
|
|
|
uint32_t tmp_bytes_read;
|
|
|
|
success = file_reader_.ReadVarint32(start_offset + 1, &extra_size,
|
|
|
|
&tmp_bytes_read);
|
|
|
|
if (!success) {
|
|
|
|
return file_reader_.status();
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
2015-09-16 23:57:43 +00:00
|
|
|
assert(tmp_bytes_read > 0);
|
2014-06-18 23:36:48 +00:00
|
|
|
*key_size = kSizeInlineLimit + extra_size;
|
2015-09-16 23:57:43 +00:00
|
|
|
*bytes_read = tmp_bytes_read + 1;
|
|
|
|
return Status::OK();
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
IOStatus PlainTableKeyEncoder::AppendKey(const Slice& key,
|
|
|
|
WritableFileWriter* file,
|
|
|
|
uint64_t* offset, char* meta_bytes_buf,
|
|
|
|
size_t* meta_bytes_buf_size) {
|
2014-06-18 23:36:48 +00:00
|
|
|
ParsedInternalKey parsed_key;
|
2020-10-28 17:11:13 +00:00
|
|
|
Status pik_status =
|
|
|
|
ParseInternalKey(key, &parsed_key, false /* log_err_key */); // TODO
|
|
|
|
if (!pik_status.ok()) {
|
|
|
|
return IOStatus::Corruption(pik_status.getState());
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
Slice key_to_write = key; // Portion of internal key to write out.
|
|
|
|
|
2015-02-23 22:10:09 +00:00
|
|
|
uint32_t user_key_size = static_cast<uint32_t>(key.size() - 8);
|
Group SST write in flush, compaction and db open with new stats (#11910)
Summary:
## Context/Summary
Similar to https://github.com/facebook/rocksdb/pull/11288, https://github.com/facebook/rocksdb/pull/11444, categorizing SST/blob file write according to different io activities allows more insight into the activity.
For that, this PR does the following:
- Tag different write IOs by passing down and converting WriteOptions to IOOptions
- Add new SST_WRITE_MICROS histogram in WritableFileWriter::Append() and breakdown FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS
Some related code refactory to make implementation cleaner:
- Blob stats
- Replace high-level write measurement with low-level WritableFileWriter::Append() measurement for BLOB_DB_BLOB_FILE_WRITE_MICROS. This is to make FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS include blob file. As a consequence, this introduces some behavioral changes on it, see HISTORY and db bench test plan below for more info.
- Fix bugs where BLOB_DB_BLOB_FILE_SYNCED/BLOB_DB_BLOB_FILE_BYTES_WRITTEN include file failed to sync and bytes failed to write.
- Refactor WriteOptions constructor for easier construction with io_activity and rate_limiter_priority
- Refactor DBImpl::~DBImpl()/BlobDBImpl::Close() to bypass thread op verification
- Build table
- TableBuilderOptions now includes Read/WriteOpitons so BuildTable() do not need to take these two variables
- Replace the io_priority passed into BuildTable() with TableBuilderOptions::WriteOpitons::rate_limiter_priority. Similar for BlobFileBuilder.
This parameter is used for dynamically changing file io priority for flush, see https://github.com/facebook/rocksdb/pull/9988?fbclid=IwAR1DtKel6c-bRJAdesGo0jsbztRtciByNlvokbxkV6h_L-AE9MACzqRTT5s for more
- Update ThreadStatus::FLUSH_BYTES_WRITTEN to use io_activity to track flush IO in flush job and db open instead of io_priority
## Test
### db bench
Flush
```
./db_bench --statistics=1 --benchmarks=fillseq --num=100000 --write_buffer_size=100
rocksdb.sst.write.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.flush.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.compaction.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.db.open.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
```
compaction, db oopen
```
Setup: ./db_bench --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
rocksdb.sst.write.micros P50 : 2.675325 P95 : 9.578788 P99 : 18.780000 P100 : 314.000000 COUNT : 638 SUM : 3279
rocksdb.file.write.flush.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.compaction.micros P50 : 2.757353 P95 : 9.610687 P99 : 19.316667 P100 : 314.000000 COUNT : 615 SUM : 3213
rocksdb.file.write.db.open.micros P50 : 2.055556 P95 : 3.925000 P99 : 9.000000 P100 : 9.000000 COUNT : 23 SUM : 66
```
blob stats - just to make sure they aren't broken by this PR
```
Integrated Blob DB
Setup: ./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 7.298246 P95 : 9.771930 P99 : 9.991813 P100 : 16.000000 COUNT : 235 SUM : 1600
rocksdb.blobdb.blob.file.synced COUNT : 1
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 2.000000 P95 : 2.829360 P99 : 2.993779 P100 : 9.000000 COUNT : 707 SUM : 1614
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 1 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842 (stay the same)
```
```
Stacked Blob DB
Run: ./db_bench --use_blob_db=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 12.808042 P95 : 19.674497 P99 : 28.539683 P100 : 51.000000 COUNT : 10000 SUM : 140876
rocksdb.blobdb.blob.file.synced COUNT : 8
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 1.657370 P95 : 2.952175 P99 : 3.877519 P100 : 24.000000 COUNT : 30001 SUM : 67924
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 8 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445 (stay the same)
```
### Rehearsal CI stress test
Trigger 3 full runs of all our CI stress tests
### Performance
Flush
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=ManualFlush/key_num:524288/per_key_size:256 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark; enable_statistics = true
Pre-pr: avg 507515519.3 ns
497686074,499444327,500862543,501389862,502994471,503744435,504142123,504224056,505724198,506610393,506837742,506955122,507695561,507929036,508307733,508312691,508999120,509963561,510142147,510698091,510743096,510769317,510957074,511053311,511371367,511409911,511432960,511642385,511691964,511730908,
Post-pr: avg 511971266.5 ns, regressed 0.88%
502744835,506502498,507735420,507929724,508313335,509548582,509994942,510107257,510715603,511046955,511352639,511458478,512117521,512317380,512766303,512972652,513059586,513804934,513808980,514059409,514187369,514389494,514447762,514616464,514622882,514641763,514666265,514716377,514990179,515502408,
```
Compaction
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_{pre|post}_pr --benchmark_filter=ManualCompaction/comp_style:0/max_data:134217728/per_key_size:256/enable_statistics:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 495346098.30 ns
492118301,493203526,494201411,494336607,495269217,495404950,496402598,497012157,497358370,498153846
Post-pr: avg 504528077.20, regressed 1.85%. "ManualCompaction" include flush so the isolated regression for compaction should be around 1.85-0.88 = 0.97%
502465338,502485945,502541789,502909283,503438601,504143885,506113087,506629423,507160414,507393007
```
Put with WAL (in case passing WriteOptions slows down this path even without collecting SST write stats)
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=DBPut/comp_style:0/max_data:107374182400/per_key_size:256/enable_statistics:1/wal:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 3848.10 ns
3814,3838,3839,3848,3854,3854,3854,3860,3860,3860
Post-pr: avg 3874.20 ns, regressed 0.68%
3863,3867,3871,3874,3875,3877,3877,3877,3880,3881
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11910
Reviewed By: ajkr
Differential Revision: D49788060
Pulled By: hx235
fbshipit-source-id: 79e73699cda5be3b66461687e5147c2484fc5eff
2023-12-29 23:29:23 +00:00
|
|
|
const IOOptions opts;
|
|
|
|
|
2014-06-18 23:36:48 +00:00
|
|
|
if (encoding_type_ == kPlain) {
|
|
|
|
if (fixed_user_key_len_ == kPlainTableVariableLength) {
|
|
|
|
// Write key length
|
|
|
|
char key_size_buf[5]; // tmp buffer for key size as varint32
|
|
|
|
char* ptr = EncodeVarint32(key_size_buf, user_key_size);
|
|
|
|
assert(ptr <= key_size_buf + sizeof(key_size_buf));
|
|
|
|
auto len = ptr - key_size_buf;
|
Group SST write in flush, compaction and db open with new stats (#11910)
Summary:
## Context/Summary
Similar to https://github.com/facebook/rocksdb/pull/11288, https://github.com/facebook/rocksdb/pull/11444, categorizing SST/blob file write according to different io activities allows more insight into the activity.
