1
by brian
clean slate |
1 |
/* Copyright (C) 2000-2006 MySQL AB & Sasha
|
2 |
||
3 |
This program is free software; you can redistribute it and/or modify
|
|
4 |
it under the terms of the GNU General Public License as published by
|
|
5 |
the Free Software Foundation; version 2 of the License.
|
|
6 |
||
7 |
This program is distributed in the hope that it will be useful,
|
|
8 |
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
9 |
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
10 |
GNU General Public License for more details.
|
|
11 |
||
12 |
You should have received a copy of the GNU General Public License
|
|
13 |
along with this program; if not, write to the Free Software
|
|
14 |
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */
|
|
15 |
||
16 |
#include "mysql_priv.h" |
|
17 |
#ifdef HAVE_REPLICATION
|
|
18 |
||
19 |
#include "rpl_mi.h" |
|
20 |
#include "sql_repl.h" |
|
21 |
#include "log_event.h" |
|
22 |
#include "rpl_filter.h" |
|
23 |
#include <my_dir.h> |
|
24 |
||
25 |
int max_binlog_dump_events = 0; // unlimited |
|
26 |
my_bool opt_sporadic_binlog_dump_fail = 0; |
|
27 |
#ifndef DBUG_OFF
|
|
28 |
static int binlog_dump_count = 0; |
|
29 |
#endif
|
|
30 |
||
31 |
/*
|
|
32 |
fake_rotate_event() builds a fake (=which does not exist physically in any
|
|
33 |
binlog) Rotate event, which contains the name of the binlog we are going to
|
|
34 |
send to the slave (because the slave may not know it if it just asked for
|
|
35 |
MASTER_LOG_FILE='', MASTER_LOG_POS=4).
|
|
36 |
< 4.0.14, fake_rotate_event() was called only if the requested pos was 4.
|
|
37 |
After this version we always call it, so that a 3.23.58 slave can rely on
|
|
38 |
it to detect if the master is 4.0 (and stop) (the _fake_ Rotate event has
|
|
39 |
zeros in the good positions which, by chance, make it possible for the 3.23
|
|
40 |
slave to detect that this event is unexpected) (this is luck which happens
|
|
41 |
because the master and slave disagree on the size of the header of
|
|
42 |
Log_event).
|
|
43 |
||
44 |
Relying on the event length of the Rotate event instead of these
|
|
45 |
well-placed zeros was not possible as Rotate events have a variable-length
|
|
46 |
part.
|
|
47 |
*/
|
|
48 |
||
49 |
static int fake_rotate_event(NET* net, String* packet, char* log_file_name, |
|
50 |
ulonglong position, const char** errmsg) |
|
51 |
{
|
|
52 |
DBUG_ENTER("fake_rotate_event"); |
|
53 |
char header[LOG_EVENT_HEADER_LEN], buf[ROTATE_HEADER_LEN+100]; |
|
54 |
/*
|
|
55 |
'when' (the timestamp) is set to 0 so that slave could distinguish between
|
|
56 |
real and fake Rotate events (if necessary)
|
|
57 |
*/
|
|
58 |
memset(header, 0, 4); |
|
59 |
header[EVENT_TYPE_OFFSET] = ROTATE_EVENT; |
|
60 |
||
61 |
char* p = log_file_name+dirname_length(log_file_name); |
|
62 |
uint ident_len = (uint) strlen(p); |
|
63 |
ulong event_len = ident_len + LOG_EVENT_HEADER_LEN + ROTATE_HEADER_LEN; |
|
64 |
int4store(header + SERVER_ID_OFFSET, server_id); |
|
65 |
int4store(header + EVENT_LEN_OFFSET, event_len); |
|
66 |
int2store(header + FLAGS_OFFSET, 0); |
|
67 |
||
68 |
// TODO: check what problems this may cause and fix them
|
|
69 |
int4store(header + LOG_POS_OFFSET, 0); |
|
70 |
||
71 |
packet->append(header, sizeof(header)); |
|
72 |
