1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
|
#include "ThreadPoolReader.h"
#include <Common/VersionNumber.h>
#include <Common/assert_cast.h>
#include <Common/Exception.h>
#include <Common/ProfileEvents.h>
#include <Common/CurrentMetrics.h>
#include <Common/Stopwatch.h>
#include <Common/setThreadName.h>
#include <Common/MemorySanitizer.h>
#include <Common/CurrentThread.h>
#include <Common/ThreadPool.h>
#include <Poco/Environment.h>
#include <base/errnoToString.h>
#include <Poco/Event.h>
#include <future>
#include <unistd.h>
#include <fcntl.h>
#if defined(OS_LINUX)
#include <sys/syscall.h>
#include <sys/uio.h>
/// We don't want to depend on specific glibc version.
#if !defined(RWF_NOWAIT)
#define RWF_NOWAIT 8
#endif
#if !defined(SYS_preadv2)
#if defined(__x86_64__)
#define SYS_preadv2 327
#elif defined(__aarch64__)
#define SYS_preadv2 286
#elif defined(__powerpc64__)
#define SYS_preadv2 380
#elif defined(__riscv)
#define SYS_preadv2 286
#else
#error "Unsupported architecture"
#endif
#endif
#endif
namespace ProfileEvents
{
extern const Event ThreadPoolReaderPageCacheHit;
extern const Event ThreadPoolReaderPageCacheHitBytes;
extern const Event ThreadPoolReaderPageCacheHitElapsedMicroseconds;
extern const Event ThreadPoolReaderPageCacheMiss;
extern const Event ThreadPoolReaderPageCacheMissBytes;
extern const Event ThreadPoolReaderPageCacheMissElapsedMicroseconds;
extern const Event ReadBufferFromFileDescriptorRead;
extern const Event ReadBufferFromFileDescriptorReadFailed;
extern const Event ReadBufferFromFileDescriptorReadBytes;
extern const Event DiskReadElapsedMicroseconds;
}
namespace CurrentMetrics
{
extern const Metric Read;
extern const Metric ThreadPoolFSReaderThreads;
extern const Metric ThreadPoolFSReaderThreadsActive;
}
namespace DB
{
namespace ErrorCodes
{
extern const int CANNOT_READ_FROM_FILE_DESCRIPTOR;
}
#if defined(OS_LINUX)
/// According to man, Linux 5.9 and 5.10 have a bug in preadv2() with the RWF_NOWAIT.
/// https://manpages.debian.org/testing/manpages-dev/preadv2.2.en.html#BUGS
/// We also disable it for older Linux kernels, because according to user's reports, RedHat-patched kernels might be also affected.
static bool hasBugInPreadV2()
{
VersionNumber linux_version(Poco::Environment::osVersion());
return linux_version < VersionNumber{5, 11, 0};
}
#endif
ThreadPoolReader::ThreadPoolReader(size_t pool_size, size_t queue_size_)
: pool(std::make_unique<ThreadPool>(CurrentMetrics::ThreadPoolFSReaderThreads, CurrentMetrics::ThreadPoolFSReaderThreadsActive, pool_size, pool_size, queue_size_))
{
}
std::future<IAsynchronousReader::Result> ThreadPoolReader::submit(Request request)
{
/// If size is zero, then read() cannot be distinguished from EOF
assert(request.size);
int fd = assert_cast<const LocalFileDescriptor &>(*request.descriptor).fd;
#if defined(OS_LINUX)
/// Check if data is already in page cache with preadv2 syscall.
/// We don't want to depend on new Linux kernel.
/// But kernels 5.9 and 5.10 have a bug where preadv2() with the
/// RWF_NOWAIT flag may return 0 even when not at end of file.
/// It can't be distinguished from the real eof, so we have to
/// disable pread with nowait.
static std::atomic<bool> has_pread_nowait_support = !hasBugInPreadV2();
if (has_pread_nowait_support.load(std::memory_order_relaxed))
{
/// It reports real time spent including the time spent while thread was preempted doing nothing.
/// And it is Ok for the purpose of this watch (it is used to lower the number of threads to read from tables).
/// Sometimes it is better to use taskstats::blkio_delay_total, but it is quite expensive to get it
/// (NetlinkMetricsProvider has about 500K RPS).