For that, this PR does the following:
- Tag different write IOs by passing down and converting WriteOptions to IOOptions
- Add new SST_WRITE_MICROS histogram in WritableFileWriter::Append() and breakdown FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS
Some related code refactory to make implementation cleaner:
- Blob stats
- Replace high-level write measurement with low-level WritableFileWriter::Append() measurement for BLOB_DB_BLOB_FILE_WRITE_MICROS. This is to make FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS include blob file. As a consequence, this introduces some behavioral changes on it, see HISTORY and db bench test plan below for more info.
- Fix bugs where BLOB_DB_BLOB_FILE_SYNCED/BLOB_DB_BLOB_FILE_BYTES_WRITTEN include file failed to sync and bytes failed to write.
- Refactor WriteOptions constructor for easier construction with io_activity and rate_limiter_priority
- Refactor DBImpl::~DBImpl()/BlobDBImpl::Close() to bypass thread op verification
- Build table
- TableBuilderOptions now includes Read/WriteOpitons so BuildTable() do not need to take these two variables
- Replace the io_priority passed into BuildTable() with TableBuilderOptions::WriteOpitons::rate_limiter_priority. Similar for BlobFileBuilder.
This parameter is used for dynamically changing file io priority for flush, see https://github.com/facebook/rocksdb/pull/9988?fbclid=IwAR1DtKel6c-bRJAdesGo0jsbztRtciByNlvokbxkV6h_L-AE9MACzqRTT5s for more
- Update ThreadStatus::FLUSH_BYTES_WRITTEN to use io_activity to track flush IO in flush job and db open instead of io_priority
## Test
### db bench
Flush
```
./db_bench --statistics=1 --benchmarks=fillseq --num=100000 --write_buffer_size=100
rocksdb.sst.write.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.flush.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.compaction.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.db.open.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
```
compaction, db oopen
```
Setup: ./db_bench --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
rocksdb.sst.write.micros P50 : 2.675325 P95 : 9.578788 P99 : 18.780000 P100 : 314.000000 COUNT : 638 SUM : 3279
rocksdb.file.write.flush.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.compaction.micros P50 : 2.757353 P95 : 9.610687 P99 : 19.316667 P100 : 314.000000 COUNT : 615 SUM : 3213
rocksdb.file.write.db.open.micros P50 : 2.055556 P95 : 3.925000 P99 : 9.000000 P100 : 9.000000 COUNT : 23 SUM : 66
```
blob stats - just to make sure they aren't broken by this PR
```
Integrated Blob DB
Setup: ./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 7.298246 P95 : 9.771930 P99 : 9.991813 P100 : 16.000000 COUNT : 235 SUM : 1600
rocksdb.blobdb.blob.file.synced COUNT : 1
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 2.000000 P95 : 2.829360 P99 : 2.993779 P100 : 9.000000 COUNT : 707 SUM : 1614
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 1 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842 (stay the same)
```
```
Stacked Blob DB
Run: ./db_bench --use_blob_db=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 12.808042 P95 : 19.674497 P99 : 28.539683 P100 : 51.000000 COUNT : 10000 SUM : 140876
rocksdb.blobdb.blob.file.synced COUNT : 8
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 1.657370 P95 : 2.952175 P99 : 3.877519 P100 : 24.000000 COUNT : 30001 SUM : 67924
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 8 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445 (stay the same)
```
### Rehearsal CI stress test
Trigger 3 full runs of all our CI stress tests
### Performance
Flush
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=ManualFlush/key_num:524288/per_key_size:256 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark; enable_statistics = true
Pre-pr: avg 507515519.3 ns
497686074,499444327,500862543,501389862,502994471,503744435,504142123,504224056,505724198,506610393,506837742,506955122,507695561,507929036,508307733,508312691,508999120,509963561,510142147,510698091,510743096,510769317,510957074,511053311,511371367,511409911,511432960,511642385,511691964,511730908,
Post-pr: avg 511971266.5 ns, regressed 0.88%
502744835,506502498,507735420,507929724,508313335,509548582,509994942,510107257,510715603,511046955,511352639,511458478,512117521,512317380,512766303,512972652,513059586,513804934,513808980,514059409,514187369,514389494,514447762,514616464,514622882,514641763,514666265,514716377,514990179,515502408,
```
Compaction
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_{pre|post}_pr --benchmark_filter=ManualCompaction/comp_style:0/max_data:134217728/per_key_size:256/enable_statistics:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 495346098.30 ns
492118301,493203526,494201411,494336607,495269217,495404950,496402598,497012157,497358370,498153846
Post-pr: avg 504528077.20, regressed 1.85%. "ManualCompaction" include flush so the isolated regression for compaction should be around 1.85-0.88 = 0.97%
502465338,502485945,502541789,502909283,503438601,504143885,506113087,506629423,507160414,507393007
```
Put with WAL (in case passing WriteOptions slows down this path even without collecting SST write stats)
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=DBPut/comp_style:0/max_data:107374182400/per_key_size:256/enable_statistics:1/wal:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 3848.10 ns
3814,3838,3839,3848,3854,3854,3854,3860,3860,3860
Post-pr: avg 3874.20 ns, regressed 0.68%
3863,3867,3871,3874,3875,3877,3877,3877,3880,3881
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11910
Reviewed By: ajkr
Differential Revision: D49788060
Pulled By: hx235
fbshipit-source-id: 79e73699cda5be3b66461687e5147c2484fc5eff
2023-12-29 23:29:23 +00:00
|
|
|
IOStatus io_s = file->Append(opts, Slice(key_size_buf, len));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
if (!io_s.ok()) {
|
|
|
|
return io_s;
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
*offset += len;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
assert(encoding_type_ == kPrefix);
|
|
|
|
char size_bytes[12];
|
|
|
|
size_t size_bytes_pos = 0;
|
|
|
|
|
|
|
|
Slice prefix =
|
|
|
|
prefix_extractor_->Transform(Slice(key.data(), user_key_size));
|
2017-04-04 21:17:16 +00:00
|
|
|
if (key_count_for_prefix_ == 0 || prefix != pre_prefix_.GetUserKey() ||
|
2014-09-06 15:21:26 +00:00
|
|
|
key_count_for_prefix_ % index_sparseness_ == 0) {
|
|
|
|
key_count_for_prefix_ = 1;
|
2017-04-04 21:17:16 +00:00
|
|
|
pre_prefix_.SetUserKey(prefix);
|
2014-06-18 23:36:48 +00:00
|
|
|
size_bytes_pos += EncodeSize(kFullKey, user_key_size, size_bytes);
|
Group SST write in flush, compaction and db open with new stats (#11910)
Summary:
## Context/Summary
Similar to https://github.com/facebook/rocksdb/pull/11288, https://github.com/facebook/rocksdb/pull/11444, categorizing SST/blob file write according to different io activities allows more insight into the activity.
For that, this PR does the following:
- Tag different write IOs by passing down and converting WriteOptions to IOOptions
- Add new SST_WRITE_MICROS histogram in WritableFileWriter::Append() and breakdown FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS
Some related code refactory to make implementation cleaner:
- Blob stats
- Replace high-level write measurement with low-level WritableFileWriter::Append() measurement for BLOB_DB_BLOB_FILE_WRITE_MICROS. This is to make FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS include blob file. As a consequence, this introduces some behavioral changes on it, see HISTORY and db bench test plan below for more info.
- Fix bugs where BLOB_DB_BLOB_FILE_SYNCED/BLOB_DB_BLOB_FILE_BYTES_WRITTEN include file failed to sync and bytes failed to write.
- Refactor WriteOptions constructor for easier construction with io_activity and rate_limiter_priority
- Refactor DBImpl::~DBImpl()/BlobDBImpl::Close() to bypass thread op verification
- Build table
- TableBuilderOptions now includes Read/WriteOpitons so BuildTable() do not need to take these two variables
- Replace the io_priority passed into BuildTable() with TableBuilderOptions::WriteOpitons::rate_limiter_priority. Similar for BlobFileBuilder.