int8store(buf+R_POS_OFFSET,position); |
|
73 |
packet->append(buf, ROTATE_HEADER_LEN); |
|
74 |
packet->append(p,ident_len); |
|
75 |
if (my_net_write(net, (uchar*) packet->ptr(), packet->length())) |
|
76 |
{
|
|
77 |
*errmsg = "failed on my_net_write()"; |
|
78 |
DBUG_RETURN(-1); |
|
79 |
}
|
|
80 |
DBUG_RETURN(0); |
|
81 |
}
|
|
82 |
||
83 |
static int send_file(THD *thd) |
|
84 |
{
|
|
85 |
NET* net = &thd->net; |
|
86 |
int fd = -1, error = 1; |
|
87 |
size_t bytes; |
|
88 |
char fname[FN_REFLEN+1]; |
|
89 |
const char *errmsg = 0; |
|
90 |
int old_timeout; |
|
91 |
unsigned long packet_len; |
|
92 |
uchar buf[IO_SIZE]; // It's safe to alloc this |
|
93 |
DBUG_ENTER("send_file"); |
|
94 |
||
95 |
/*
|
|
96 |
The client might be slow loading the data, give him wait_timeout to do
|
|
97 |
the job
|
|
98 |
*/
|
|
99 |
old_timeout= net->read_timeout; |
|
100 |
my_net_set_read_timeout(net, thd->variables.net_wait_timeout); |
|
101 |
||
102 |
/*
|
|
103 |
We need net_flush here because the client will not know it needs to send
|
|
104 |
us the file name until it has processed the load event entry
|
|
105 |
*/
|
|
106 |
if (net_flush(net) || (packet_len = my_net_read(net)) == packet_error) |
|
107 |
{
|
|
108 |
errmsg = "while reading file name"; |
|
109 |
goto err; |
|
110 |
}
|
|
111 |
||
112 |
// terminate with \0 for fn_format
|
|
113 |
*((char*)net->read_pos + packet_len) = 0; |
|
114 |
fn_format(fname, (char*) net->read_pos + 1, "", "", 4); |
|
115 |
// this is needed to make replicate-ignore-db
|
|
116 |
if (!strcmp(fname,"/dev/null")) |
|
117 |
goto end; |
|
118 |
||
119 |
if ((fd = my_open(fname, O_RDONLY, MYF(0))) < 0) |
|
120 |
{
|
|
121 |
errmsg = "on open of file"; |
|
122 |
goto err; |
|
123 |
}
|
|
124 |
||
125 |
while ((long) (bytes= my_read(fd, buf, IO_SIZE, MYF(0))) > 0) |
|
126 |
{
|
|
127 |
if (my_net_write(net, buf, bytes)) |
|
128 |
{
|
|
129 |
errmsg = "while writing data to client"; |
|
130 |
goto err; |
|
131 |
}
|
|
132 |
}
|
|
133 |
||
134 |
end: |
|
135 |
if (my_net_write(net, (uchar*) "", 0) || net_flush(net) || |
|
136 |
(my_net_read(net) == packet_error)) |
|
137 |
{
|
|
138 |
errmsg = "while negotiating file transfer close"; |
|
139 |
goto err; |
|
140 |
}
|
|
141 |
error = 0; |
|
142 |
||
143 |
err: |
|
144 |
my_net_set_read_timeout(net, old_timeout); |
|
145 |
if (fd >= 0) |
|
146 |
(void) my_close(fd, MYF(0)); |
|
147 |
if (errmsg) |
|
148 |
{
|
|
149 |
sql_print_error("Failed in send_file() %s", errmsg); |
|
150 |
DBUG_PRINT("error", (errmsg)); |
|
151 |
}
|
|
152 |
DBUG_RETURN(error); |
|
153 |
}
|
|
154 |
||
155 |
||
156 |
/*
|
|
157 |
Adjust the position pointer in the binary log file for all running slaves
|
|
158 |
||
159 |
SYNOPSIS
|
|
160 |
adjust_linfo_offsets()
|
|
161 |
purge_offset Number of bytes removed from start of log index file
|
|
162 |
||
163 |
NOTES
|
|
164 |
- This is called when doing a PURGE when we delete lines from the
|
|
165 |
index log file
|
|
166 |
||
167 |
REQUIREMENTS
|
|
168 |
- Before calling this function, we have to ensure that no threads are
|
|
169 |
using any binary log file before purge_offset.a
|
|
170 |
||
171 |
TODO
|
|
172 |
- Inform the slave threads that they should sync the position
|
|
173 |
in the binary log file with flush_relay_log_info.