Stopwatch watch(CLOCK_MONOTONIC);
SCOPE_EXIT({
watch.stop();
ProfileEvents::increment(ProfileEvents::ThreadPoolReaderPageCacheHitElapsedMicroseconds, watch.elapsedMicroseconds());
ProfileEvents::increment(ProfileEvents::DiskReadElapsedMicroseconds, watch.elapsedMicroseconds());
});
std::promise<Result> promise;
std::future<Result> future = promise.get_future();
size_t bytes_read = 0;
while (!bytes_read)
{
ssize_t res = 0;
{
CurrentMetrics::Increment metric_increment{CurrentMetrics::Read};
struct iovec io_vec{ .iov_base = request.buf, .iov_len = request.size };
res = syscall(
SYS_preadv2, fd,
&io_vec, 1,
/// This is kind of weird calling convention for syscall.
static_cast<int64_t>(request.offset), static_cast<int64_t>(request.offset >> 32),
/// This flag forces read from page cache or returning EAGAIN.
RWF_NOWAIT);
}
if (!res)
{
/// The file has ended.
promise.set_value({0, 0, nullptr});
return future;
}
if (-1 == res)
{
if (errno == ENOSYS || errno == EOPNOTSUPP)
{
/// No support for the syscall or the flag in the Linux kernel.
has_pread_nowait_support.store(false, std::memory_order_relaxed);
break;
}
else if (errno == EAGAIN)
{
/// Data is not available in page cache. Will hand off to thread pool.
break;
}
else if (errno == EINTR)
{
/// Interrupted by a signal.
continue;
}
else
{
ProfileEvents::increment(ProfileEvents::ReadBufferFromFileDescriptorReadFailed);
promise.set_exception(std::make_exception_ptr(ErrnoException(
fmt::format("Cannot read from file {}, {}", fd, errnoToString()),
ErrorCodes::CANNOT_READ_FROM_FILE_DESCRIPTOR, errno)));
return future;
}
}
else
{
bytes_read += res;
__msan_unpoison(request.buf, res);
}
}
if (bytes_read)
{
/// Read successfully from page cache.
ProfileEvents::increment(ProfileEvents::ThreadPoolReaderPageCacheHit);
ProfileEvents::increment(ProfileEvents::ThreadPoolReaderPageCacheHitBytes, bytes_read);
ProfileEvents::increment(ProfileEvents::ReadBufferFromFileDescriptorReadBytes, bytes_read);
promise.set_value({bytes_read, request.ignore, nullptr});
return future;
}
}
#endif
ProfileEvents::increment(ProfileEvents::ThreadPoolReaderPageCacheMiss);
auto schedule = threadPoolCallbackRunner<Result>(*pool, "ThreadPoolRead");
return schedule([request, fd]() -> Result
{
Stopwatch watch(CLOCK_MONOTONIC);
SCOPE_EXIT({
watch.stop();
ProfileEvents::increment(ProfileEvents::ThreadPoolReaderPageCacheMissElapsedMicroseconds, watch.elapsedMicroseconds());
ProfileEvents::increment(ProfileEvents::DiskReadElapsedMicroseconds, watch.elapsedMicroseconds());
});
size_t bytes_read = 0;
while (!bytes_read)
{
ssize_t res = 0;
{
CurrentMetrics::Increment metric_increment{CurrentMetrics::Read};
res = ::pread(fd, request.buf, request.size, request.offset);
}
/// File has ended.
if (!res)
break;
if (-1 == res && errno != EINTR)
{
ProfileEvents::increment(ProfileEvents::ReadBufferFromFileDescriptorReadFailed);
throwFromErrno(fmt::format("Cannot read from file {}", fd), ErrorCodes::CANNOT_READ_FROM_FILE_DESCRIPTOR);
}
bytes_read += res;
}
watch.stop();
ProfileEvents::increment(ProfileEvents::ThreadPoolReaderPageCacheMissBytes, bytes_read);
ProfileEvents::increment(ProfileEvents::ReadBufferFromFileDescriptorReadBytes, bytes_read);
return Result{ .size = bytes_read, .offset = request.ignore };
}, request.priority);
}
void ThreadPoolReader::wait()
{
pool->wait();
}
}
|