This parameter is used for dynamically changing file io priority for flush, see https://github.com/facebook/rocksdb/pull/9988?fbclid=IwAR1DtKel6c-bRJAdesGo0jsbztRtciByNlvokbxkV6h_L-AE9MACzqRTT5s for more
- Update ThreadStatus::FLUSH_BYTES_WRITTEN to use io_activity to track flush IO in flush job and db open instead of io_priority
## Test
### db bench
Flush
```
./db_bench --statistics=1 --benchmarks=fillseq --num=100000 --write_buffer_size=100
rocksdb.sst.write.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.flush.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.compaction.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.db.open.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
```
compaction, db oopen
```
Setup: ./db_bench --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
rocksdb.sst.write.micros P50 : 2.675325 P95 : 9.578788 P99 : 18.780000 P100 : 314.000000 COUNT : 638 SUM : 3279
rocksdb.file.write.flush.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.compaction.micros P50 : 2.757353 P95 : 9.610687 P99 : 19.316667 P100 : 314.000000 COUNT : 615 SUM : 3213
rocksdb.file.write.db.open.micros P50 : 2.055556 P95 : 3.925000 P99 : 9.000000 P100 : 9.000000 COUNT : 23 SUM : 66
```
blob stats - just to make sure they aren't broken by this PR
```
Integrated Blob DB
Setup: ./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 7.298246 P95 : 9.771930 P99 : 9.991813 P100 : 16.000000 COUNT : 235 SUM : 1600
rocksdb.blobdb.blob.file.synced COUNT : 1
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 2.000000 P95 : 2.829360 P99 : 2.993779 P100 : 9.000000 COUNT : 707 SUM : 1614
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 1 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842 (stay the same)
```
```
Stacked Blob DB
Run: ./db_bench --use_blob_db=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 12.808042 P95 : 19.674497 P99 : 28.539683 P100 : 51.000000 COUNT : 10000 SUM : 140876
rocksdb.blobdb.blob.file.synced COUNT : 8
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 1.657370 P95 : 2.952175 P99 : 3.877519 P100 : 24.000000 COUNT : 30001 SUM : 67924
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 8 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445 (stay the same)
```
### Rehearsal CI stress test
Trigger 3 full runs of all our CI stress tests
### Performance
Flush
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=ManualFlush/key_num:524288/per_key_size:256 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark; enable_statistics = true
Pre-pr: avg 507515519.3 ns
497686074,499444327,500862543,501389862,502994471,503744435,504142123,504224056,505724198,506610393,506837742,506955122,507695561,507929036,508307733,508312691,508999120,509963561,510142147,510698091,510743096,510769317,510957074,511053311,511371367,511409911,511432960,511642385,511691964,511730908,
Post-pr: avg 511971266.5 ns, regressed 0.88%
502744835,506502498,507735420,507929724,508313335,509548582,509994942,510107257,510715603,511046955,511352639,511458478,512117521,512317380,512766303,512972652,513059586,513804934,513808980,514059409,514187369,514389494,514447762,514616464,514622882,514641763,514666265,514716377,514990179,515502408,
```
Compaction
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_{pre|post}_pr --benchmark_filter=ManualCompaction/comp_style:0/max_data:134217728/per_key_size:256/enable_statistics:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 495346098.30 ns
492118301,493203526,494201411,494336607,495269217,495404950,496402598,497012157,497358370,498153846
Post-pr: avg 504528077.20, regressed 1.85%. "ManualCompaction" include flush so the isolated regression for compaction should be around 1.85-0.88 = 0.97%
502465338,502485945,502541789,502909283,503438601,504143885,506113087,506629423,507160414,507393007
```
Put with WAL (in case passing WriteOptions slows down this path even without collecting SST write stats)
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=DBPut/comp_style:0/max_data:107374182400/per_key_size:256/enable_statistics:1/wal:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 3848.10 ns
3814,3838,3839,3848,3854,3854,3854,3860,3860,3860
Post-pr: avg 3874.20 ns, regressed 0.68%
3863,3867,3871,3874,3875,3877,3877,3877,3880,3881
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11910
Reviewed By: ajkr
Differential Revision: D49788060
Pulled By: hx235
fbshipit-source-id: 79e73699cda5be3b66461687e5147c2484fc5eff
2023-12-29 23:29:23 +00:00
|
|
|
IOStatus io_s = file->Append(opts, Slice(size_bytes, size_bytes_pos));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
if (!io_s.ok()) {
|
|
|
|
return io_s;
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
*offset += size_bytes_pos;
|
|
|
|
} else {
|
2014-09-06 15:21:26 +00:00
|
|
|
key_count_for_prefix_++;
|
|
|
|
if (key_count_for_prefix_ == 2) {
|
2014-06-18 23:36:48 +00:00
|
|
|
// For second key within a prefix, need to encode prefix length
|
|
|
|
size_bytes_pos +=
|
2014-11-11 21:47:22 +00:00
|
|
|
EncodeSize(kPrefixFromPreviousKey,
|
2017-04-04 21:17:16 +00:00
|
|
|
static_cast<uint32_t>(pre_prefix_.GetUserKey().size()),
|
2014-06-18 23:36:48 +00:00
|
|
|
size_bytes + size_bytes_pos);
|
|
|
|
}
|
2017-04-04 21:17:16 +00:00
|
|
|
uint32_t prefix_len =
|
|
|
|
static_cast<uint32_t>(pre_prefix_.GetUserKey().size());
|
2014-06-18 23:36:48 +00:00
|
|
|
size_bytes_pos += EncodeSize(kKeySuffix, user_key_size - prefix_len,
|
|
|
|
size_bytes + size_bytes_pos);
|
Group SST write in flush, compaction and db open with new stats (#11910)
Summary:
## Context/Summary
Similar to https://github.com/facebook/rocksdb/pull/11288, https://github.com/facebook/rocksdb/pull/11444, categorizing SST/blob file write according to different io activities allows more insight into the activity.
For that, this PR does the following:
- Tag different write IOs by passing down and converting WriteOptions to IOOptions
- Add new SST_WRITE_MICROS histogram in WritableFileWriter::Append() and breakdown FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS
Some related code refactory to make implementation cleaner:
- Blob stats
- Replace high-level write measurement with low-level WritableFileWriter::Append() measurement for BLOB_DB_BLOB_FILE_WRITE_MICROS. This is to make FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS include blob file. As a consequence, this introduces some behavioral changes on it, see HISTORY and db bench test plan below for more info.
- Fix bugs where BLOB_DB_BLOB_FILE_SYNCED/BLOB_DB_BLOB_FILE_BYTES_WRITTEN include file failed to sync and bytes failed to write.
- Refactor WriteOptions constructor for easier construction with io_activity and rate_limiter_priority
- Refactor DBImpl::~DBImpl()/BlobDBImpl::Close() to bypass thread op verification
- Build table
- TableBuilderOptions now includes Read/WriteOpitons so BuildTable() do not need to take these two variables
- Replace the io_priority passed into BuildTable() with TableBuilderOptions::WriteOpitons::rate_limiter_priority. Similar for BlobFileBuilder.