|
|
174 |
Now they sync is done for next read.
|
|
175 |
*/
|
|
176 |
||
177 |
void adjust_linfo_offsets(my_off_t purge_offset) |
|
178 |
{
|
|
179 |
THD *tmp; |
|
180 |
||
181 |
pthread_mutex_lock(&LOCK_thread_count); |
|
182 |
I_List_iterator<THD> it(threads); |
|
183 |
||
184 |
while ((tmp=it++)) |
|
185 |
{
|
|
186 |
LOG_INFO* linfo; |
|
187 |
if ((linfo = tmp->current_linfo)) |
|
188 |
{
|
|
189 |
pthread_mutex_lock(&linfo->lock); |
|
190 |
/*
|
|
191 |
Index file offset can be less that purge offset only if
|
|
192 |
we just started reading the index file. In that case
|
|
193 |
we have nothing to adjust
|
|
194 |
*/
|
|
195 |
if (linfo->index_file_offset < purge_offset) |
|
196 |
linfo->fatal = (linfo->index_file_offset != 0); |
|
197 |
else
|
|
198 |
linfo->index_file_offset -= purge_offset; |
|
199 |
pthread_mutex_unlock(&linfo->lock); |
|
200 |
}
|
|
201 |
}
|
|
202 |
pthread_mutex_unlock(&LOCK_thread_count); |
|
203 |
}
|
|
204 |
||
205 |
||
206 |
bool log_in_use(const char* log_name) |
|
207 |
{
|
|
208 |
int log_name_len = strlen(log_name) + 1; |
|
209 |
THD *tmp; |
|
210 |
bool result = 0; |
|
211 |
||
212 |
pthread_mutex_lock(&LOCK_thread_count); |
|
213 |
I_List_iterator<THD> it(threads); |
|
214 |
||
215 |
while ((tmp=it++)) |
|
216 |
{
|
|
217 |
LOG_INFO* linfo; |
|
218 |
if ((linfo = tmp->current_linfo)) |
|
219 |
{
|
|
220 |
pthread_mutex_lock(&linfo->lock); |
|
221 |
result = !bcmp((uchar*) log_name, (uchar*) linfo->log_file_name, |
|
222 |
log_name_len); |
|
223 |
pthread_mutex_unlock(&linfo->lock); |
|
224 |
if (result) |
|
225 |
break; |
|
226 |
}
|
|
227 |
}
|
|
228 |
||
229 |
pthread_mutex_unlock(&LOCK_thread_count); |
|
230 |
return result; |
|
231 |
}
|
|
232 |
||
233 |
bool purge_error_message(THD* thd, int res) |
|
234 |
{
|
|
235 |
uint errmsg= 0; |
|
236 |
||
237 |
switch (res) { |
|
238 |
case 0: break; |
|
239 |
case LOG_INFO_EOF: errmsg= ER_UNKNOWN_TARGET_BINLOG; break; |
|
240 |
case LOG_INFO_IO: errmsg= ER_IO_ERR_LOG_INDEX_READ; break; |
|
241 |
case LOG_INFO_INVALID:errmsg= ER_BINLOG_PURGE_PROHIBITED; break; |
|
242 |
case LOG_INFO_SEEK: errmsg= ER_FSEEK_FAIL; break; |
|
243 |
case LOG_INFO_MEM: errmsg= ER_OUT_OF_RESOURCES; break; |
|
244 |
case LOG_INFO_FATAL: errmsg= ER_BINLOG_PURGE_FATAL_ERR; break; |
|
245 |
case LOG_INFO_IN_USE: errmsg= ER_LOG_IN_USE; break; |
|
246 |
case LOG_INFO_EMFILE: errmsg= ER_BINLOG_PURGE_EMFILE; break; |
|
247 |
default: errmsg= ER_LOG_PURGE_UNKNOWN_ERR; break; |
|
248 |
}
|
|
249 |
||
250 |
if (errmsg) |
|
251 |
{
|
|
252 |
my_message(errmsg, ER(errmsg), MYF(0)); |
|
253 |
return TRUE; |
|
254 |
}
|
|
255 |
my_ok(thd); |
|
256 |
return FALSE; |
|
257 |
}
|
|
258 |
||
259 |
||
260 |
bool purge_master_logs(THD* thd, const char* to_log) |
|
261 |
{
|
|
262 |
char search_file_name[FN_REFLEN]; |
|
263 |
if (!mysql_bin_log.is_open()) |
|
264 |
{
|
|
265 |
my_ok(thd); |
|
266 |
return FALSE; |
|
267 |
}
|
|
268 |
||
269 |
mysql_bin_log.make_log_name(search_file_name, to_log); |
|
270 |
return purge_error_message(thd, |
|
271 |
mysql_bin_log.purge_logs(search_file_name, 0, 1, |
|
272 |
1, NULL)); |
|
273 |
}
|
|
274 |
||
275 |
||
276 |
bool purge_master_logs_before_date(THD* thd, time_t purge_time) |
|
277 |
{
|
|
278 |
if (!mysql_bin_log.is_open()) |
|
279 |
{
|
|
280 |
my_ok(thd); |
|
281 |
return 0; |
|
282 |
}
|
|
283 |
return purge_error_message(thd, |
|
284 |
mysql_bin_log.purge_logs_before_date(purge_time)); |
|
285 |
}
|
|
286 |
||
287 |
int test_for_non_eof_log_read_errors(int error, const char **errmsg) |
|
288 |
{
|
|
289 |
if (error == LOG_READ_EOF) |
|
290 |
return 0; |
|
291 |
my_errno= ER_MASTER_FATAL_ERROR_READING_BINLOG; |
|
292 |
switch (error) { |
|
293 |
case LOG_READ_BOGUS: |
|
294 |
*errmsg = "bogus data in log event"; |
|
295 |
break; |
|
296 |
case LOG_READ_TOO_LARGE: |
|
297 |
*errmsg = "log event entry exceeded max_allowed_packet; \ |
|
298 |
Increase max_allowed_packet on master"; |
|
299 |
break; |
|
300 |
case LOG_READ_IO: |
|
301 |
*errmsg = "I/O error reading log event"; |
|
302 |
break; |
|
303 |
case LOG_READ_MEM: |
|
304 |
*errmsg = "memory allocation failed reading log event"; |
|
305 |
break; |
|
306 |
case LOG_READ_TRUNC: |
|
307 |
*errmsg = "binlog truncated in the middle of event"; |
|
308 |
break; |
|
309 |
default: |
|
310 |
*errmsg = "unknown error reading log event on the master"; |
|
311 |
break; |
|
312 |
}
|
|
313 |
return error; |
|
314 |
}
|
|
315 |
||
316 |
||
317 |
/**
|
|
318 |
An auxiliary function for calling in mysql_binlog_send
|
|
319 |
to initialize the heartbeat timeout in waiting for a binlogged event.
|
|
320 |
||
321 |
@param[in] thd THD to access a user variable
|
|
322 |
||
323 |
@return heartbeat period an ulonglong of nanoseconds
|
|
324 |
or zero if heartbeat was not demanded by slave
|
|
325 |
*/
|
|
326 |
static ulonglong get_heartbeat_period(THD * thd) |
|
327 |
{
|
|
328 |
my_bool null_value; |
|
329 |
LEX_STRING name= { C_STRING_WITH_LEN("master_heartbeat_period")}; |
|
330 |
user_var_entry *entry= |
|
331 |
(user_var_entry*) hash_search(&thd->user_vars, (uchar*) name.str, |
|
332 |
name.length); |
|
333 |
return entry? entry->val_int(&null_value) : 0; |
|
334 |
}
|
|
335 |
||
336 |
/*
|
|
337 |
Function prepares and sends repliation heartbeat event.