This parameter is used for dynamically changing file io priority for flush, see https://github.com/facebook/rocksdb/pull/9988?fbclid=IwAR1DtKel6c-bRJAdesGo0jsbztRtciByNlvokbxkV6h_L-AE9MACzqRTT5s for more
- Update ThreadStatus::FLUSH_BYTES_WRITTEN to use io_activity to track flush IO in flush job and db open instead of io_priority
## Test
### db bench
Flush
```
./db_bench --statistics=1 --benchmarks=fillseq --num=100000 --write_buffer_size=100
rocksdb.sst.write.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.flush.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.compaction.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.db.open.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
```
compaction, db oopen
```
Setup: ./db_bench --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
rocksdb.sst.write.micros P50 : 2.675325 P95 : 9.578788 P99 : 18.780000 P100 : 314.000000 COUNT : 638 SUM : 3279
rocksdb.file.write.flush.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.compaction.micros P50 : 2.757353 P95 : 9.610687 P99 : 19.316667 P100 : 314.000000 COUNT : 615 SUM : 3213
rocksdb.file.write.db.open.micros P50 : 2.055556 P95 : 3.925000 P99 : 9.000000 P100 : 9.000000 COUNT : 23 SUM : 66
```
blob stats - just to make sure they aren't broken by this PR
```
Integrated Blob DB
Setup: ./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 7.298246 P95 : 9.771930 P99 : 9.991813 P100 : 16.000000 COUNT : 235 SUM : 1600
rocksdb.blobdb.blob.file.synced COUNT : 1
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 2.000000 P95 : 2.829360 P99 : 2.993779 P100 : 9.000000 COUNT : 707 SUM : 1614
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 1 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842 (stay the same)
```
```
Stacked Blob DB
Run: ./db_bench --use_blob_db=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 12.808042 P95 : 19.674497 P99 : 28.539683 P100 : 51.000000 COUNT : 10000 SUM : 140876
rocksdb.blobdb.blob.file.synced COUNT : 8
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 1.657370 P95 : 2.952175 P99 : 3.877519 P100 : 24.000000 COUNT : 30001 SUM : 67924
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 8 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445 (stay the same)
```
### Rehearsal CI stress test
Trigger 3 full runs of all our CI stress tests
### Performance
Flush
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=ManualFlush/key_num:524288/per_key_size:256 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark; enable_statistics = true
Pre-pr: avg 507515519.3 ns
497686074,499444327,500862543,501389862,502994471,503744435,504142123,504224056,505724198,506610393,506837742,506955122,507695561,507929036,508307733,508312691,508999120,509963561,510142147,510698091,510743096,510769317,510957074,511053311,511371367,511409911,511432960,511642385,511691964,511730908,
Post-pr: avg 511971266.5 ns, regressed 0.88%
502744835,506502498,507735420,507929724,508313335,509548582,509994942,510107257,510715603,511046955,511352639,511458478,512117521,512317380,512766303,512972652,513059586,513804934,513808980,514059409,514187369,514389494,514447762,514616464,514622882,514641763,514666265,514716377,514990179,515502408,
```
Compaction
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_{pre|post}_pr --benchmark_filter=ManualCompaction/comp_style:0/max_data:134217728/per_key_size:256/enable_statistics:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 495346098.30 ns
492118301,493203526,494201411,494336607,495269217,495404950,496402598,497012157,497358370,498153846
Post-pr: avg 504528077.20, regressed 1.85%. "ManualCompaction" include flush so the isolated regression for compaction should be around 1.85-0.88 = 0.97%
502465338,502485945,502541789,502909283,503438601,504143885,506113087,506629423,507160414,507393007
```
Put with WAL (in case passing WriteOptions slows down this path even without collecting SST write stats)
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=DBPut/comp_style:0/max_data:107374182400/per_key_size:256/enable_statistics:1/wal:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 3848.10 ns
3814,3838,3839,3848,3854,3854,3854,3860,3860,3860
Post-pr: avg 3874.20 ns, regressed 0.68%
3863,3867,3871,3874,3875,3877,3877,3877,3880,3881
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11910
Reviewed By: ajkr
Differential Revision: D49788060
Pulled By: hx235
fbshipit-source-id: 79e73699cda5be3b66461687e5147c2484fc5eff
2023-12-29 23:29:23 +00:00
|
|
|
IOStatus io_s = file->Append(opts, Slice(size_bytes, size_bytes_pos));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
if (!io_s.ok()) {
|
|
|
|
return io_s;
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
*offset += size_bytes_pos;
|
|
|
|
key_to_write = Slice(key.data() + prefix_len, key.size() - prefix_len);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Encode full key
|
|
|
|
// For value size as varint32 (up to 5 bytes).
|
|
|
|
// If the row is of value type with seqId 0, flush the special flag together
|
|
|
|
// in this buffer to safe one file append call, which takes 1 byte.
|
|
|
|
if (parsed_key.sequence == 0 && parsed_key.type == kTypeValue) {
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
IOStatus io_s =
|
Group SST write in flush, compaction and db open with new stats (#11910)
Summary:
## Context/Summary
Similar to https://github.com/facebook/rocksdb/pull/11288, https://github.com/facebook/rocksdb/pull/11444, categorizing SST/blob file write according to different io activities allows more insight into the activity.
For that, this PR does the following:
- Tag different write IOs by passing down and converting WriteOptions to IOOptions
- Add new SST_WRITE_MICROS histogram in WritableFileWriter::Append() and breakdown FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS
Some related code refactory to make implementation cleaner:
- Blob stats
- Replace high-level write measurement with low-level WritableFileWriter::Append() measurement for BLOB_DB_BLOB_FILE_WRITE_MICROS. This is to make FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS include blob file. As a consequence, this introduces some behavioral changes on it, see HISTORY and db bench test plan below for more info.
- Fix bugs where BLOB_DB_BLOB_FILE_SYNCED/BLOB_DB_BLOB_FILE_BYTES_WRITTEN include file failed to sync and bytes failed to write.
- Refactor WriteOptions constructor for easier construction with io_activity and rate_limiter_priority
- Refactor DBImpl::~DBImpl()/BlobDBImpl::Close() to bypass thread op verification
- Build table
- TableBuilderOptions now includes Read/WriteOpitons so BuildTable() do not need to take these two variables
- Replace the io_priority passed into BuildTable() with TableBuilderOptions::WriteOpitons::rate_limiter_priority. Similar for BlobFileBuilder.
This parameter is used for dynamically changing file io priority for flush, see https://github.com/facebook/rocksdb/pull/9988?fbclid=IwAR1DtKel6c-bRJAdesGo0jsbztRtciByNlvokbxkV6h_L-AE9MACzqRTT5s for more
- Update ThreadStatus::FLUSH_BYTES_WRITTEN to use io_activity to track flush IO in flush job and db open instead of io_priority
## Test
### db bench
Flush
```
./db_bench --statistics=1 --benchmarks=fillseq --num=100000 --write_buffer_size=100
rocksdb.sst.write.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.flush.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.compaction.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.db.open.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
```
compaction, db oopen
```
Setup: ./db_bench --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
rocksdb.sst.write.micros P50 : 2.675325 P95 : 9.578788 P99 : 18.780000 P100 : 314.000000 COUNT : 638 SUM : 3279
rocksdb.file.write.flush.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.compaction.micros P50 : 2.757353 P95 : 9.610687 P99 : 19.316667 P100 : 314.000000 COUNT : 615 SUM : 3213
rocksdb.file.write.db.open.micros P50 : 2.055556 P95 : 3.925000 P99 : 9.000000 P100 : 9.000000 COUNT : 23 SUM : 66
```
blob stats - just to make sure they aren't broken by this PR
```
Integrated Blob DB
Setup: ./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 7.298246 P95 : 9.771930 P99 : 9.991813 P100 : 16.000000 COUNT : 235 SUM : 1600
rocksdb.blobdb.blob.file.synced COUNT : 1
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 2.000000 P95 : 2.829360 P99 : 2.993779 P100 : 9.000000 COUNT : 707 SUM : 1614
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 1 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842 (stay the same)
```
```
Stacked Blob DB
Run: ./db_bench --use_blob_db=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 12.808042 P95 : 19.674497 P99 : 28.539683 P100 : 51.000000 COUNT : 10000 SUM : 140876
rocksdb.blobdb.blob.file.synced COUNT : 8
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 1.657370 P95 : 2.952175 P99 : 3.877519 P100 : 24.000000 COUNT : 30001 SUM : 67924
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 8 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445 (stay the same)
```
### Rehearsal CI stress test
Trigger 3 full runs of all our CI stress tests
### Performance
Flush
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=ManualFlush/key_num:524288/per_key_size:256 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark; enable_statistics = true
Pre-pr: avg 507515519.3 ns
497686074,499444327,500862543,501389862,502994471,503744435,504142123,504224056,505724198,506610393,506837742,506955122,507695561,507929036,508307733,508312691,508999120,509963561,510142147,510698091,510743096,510769317,510957074,511053311,511371367,511409911,511432960,511642385,511691964,511730908,
Post-pr: avg 511971266.5 ns, regressed 0.88%
502744835,506502498,507735420,507929724,508313335,509548582,509994942,510107257,510715603,511046955,511352639,511458478,512117521,512317380,512766303,512972652,513059586,513804934,513808980,514059409,514187369,514389494,514447762,514616464,514622882,514641763,514666265,514716377,514990179,515502408,
```
Compaction
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_{pre|post}_pr --benchmark_filter=ManualCompaction/comp_style:0/max_data:134217728/per_key_size:256/enable_statistics:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 495346098.30 ns
492118301,493203526,494201411,494336607,495269217,495404950,496402598,497012157,497358370,498153846
Post-pr: avg 504528077.20, regressed 1.85%. "ManualCompaction" include flush so the isolated regression for compaction should be around 1.85-0.88 = 0.97%
502465338,502485945,502541789,502909283,503438601,504143885,506113087,506629423,507160414,507393007
```
Put with WAL (in case passing WriteOptions slows down this path even without collecting SST write stats)
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=DBPut/comp_style:0/max_data:107374182400/per_key_size:256/enable_statistics:1/wal:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 3848.10 ns
3814,3838,3839,3848,3854,3854,3854,3860,3860,3860
Post-pr: avg 3874.20 ns, regressed 0.68%
3863,3867,3871,3874,3875,3877,3877,3877,3880,3881
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11910
Reviewed By: ajkr
Differential Revision: D49788060
Pulled By: hx235
fbshipit-source-id: 79e73699cda5be3b66461687e5147c2484fc5eff
2023-12-29 23:29:23 +00:00
|
|
|
file->Append(opts, Slice(key_to_write.data(), key_to_write.size() - 8));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
if (!io_s.ok()) {
|
|
|
|
return io_s;
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
*offset += key_to_write.size() - 8;
|
|
|
|
meta_bytes_buf[*meta_bytes_buf_size] = PlainTableFactory::kValueTypeSeqId0;
|
|
|
|
*meta_bytes_buf_size += 1;
|
|
|
|
} else {
|
Group SST write in flush, compaction and db open with new stats (#11910)
Summary:
## Context/Summary
Similar to https://github.com/facebook/rocksdb/pull/11288, https://github.com/facebook/rocksdb/pull/11444, categorizing SST/blob file write according to different io activities allows more insight into the activity.