|
|
338 |
||
339 |
@param net net object of THD
|
|
340 |
@param packet buffer to store the heartbeat instance
|
|
341 |
@param event_coordinates binlog file name and position of the last
|
|
342 |
real event master sent from binlog
|
|
343 |
||
344 |
@note
|
|
345 |
Among three essential pieces of heartbeat data Log_event::when
|
|
346 |
is computed locally.
|
|
347 |
The error to send is serious and should force terminating
|
|
348 |
the dump thread.
|
|
349 |
*/
|
|
350 |
static int send_heartbeat_event(NET* net, String* packet, |
|
351 |
const struct event_coordinates *coord) |
|
352 |
{
|
|
353 |
DBUG_ENTER("send_heartbeat_event"); |
|
354 |
char header[LOG_EVENT_HEADER_LEN]; |
|
355 |
/*
|
|
356 |
'when' (the timestamp) is set to 0 so that slave could distinguish between
|
|
357 |
real and fake Rotate events (if necessary)
|
|
358 |
*/
|
|
359 |
memset(header, 0, 4); // when |
|
360 |
||
361 |
header[EVENT_TYPE_OFFSET] = HEARTBEAT_LOG_EVENT; |
|
362 |
||
363 |
char* p= coord->file_name + dirname_length(coord->file_name); |
|
364 |
||
365 |
uint ident_len = strlen(p); |
|
366 |
ulong event_len = ident_len + LOG_EVENT_HEADER_LEN; |
|
367 |
int4store(header + SERVER_ID_OFFSET, server_id); |
|
368 |
int4store(header + EVENT_LEN_OFFSET, event_len); |
|
369 |
int2store(header + FLAGS_OFFSET, 0); |
|
370 |
||
371 |
int4store(header + LOG_POS_OFFSET, coord->pos); // log_pos |
|
372 |
||
373 |
packet->append(header, sizeof(header)); |
|
374 |
packet->append(p, ident_len); // log_file_name |
|
375 |
||
376 |
if (my_net_write(net, (uchar*) packet->ptr(), packet->length()) || |
|
377 |
net_flush(net)) |
|
378 |
{
|
|
379 |
DBUG_RETURN(-1); |
|
380 |
}
|
|
381 |
packet->set("\0", 1, &my_charset_bin); |
|
382 |
DBUG_RETURN(0); |
|
383 |
}
|
|
384 |
||
385 |
/*
|
|
386 |
TODO: Clean up loop to only have one call to send_file()
|
|
387 |
*/
|
|
388 |
||
389 |
void mysql_binlog_send(THD* thd, char* log_ident, my_off_t pos, |
|
390 |
ushort flags) |
|
391 |
{
|
|
392 |
LOG_INFO linfo; |
|
393 |
char *log_file_name = linfo.log_file_name; |
|
394 |
char search_file_name[FN_REFLEN], *name; |
|
395 |
IO_CACHE log; |
|
396 |
File file = -1; |
|
397 |
String* packet = &thd->packet; |
|
398 |
int error; |
|
399 |
const char *errmsg = "Unknown error"; |
|
400 |
NET* net = &thd->net; |
|
401 |
pthread_mutex_t *log_lock; |
|
402 |
bool binlog_can_be_corrupted= FALSE; |
|
403 |
#ifndef DBUG_OFF
|
|
404 |
int left_events = max_binlog_dump_events; |
|
405 |
#endif
|
|
406 |
DBUG_ENTER("mysql_binlog_send"); |
|
407 |
DBUG_PRINT("enter",("log_ident: '%s' pos: %ld", log_ident, (long) pos)); |
|
408 |
||
409 |
bzero((char*) &log,sizeof(log)); |
|
410 |
/*
|
|
411 |
heartbeat_period from @master_heartbeat_period user variable
|
|
412 |
*/
|
|
413 |
ulonglong heartbeat_period= get_heartbeat_period(thd); |
|
414 |
struct timespec heartbeat_buf; |
|
415 |
struct event_coordinates coord_buf; |
|
416 |
struct timespec *heartbeat_ts= NULL; |
|
417 |