For that, this PR does the following:
- Tag different write IOs by passing down and converting WriteOptions to IOOptions
- Add new SST_WRITE_MICROS histogram in WritableFileWriter::Append() and breakdown FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS
Some related code refactory to make implementation cleaner:
- Blob stats
- Replace high-level write measurement with low-level WritableFileWriter::Append() measurement for BLOB_DB_BLOB_FILE_WRITE_MICROS. This is to make FILE_WRITE_{FLUSH|COMPACTION|DB_OPEN}_MICROS include blob file. As a consequence, this introduces some behavioral changes on it, see HISTORY and db bench test plan below for more info.
- Fix bugs where BLOB_DB_BLOB_FILE_SYNCED/BLOB_DB_BLOB_FILE_BYTES_WRITTEN include file failed to sync and bytes failed to write.
- Refactor WriteOptions constructor for easier construction with io_activity and rate_limiter_priority
- Refactor DBImpl::~DBImpl()/BlobDBImpl::Close() to bypass thread op verification
- Build table
- TableBuilderOptions now includes Read/WriteOpitons so BuildTable() do not need to take these two variables
- Replace the io_priority passed into BuildTable() with TableBuilderOptions::WriteOpitons::rate_limiter_priority. Similar for BlobFileBuilder.
This parameter is used for dynamically changing file io priority for flush, see https://github.com/facebook/rocksdb/pull/9988?fbclid=IwAR1DtKel6c-bRJAdesGo0jsbztRtciByNlvokbxkV6h_L-AE9MACzqRTT5s for more
- Update ThreadStatus::FLUSH_BYTES_WRITTEN to use io_activity to track flush IO in flush job and db open instead of io_priority
## Test
### db bench
Flush
```
./db_bench --statistics=1 --benchmarks=fillseq --num=100000 --write_buffer_size=100
rocksdb.sst.write.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.flush.micros P50 : 1.830863 P95 : 4.094720 P99 : 6.578947 P100 : 26.000000 COUNT : 7875 SUM : 20377
rocksdb.file.write.compaction.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.db.open.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
```
compaction, db oopen
```
Setup: ./db_bench --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
rocksdb.sst.write.micros P50 : 2.675325 P95 : 9.578788 P99 : 18.780000 P100 : 314.000000 COUNT : 638 SUM : 3279
rocksdb.file.write.flush.micros P50 : 0.000000 P95 : 0.000000 P99 : 0.000000 P100 : 0.000000 COUNT : 0 SUM : 0
rocksdb.file.write.compaction.micros P50 : 2.757353 P95 : 9.610687 P99 : 19.316667 P100 : 314.000000 COUNT : 615 SUM : 3213
rocksdb.file.write.db.open.micros P50 : 2.055556 P95 : 3.925000 P99 : 9.000000 P100 : 9.000000 COUNT : 23 SUM : 66
```
blob stats - just to make sure they aren't broken by this PR
```
Integrated Blob DB
Setup: ./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
Run:./db_bench --enable_blob_files=1 --statistics=1 --benchmarks=compact --db=../db_bench --use_existing_db=1
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 7.298246 P95 : 9.771930 P99 : 9.991813 P100 : 16.000000 COUNT : 235 SUM : 1600
rocksdb.blobdb.blob.file.synced COUNT : 1
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 2.000000 P95 : 2.829360 P99 : 2.993779 P100 : 9.000000 COUNT : 707 SUM : 1614
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 1 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 34842 (stay the same)
```
```
Stacked Blob DB
Run: ./db_bench --use_blob_db=1 --statistics=1 --benchmarks=fillseq --num=10000 --disable_auto_compactions=1 -write_buffer_size=100 --db=../db_bench
pre-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 12.808042 P95 : 19.674497 P99 : 28.539683 P100 : 51.000000 COUNT : 10000 SUM : 140876
rocksdb.blobdb.blob.file.synced COUNT : 8
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445
post-PR:
rocksdb.blobdb.blob.file.write.micros P50 : 1.657370 P95 : 2.952175 P99 : 3.877519 P100 : 24.000000 COUNT : 30001 SUM : 67924
- COUNT is higher and values are smaller as it includes header and footer write
- COUNT is 3X higher due to each Append() count as one post-PR, while in pre-PR, 3 Append()s counts as one. See https://github.com/facebook/rocksdb/pull/11910/files#diff-32b811c0a1c000768cfb2532052b44dc0b3bf82253f3eab078e15ff201a0dabfL157-L164
rocksdb.blobdb.blob.file.synced COUNT : 8 (stay the same)
rocksdb.blobdb.blob.file.bytes.written COUNT : 1043445 (stay the same)
```
### Rehearsal CI stress test
Trigger 3 full runs of all our CI stress tests
### Performance
Flush
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=ManualFlush/key_num:524288/per_key_size:256 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark; enable_statistics = true
Pre-pr: avg 507515519.3 ns
497686074,499444327,500862543,501389862,502994471,503744435,504142123,504224056,505724198,506610393,506837742,506955122,507695561,507929036,508307733,508312691,508999120,509963561,510142147,510698091,510743096,510769317,510957074,511053311,511371367,511409911,511432960,511642385,511691964,511730908,
Post-pr: avg 511971266.5 ns, regressed 0.88%
502744835,506502498,507735420,507929724,508313335,509548582,509994942,510107257,510715603,511046955,511352639,511458478,512117521,512317380,512766303,512972652,513059586,513804934,513808980,514059409,514187369,514389494,514447762,514616464,514622882,514641763,514666265,514716377,514990179,515502408,
```
Compaction
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_{pre|post}_pr --benchmark_filter=ManualCompaction/comp_style:0/max_data:134217728/per_key_size:256/enable_statistics:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 495346098.30 ns
492118301,493203526,494201411,494336607,495269217,495404950,496402598,497012157,497358370,498153846
Post-pr: avg 504528077.20, regressed 1.85%. "ManualCompaction" include flush so the isolated regression for compaction should be around 1.85-0.88 = 0.97%
502465338,502485945,502541789,502909283,503438601,504143885,506113087,506629423,507160414,507393007
```
Put with WAL (in case passing WriteOptions slows down this path even without collecting SST write stats)
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_pre_pr --benchmark_filter=DBPut/comp_style:0/max_data:107374182400/per_key_size:256/enable_statistics:1/wal:1 --benchmark_repetitions=1000
-- default: 1 thread is used to run benchmark
Pre-pr: avg 3848.10 ns
3814,3838,3839,3848,3854,3854,3854,3860,3860,3860
Post-pr: avg 3874.20 ns, regressed 0.68%
3863,3867,3871,3874,3875,3877,3877,3877,3880,3881
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11910
Reviewed By: ajkr
Differential Revision: D49788060
Pulled By: hx235
fbshipit-source-id: 79e73699cda5be3b66461687e5147c2484fc5eff
2023-12-29 23:29:23 +00:00
|
|
|
IOStatus io_s = file->Append(opts, key_to_write);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
if (!io_s.ok()) {
|
|
|
|
return io_s;
|
|
|
|
}
|
2014-06-18 23:36:48 +00:00
|
|
|
*offset += key_to_write.size();
|
|
|
|
}
|
|
|
|
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
return IOStatus::OK();
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
Slice PlainTableFileReader::GetFromBuffer(Buffer* buffer, uint32_t file_offset,
|
|
|
|
uint32_t len) {
|
|
|
|
assert(file_offset + len <= file_info_->data_end_offset);
|
|
|
|
return Slice(buffer->buf.get() + (file_offset - buffer->buf_start_offset),
|
|
|
|
len);
|
2015-09-16 23:57:43 +00:00
|
|
|
}
|
|
|
|
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
bool PlainTableFileReader::ReadNonMmap(uint32_t file_offset, uint32_t len,
|
|
|
|
Slice* out) {
|
2015-09-16 23:57:43 +00:00
|
|
|
const uint32_t kPrefetchSize = 256u;
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
|
|
|
|
// Try to read from buffers.
|
|
|
|
for (uint32_t i = 0; i < num_buf_; i++) {
|
|
|
|
Buffer* buffer = buffers_[num_buf_ - 1 - i].get();
|
|
|
|
if (file_offset >= buffer->buf_start_offset &&
|
|
|
|
file_offset + len <= buffer->buf_start_offset + buffer->buf_len) {
|
|
|
|
*out = GetFromBuffer(buffer, file_offset, len);
|
|
|
|
return true;
|
2015-09-16 23:57:43 +00:00
|
|
|
}
|
|
|
|
}
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
|
|
|
|
Buffer* new_buffer;
|
|
|
|
// Data needed is not in any of the buffer. Allocate a new buffer.
|
|
|
|
if (num_buf_ < buffers_.size()) {
|
|
|
|
// Add a new buffer
|
|
|
|
new_buffer = new Buffer();
|
|
|
|
buffers_[num_buf_++].reset(new_buffer);
|
|
|
|
} else {
|
|
|
|
// Now simply replace the last buffer. Can improve the placement policy
|
|
|
|
// if needed.
|
|
|
|
new_buffer = buffers_[num_buf_ - 1].get();
|
|
|
|
}
|
|
|
|
|
|
|
|
assert(file_offset + len <= file_info_->data_end_offset);
|
|
|
|
uint32_t size_to_read = std::min(file_info_->data_end_offset - file_offset,
|
|
|
|
std::max(kPrefetchSize, len));
|
|
|
|
if (size_to_read > new_buffer->buf_capacity) {
|
|
|
|
new_buffer->buf.reset(new char[size_to_read]);
|
|
|
|
new_buffer->buf_capacity = size_to_read;
|
|
|
|
new_buffer->buf_len = 0;
|
|
|
|
}
|
|
|
|
Slice read_result;
|
2022-02-17 07:17:03 +00:00
|
|
|
// TODO: rate limit plain table reads.
|
2020-04-30 21:48:51 +00:00
|
|
|
Status s =
|
|
|
|
file_info_->file->Read(IOOptions(), file_offset, size_to_read,
|
Group rocksdb.sst.read.micros stat by different user read IOActivity + misc (#11444)
Summary:
**Context/Summary:**
- Similar to https://github.com/facebook/rocksdb/pull/11288 but for user read such as `Get(), MultiGet(), DBIterator::XXX(), Verify(File)Checksum()`.
- For this, I refactored some user-facing `MultiGet` calls in `TransactionBase` and various types of `DB` so that it does not call a user-facing `Get()` but `GetImpl()` for passing the `ReadOptions::io_activity` check (see PR conversation)
- New user read stats breakdown are guarded by `kExceptDetailedTimers` since measurement shows they have 4-5% regression to the upstream/main.
- Misc
- More refactoring: with https://github.com/facebook/rocksdb/pull/11288, we complete passing `ReadOptions/IOOptions` to FS level. So we can now replace the previously [added](https://github.com/facebook/rocksdb/pull/9424) `rate_limiter_priority` parameter in `RandomAccessFileReader`'s `Read/MultiRead/Prefetch()` with `IOOptions::rate_limiter_priority`
- Also, `ReadAsync()` call time is measured in `SST_READ_MICRO` now
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11444
Test Plan:
- CI fake db crash/stress test
- Microbenchmarking
**Build** `make clean && ROCKSDB_NO_FBCODE=1 DEBUG_LEVEL=0 make -jN db_basic_bench`
- google benchmark version: https://github.com/google/benchmark/commit/604f6fd3f4b34a84ec4eb4db81d842fa4db829cd
- db_basic_bench_base: upstream
- db_basic_bench_pr: db_basic_bench_base + this PR
- asyncread_db_basic_bench_base: upstream + [db basic bench patch for IteratorNext](https://github.com/facebook/rocksdb/compare/main...hx235:rocksdb:micro_bench_async_read)
- asyncread_db_basic_bench_pr: asyncread_db_basic_bench_base + this PR
**Test**
Get
```
TEST_TMPDIR=/dev/shm ./db_basic_bench_{null_stat|base|pr} --benchmark_filter=DBGet/comp_style:0/max_data:134217728/per_key_size:256/enable_statistics:1/negative_query:0/enable_filter:0/mmap:1/threads:1 --benchmark_repetitions=1000
```
Result
```
Coming soon
```
AsyncRead
```
TEST_TMPDIR=/dev/shm ./asyncread_db_basic_bench_{base|pr} --benchmark_filter=IteratorNext/comp_style:0/max_data:134217728/per_key_size:256/enable_statistics:1/async_io:1/include_detailed_timers:0 --benchmark_repetitions=1000 > syncread_db_basic_bench_{base|pr}.out
```
Result
```
Base:
1956,1956,1968,1977,1979,1986,1988,1988,1988,1990,1991,1991,1993,1993,1993,1993,1994,1996,1997,1997,1997,1998,1999,2001,2001,2002,2004,2007,2007,2008,
PR (2.3% regression, due to measuring `SST_READ_MICRO` that wasn't measured before):
1993,2014,2016,2022,2024,2027,2027,2028,2028,2030,2031,2031,2032,2032,2038,2039,2042,2044,2044,2047,2047,2047,2048,2049,2050,2052,2052,2052,2053,2053,
```
Reviewed By: ajkr
Differential Revision: D45918925
Pulled By: hx235
fbshipit-source-id: 58a54560d9ebeb3a59b6d807639692614dad058a
2023-08-09 00:26:50 +00:00
|
|
|
&read_result, new_buffer->buf.get(), nullptr);
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
if (!s.ok()) {
|
|
|
|
status_ = s;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
new_buffer->buf_start_offset = file_offset;
|
|
|
|
new_buffer->buf_len = size_to_read;
|
|
|
|
*out = GetFromBuffer(new_buffer, file_offset, len);
|
2015-09-16 23:57:43 +00:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
inline bool PlainTableFileReader::ReadVarint32(uint32_t offset, uint32_t* out,
|
|
|
|
uint32_t* bytes_read) {
|
2015-09-16 23:57:43 +00:00
|
|
|
if (file_info_->is_mmap_mode) {
|
|
|
|
const char* start = file_info_->file_data.data() + offset;
|
|
|
|
const char* limit =
|
|
|
|
file_info_->file_data.data() + file_info_->data_end_offset;
|
|
|
|
const char* key_ptr = GetVarint32Ptr(start, limit, out);
|
|
|
|
assert(key_ptr != nullptr);
|
|
|
|
*bytes_read = static_cast<uint32_t>(key_ptr - start);
|
|
|
|
return true;
|
|
|
|
} else {
|
|
|
|
return ReadVarint32NonMmap(offset, out, bytes_read);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
bool PlainTableFileReader::ReadVarint32NonMmap(uint32_t offset, uint32_t* out,
|
|
|
|
uint32_t* bytes_read) {
|
2015-09-16 23:57:43 +00:00
|
|
|
const char* start;
|
|
|
|
const char* limit;
|
|
|
|
const uint32_t kMaxVarInt32Size = 6u;
|
|
|
|
uint32_t bytes_to_read =
|
|
|
|
std::min(file_info_->data_end_offset - offset, kMaxVarInt32Size);
|
|
|
|
Slice bytes;
|
|
|
|