struct event_coordinates *coord= NULL; |
|
80.1.1
by Brian Aker
LL() cleanup |
418 |
if (heartbeat_period != 0LL) |
1
by brian
clean slate |
419 |
{
|
420 |
heartbeat_ts= &heartbeat_buf; |
|
421 |
set_timespec_nsec(*heartbeat_ts, 0); |
|
422 |
coord= &coord_buf; |
|
423 |
coord->file_name= log_file_name; // initialization basing on what slave remembers |
|
424 |
coord->pos= pos; |
|
425 |
}
|
|
426 |
#ifndef DBUG_OFF
|
|
427 |
if (opt_sporadic_binlog_dump_fail && (binlog_dump_count++ % 2)) |
|
428 |
{
|
|
429 |
errmsg = "Master failed COM_BINLOG_DUMP to test if slave can recover"; |
|
430 |
my_errno= ER_UNKNOWN_ERROR; |
|
431 |
goto err; |
|
432 |
}
|
|
433 |
#endif
|
|
434 |
||
435 |
if (!mysql_bin_log.is_open()) |
|
436 |
{
|
|
437 |
errmsg = "Binary log is not open"; |
|
438 |
my_errno= ER_MASTER_FATAL_ERROR_READING_BINLOG; |
|
439 |
goto err; |
|
440 |
}
|
|
441 |
if (!server_id_supplied) |
|
442 |
{
|
|
443 |
errmsg = "Misconfigured master - server id was not set"; |
|
444 |
my_errno= ER_MASTER_FATAL_ERROR_READING_BINLOG; |
|
445 |
goto err; |
|
446 |
}
|
|
447 |
||
448 |
name=search_file_name; |
|
449 |
if (log_ident[0]) |
|
450 |
mysql_bin_log.make_log_name(search_file_name, log_ident); |
|
451 |
else
|
|
452 |
name=0; // Find first log |
|
453 |
||
454 |
linfo.index_file_offset = 0; |
|
455 |
||
456 |
if (mysql_bin_log.find_log_pos(&linfo, name, 1)) |
|
457 |
{
|
|
458 |
errmsg = "Could not find first log file name in binary log index file"; |
|
459 |
my_errno= ER_MASTER_FATAL_ERROR_READING_BINLOG; |
|
460 |
goto err; |
|
461 |
}
|
|
462 |
||
463 |
pthread_mutex_lock(&LOCK_thread_count); |
|
464 |
thd->current_linfo = &linfo; |
|
465 |
pthread_mutex_unlock(&LOCK_thread_count); |
|
466 |
||
467 |
if ((file=open_binlog(&log, log_file_name, &errmsg)) < 0) |
|
468 |
{
|
|
469 |
my_errno= ER_MASTER_FATAL_ERROR_READING_BINLOG; |
|
470 |
goto err; |
|
471 |
}
|
|
472 |
if (pos < BIN_LOG_HEADER_SIZE || pos > my_b_filelength(&log)) |
|
473 |
{
|
|
474 |
errmsg= "Client requested master to start replication from \ |
|
475 |
impossible position"; |
|
476 |
my_errno= ER_MASTER_FATAL_ERROR_READING_BINLOG; |
|
477 |
goto err; |
|
478 |
}
|
|
479 |
||
480 |
/*
|
|
481 |
We need to start a packet with something other than 255
|
|
482 |
to distinguish it from error
|
|
483 |
*/
|
|
484 |
packet->set("\0", 1, &my_charset_bin); /* This is the start of a new packet */ |
|
485 |
||
486 |
/*
|
|
487 |
Tell the client about the log name with a fake Rotate event;
|
|
488 |
this is needed even if we also send a Format_description_log_event
|
|
489 |
just after, because that event does not contain the binlog's name.
|
|
490 |
Note that as this Rotate event is sent before
|
|
491 |
Format_description_log_event, the slave cannot have any info to
|
|
492 |
understand this event's format, so the header len of
|
|
493 |
Rotate_log_event is FROZEN (so in 5.0 it will have a header shorter
|
|
494 |
than other events except FORMAT_DESCRIPTION_EVENT).