if (!Read(offset, bytes_to_read, &bytes)) {
|
|
|
|
return false;
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
2015-09-16 23:57:43 +00:00
|
|
|
start = bytes.data();
|
|
|
|
limit = bytes.data() + bytes.size();
|
|
|
|
|
|
|
|
const char* key_ptr = GetVarint32Ptr(start, limit, out);
|
|
|
|
*bytes_read =
|
|
|
|
(key_ptr != nullptr) ? static_cast<uint32_t>(key_ptr - start) : 0;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status PlainTableKeyDecoder::ReadInternalKey(
|
|
|
|
uint32_t file_offset, uint32_t user_key_size, ParsedInternalKey* parsed_key,
|
|
|
|
uint32_t* bytes_read, bool* internal_key_valid, Slice* internal_key) {
|
|
|
|
Slice tmp_slice;
|
|
|
|
bool success = file_reader_.Read(file_offset, user_key_size + 1, &tmp_slice);
|
|
|
|
if (!success) {
|
|
|
|
return file_reader_.status();
|
|
|
|
}
|
|
|
|
if (tmp_slice[user_key_size] == PlainTableFactory::kValueTypeSeqId0) {
|
2014-06-18 23:36:48 +00:00
|
|
|
// Special encoding for the row with seqID=0
|
2015-09-16 23:57:43 +00:00
|
|
|
parsed_key->user_key = Slice(tmp_slice.data(), user_key_size);
|
2014-06-18 23:36:48 +00:00
|
|
|
parsed_key->sequence = 0;
|
|
|
|
parsed_key->type = kTypeValue;
|
|
|
|
*bytes_read += user_key_size + 1;
|
|
|
|
*internal_key_valid = false;
|
|
|
|
} else {
|
2015-09-16 23:57:43 +00:00
|
|
|
success = file_reader_.Read(file_offset, user_key_size + 8, internal_key);
|
|
|
|
if (!success) {
|
|
|
|
return file_reader_.status();
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
*internal_key_valid = true;
|
2020-10-28 17:11:13 +00:00
|
|
|
Status pik_status = ParseInternalKey(*internal_key, parsed_key,
|
|
|
|
false /* log_err_key */); // TODO
|
|
|
|
if (!pik_status.ok()) {
|
2014-06-18 23:36:48 +00:00
|
|
|
return Status::Corruption(
|
2020-10-28 17:11:13 +00:00
|
|
|
Slice("Corrupted key found during next key read. "),
|
|
|
|
pik_status.getState());
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
*bytes_read += user_key_size + 8;
|
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
2015-09-16 23:57:43 +00:00
|
|
|
Status PlainTableKeyDecoder::NextPlainEncodingKey(uint32_t start_offset,
|
|
|
|
ParsedInternalKey* parsed_key,
|
|
|
|
Slice* internal_key,
|
|
|
|
uint32_t* bytes_read,
|
2018-03-05 21:08:17 +00:00
|
|
|
bool* /*seekable*/) {
|
2014-11-11 21:47:22 +00:00
|
|
|
uint32_t user_key_size = 0;
|
2015-09-16 23:57:43 +00:00
|
|
|
Status s;
|
2014-06-18 23:36:48 +00:00
|
|
|
if (fixed_user_key_len_ != kPlainTableVariableLength) {
|
|
|
|
user_key_size = fixed_user_key_len_;
|
|
|
|
} else {
|
|
|
|
uint32_t tmp_size = 0;
|
2015-09-16 23:57:43 +00:00
|
|
|
uint32_t tmp_read;
|
|
|
|
bool success =
|
|
|
|
file_reader_.ReadVarint32(start_offset, &tmp_size, &tmp_read);
|
|
|
|
if (!success) {
|
|
|
|
return file_reader_.status();
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
2015-09-16 23:57:43 +00:00
|
|
|
assert(tmp_read > 0);
|
2014-11-11 21:47:22 +00:00
|
|
|
user_key_size = tmp_size;
|
2015-09-16 23:57:43 +00:00
|
|
|
*bytes_read = tmp_read;
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
2014-07-04 06:13:08 +00:00
|
|
|
// dummy initial value to avoid compiler complain
|
|
|
|
bool decoded_internal_key_valid = true;
|
2014-06-18 23:36:48 +00:00
|
|
|
Slice decoded_internal_key;
|
2015-09-16 23:57:43 +00:00
|
|
|
s = ReadInternalKey(start_offset + *bytes_read, user_key_size, parsed_key,
|
|
|
|
bytes_read, &decoded_internal_key_valid,
|
|
|
|
&decoded_internal_key);
|
2014-06-18 23:36:48 +00:00
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
if (!file_reader_.file_info()->is_mmap_mode) {
|
2015-09-16 23:57:43 +00:00
|
|
|
cur_key_.SetInternalKey(*parsed_key);
|
2017-04-04 21:17:16 +00:00
|
|
|
parsed_key->user_key =
|
|
|
|
Slice(cur_key_.GetInternalKey().data(), user_key_size);
|
2015-09-16 23:57:43 +00:00
|
|
|
if (internal_key != nullptr) {
|
2017-04-04 21:17:16 +00:00
|
|
|
*internal_key = cur_key_.GetInternalKey();
|
2015-09-16 23:57:43 +00:00
|
|
|
}
|
|
|
|
} else if (internal_key != nullptr) {
|
2014-06-18 23:36:48 +00:00
|
|
|
if (decoded_internal_key_valid) {
|
|
|
|
*internal_key = decoded_internal_key;
|
|
|
|
} else {
|
|
|
|
// Need to copy out the internal key
|
|
|
|
cur_key_.SetInternalKey(*parsed_key);
|
2017-04-04 21:17:16 +00:00
|
|
|
*internal_key = cur_key_.GetInternalKey();
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
|
|
|
Status PlainTableKeyDecoder::NextPrefixEncodingKey(
|
2015-09-16 23:57:43 +00:00
|
|
|
uint32_t start_offset, ParsedInternalKey* parsed_key, Slice* internal_key,
|
|
|
|
uint32_t* bytes_read, bool* seekable) {
|
Avoid naming conflict of EntryType
Summary:
Fix build break on travis build:
$ OPT=-DTRAVIS V=1 make unity && make clean && OPT=-DTRAVIS V=1 make db_test && ./db_test
......
In file included from unity.cc:65:0:
./table/plain_table_key_coding.cc: In member function ‘rocksdb::Status rocksdb::PlainTableKeyDecoder::NextPrefixEncodingKey(const char*, const char*, rocksdb::ParsedInternalKey*, rocksdb::Slice*, size_t*, bool*)’:
./table/plain_table_key_coding.cc:224:3: error: reference to ‘EntryType’ is ambiguous
EntryType entry_type;
^
In file included from ./db/table_properties_collector.h:9:0,
from ./db/builder.h:11,
from ./db/builder.cc:10,
from unity.cc:1:
./include/rocksdb/table_properties.h:81:6: note: candidates are: enum rocksdb::EntryType
enum EntryType {
^
In file included from unity.cc:65:0:
./table/plain_table_key_coding.cc:16:6: note: enum rocksdb::{anonymous}::EntryType
enum EntryType : unsigned char {
^
./table/plain_table_key_coding.cc:231:51: error: ‘entry_type’ was not declared in this scope
const char* pos = DecodeSize(key_ptr, limit, &entry_type, &size);
^
make: *** [unity.o] Error 1
Test Plan:
OPT=-DTRAVIS V=1 make unity
And make sure it doesn't break anymore.
Reviewers: yhchiang, kradhakrishnan, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D36549
2015-04-06 18:33:01 +00:00
|
|
|
PlainTableEntryType entry_type;
|
2014-06-18 23:36:48 +00:00
|
|
|
|
|
|
|
bool expect_suffix = false;
|
2015-09-16 23:57:43 +00:00
|
|
|
Status s;
|
2014-06-18 23:36:48 +00:00
|
|
|
do {
|
2014-11-11 21:47:22 +00:00
|
|
|
uint32_t size = 0;
|
2014-07-04 06:13:08 +00:00
|
|
|
// dummy initial value to avoid compiler complain
|
|
|
|
bool decoded_internal_key_valid = true;
|
2015-09-16 23:57:43 +00:00
|
|
|
uint32_t my_bytes_read = 0;
|
|
|
|
s = DecodeSize(start_offset + *bytes_read, &entry_type, &size,
|
|
|
|
&my_bytes_read);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
if (my_bytes_read == 0) {
|
2014-06-18 23:36:48 +00:00
|
|
|
return Status::Corruption("Unexpected EOF when reading size of the key");
|
|
|
|
}
|
2015-09-16 23:57:43 +00:00
|
|
|
*bytes_read += my_bytes_read;
|
2014-06-18 23:36:48 +00:00
|
|
|
|
|
|
|
switch (entry_type) {
|
|
|
|
case kFullKey: {
|
|
|
|
expect_suffix = false;
|
|
|
|
Slice decoded_internal_key;
|
2015-09-16 23:57:43 +00:00
|
|
|
s = ReadInternalKey(start_offset + *bytes_read, size, parsed_key,
|
|
|
|
bytes_read, &decoded_internal_key_valid,
|
|
|
|
&decoded_internal_key);
|
2014-06-18 23:36:48 +00:00
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
if (!file_reader_.file_info()->is_mmap_mode ||
|
2015-09-16 23:57:43 +00:00
|
|
|
(internal_key != nullptr && !decoded_internal_key_valid)) {
|
|
|
|
// In non-mmap mode, always need to make a copy of keys returned to
|
|
|
|
// users, because after reading value for the key, the key might
|
|
|
|
// be invalid.
|
|
|
|
cur_key_.SetInternalKey(*parsed_key);
|
2017-04-04 21:17:16 +00:00
|
|
|
saved_user_key_ = cur_key_.GetUserKey();
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
if (!file_reader_.file_info()->is_mmap_mode) {
|
2017-04-04 21:17:16 +00:00
|
|
|
parsed_key->user_key =
|
|
|
|
Slice(cur_key_.GetInternalKey().data(), size);
|
2015-09-16 23:57:43 +00:00
|
|
|
}
|
|
|
|
if (internal_key != nullptr) {
|
2017-04-04 21:17:16 +00:00
|
|
|
*internal_key = cur_key_.GetInternalKey();
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
2015-09-16 23:57:43 +00:00
|
|
|
} else {
|
|
|
|
if (internal_key != nullptr) {
|
|
|
|
*internal_key = decoded_internal_key;
|
|
|
|
}
|
|
|
|
saved_user_key_ = parsed_key->user_key;
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
case kPrefixFromPreviousKey: {
|
|
|
|
if (seekable != nullptr) {
|
|
|
|
*seekable = false;
|
|
|
|
}
|
|
|
|
prefix_len_ = size;
|
|
|
|
assert(prefix_extractor_ == nullptr ||
|
|
|
|
prefix_extractor_->Transform(saved_user_key_).size() ==
|
|
|
|
prefix_len_);
|
|
|
|
// Need read another size flag for suffix
|
|
|
|
expect_suffix = true;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
case kKeySuffix: {
|
|
|
|
expect_suffix = false;
|
|
|
|
if (seekable != nullptr) {
|
|
|
|
*seekable = false;
|
|
|
|
}
|
|
|
|
|
|
|
|
Slice tmp_slice;
|
2015-09-16 23:57:43 +00:00
|
|
|
s = ReadInternalKey(start_offset + *bytes_read, size, parsed_key,
|
|
|
|
bytes_read, &decoded_internal_key_valid,
|
|
|
|
&tmp_slice);
|
2014-06-18 23:36:48 +00:00
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
plain table reader: non-mmap mode to keep two recent buffers
Summary: In plain table reader's non-mmap mode, we only keep the most recent read buffer. However, for binary search, it is likely we come back to a location to read. To avoid one pread in such a case, we keep two read buffers. It should cover most of the cases.
Test Plan:
1. run tests
2. check the optimization works through strace when running
./table_reader_bench -mmap_read=false --num_keys2=1 -num_keys1=5000 -table_factory=plain_table --iterator --through_db
Reviewers: anthony, rven, kradhakrishnan, igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D51171
2015-12-24 01:30:10 +00:00
|
|
|
if (!file_reader_.file_info()->is_mmap_mode) {
|
2015-09-16 23:57:43 +00:00
|
|
|
// In non-mmap mode, we need to make a copy of keys returned to
|
|
|
|
// users, because after reading value for the key, the key might
|
|
|
|
// be invalid.
|
|
|
|
// saved_user_key_ points to cur_key_. We are making a copy of
|
|
|
|
// the prefix part to another string, and construct the current
|
|
|
|
// key from the prefix part and the suffix part back to cur_key_.
|
|
|
|
std::string tmp =
|
|
|
|
Slice(saved_user_key_.data(), prefix_len_).ToString();
|
|
|
|
cur_key_.Reserve(prefix_len_ + size);
|
|
|
|
cur_key_.SetInternalKey(tmp, *parsed_key);
|
|
|
|
parsed_key->user_key =
|
2017-04-04 21:17:16 +00:00
|
|
|
Slice(cur_key_.GetInternalKey().data(), prefix_len_ + size);
|
|
|
|
saved_user_key_ = cur_key_.GetUserKey();
|
2015-09-16 23:57:43 +00:00
|
|
|
} else {
|
|
|
|
cur_key_.Reserve(prefix_len_ + size);
|
|
|
|
cur_key_.SetInternalKey(Slice(saved_user_key_.data(), prefix_len_),
|
|
|
|
*parsed_key);
|
|
|
|
}
|
2017-04-04 21:17:16 +00:00
|
|
|
parsed_key->user_key = cur_key_.GetUserKey();
|
2014-06-18 23:36:48 +00:00
|
|
|
if (internal_key != nullptr) {
|
2017-04-04 21:17:16 +00:00
|
|
|
*internal_key = cur_key_.GetInternalKey();
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
default:
|
2015-09-16 23:57:43 +00:00
|
|
|
return Status::Corruption("Un-identified size flag.");
|
2014-06-18 23:36:48 +00:00
|
|
|
}
|
|
|
|
} while (expect_suffix); // Another round if suffix is expected.
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
2015-09-16 23:57:43 +00:00
|
|
|
Status PlainTableKeyDecoder::NextKey(uint32_t start_offset,
|
2014-06-18 23:36:48 +00:00
|
|
|
ParsedInternalKey* parsed_key,
|
2015-09-16 23:57:43 +00:00
|
|
|
Slice* internal_key, Slice* value,
|
|
|
|
uint32_t* bytes_read, bool* seekable) {
|
|
|
|
assert(value != nullptr);
|
|
|
|
Status s = NextKeyNoValue(start_offset, parsed_key, internal_key, bytes_read,
|
|
|
|
seekable);
|
|
|
|
if (s.ok()) {
|
|
|
|
assert(bytes_read != nullptr);
|
|
|
|
uint32_t value_size;
|
|
|
|
uint32_t value_size_bytes;
|
|
|
|
bool success = file_reader_.ReadVarint32(start_offset + *bytes_read,
|
|
|
|
&value_size, &value_size_bytes);
|
|
|
|
if (!success) {
|
|
|
|
return file_reader_.status();
|
|
|
|
}
|
|
|
|
if (value_size_bytes == 0) {
|
|
|
|
return Status::Corruption(
|
|
|
|
"Unexpected EOF when reading the next value's size.");
|
|
|
|
}
|
|
|
|
*bytes_read += value_size_bytes;
|
|
|
|
success = file_reader_.Read(start_offset + *bytes_read, value_size, value);
|
|
|
|
if (!success) {
|
|
|
|
return file_reader_.status();
|
|
|
|
}
|
|
|
|
*bytes_read += value_size;
|
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status PlainTableKeyDecoder::NextKeyNoValue(uint32_t start_offset,
|
|
|
|
ParsedInternalKey* parsed_key,
|
|
|
|
Slice* internal_key,
|
|
|
|
uint32_t* bytes_read,
|
|
|
|
bool* seekable) {
|
2014-06-18 23:36:48 +00:00
|
|
|
*bytes_read = 0;
|
|
|
|
if (seekable != nullptr) {
|
|
|
|
*seekable = true;
|
|
|
|
}
|
|
|
|
if (encoding_type_ == kPlain) {
|
2015-09-16 23:57:43 +00:00
|
|
|
return NextPlainEncodingKey(start_offset, parsed_key, internal_key,
|
2014-06-18 23:36:48 +00:00
|
|
|
bytes_read, seekable);
|
|
|
|
} else {
|
|
|
|
assert(encoding_type_ == kPrefix);
|
2015-09-16 23:57:43 +00:00
|
|
|
return NextPrefixEncodingKey(start_offset, parsed_key, internal_key,
|
2014-06-18 23:36:48 +00:00
|
|
|
bytes_read, seekable);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-02-20 20:07:53 +00:00
|
|
|
} // namespace ROCKSDB_NAMESPACE
|