|
|
495 |
Before 4.0.14 we called fake_rotate_event below only if (pos ==
|
|
496 |
BIN_LOG_HEADER_SIZE), because if this is false then the slave
|
|
497 |
already knows the binlog's name.
|
|
498 |
Since, we always call fake_rotate_event; if the slave already knew
|
|
499 |
the log's name (ex: CHANGE MASTER TO MASTER_LOG_FILE=...) this is
|
|
500 |
useless but does not harm much. It is nice for 3.23 (>=.58) slaves
|
|
501 |
which test Rotate events to see if the master is 4.0 (then they
|
|
502 |
choose to stop because they can't replicate 4.0); by always calling
|
|
503 |
fake_rotate_event we are sure that 3.23.58 and newer will detect the
|
|
504 |
problem as soon as replication starts (BUG#198).
|
|
505 |
Always calling fake_rotate_event makes sending of normal
|
|
506 |
(=from-binlog) Rotate events a priori unneeded, but it is not so
|
|
507 |
simple: the 2 Rotate events are not equivalent, the normal one is
|
|
508 |
before the Stop event, the fake one is after. If we don't send the
|
|
509 |
normal one, then the Stop event will be interpreted (by existing 4.0
|
|
510 |
slaves) as "the master stopped", which is wrong. So for safety,
|
|
511 |
given that we want minimum modification of 4.0, we send the normal
|
|
512 |
and fake Rotates.
|
|
513 |
*/
|
|
514 |
if (fake_rotate_event(net, packet, log_file_name, pos, &errmsg)) |
|
515 |
{
|
|
516 |
/*
|
|
517 |
This error code is not perfect, as fake_rotate_event() does not
|
|
518 |
read anything from the binlog; if it fails it's because of an
|
|
519 |
error in my_net_write(), fortunately it will say so in errmsg.
|
|
520 |
*/
|
|
521 |
my_errno= ER_MASTER_FATAL_ERROR_READING_BINLOG; |
|
522 |
goto err; |
|
523 |
}
|
|
524 |
packet->set("\0", 1, &my_charset_bin); |
|
525 |
/*
|
|
526 |
Adding MAX_LOG_EVENT_HEADER_LEN, since a binlog event can become
|
|
527 |
this larger than the corresponding packet (query) sent
|
|
528 |
from client to master.
|
|
529 |
*/
|
|
530 |
thd->variables.max_allowed_packet+= MAX_LOG_EVENT_HEADER; |
|
531 |
||
532 |
/*
|
|
533 |
We can set log_lock now, it does not move (it's a member of
|
|
534 |
mysql_bin_log, and it's already inited, and it will be destroyed
|
|
535 |
only at shutdown).
|
|
536 |
*/
|
|
537 |
log_lock = mysql_bin_log.get_log_lock(); |
|
538 |
if (pos > BIN_LOG_HEADER_SIZE) |
|
539 |
{
|
|
540 |
/*
|
|
541 |
Try to find a Format_description_log_event at the beginning of
|
|
542 |
the binlog
|
|
543 |
*/
|
|
544 |
if (!(error = Log_event::read_log_event(&log, packet, log_lock))) |
|
545 |
{
|
|
546 |
/*
|
|
547 |
The packet has offsets equal to the normal offsets in a binlog
|
|
548 |
event +1 (the first character is \0).
|
|
549 |
*/
|
|
550 |
if ((*packet)[EVENT_TYPE_OFFSET+1] == FORMAT_DESCRIPTION_EVENT) |
|
551 |
{
|
|
552 |
binlog_can_be_corrupted= test((*packet)[FLAGS_OFFSET+1] & |
|
553 |
LOG_EVENT_BINLOG_IN_USE_F); |
|
554 |
(*packet)[FLAGS_OFFSET+1] &= ~LOG_EVENT_BINLOG_IN_USE_F; |
|
555 |
/*
|
|
556 |
mark that this event with "log_pos=0", so the slave
|
|
557 |
should not increment master's binlog position
|
|
558 |
(rli->group_master_log_pos)
|
|
559 |
*/
|
|
560 |
int4store((char*) packet->ptr()+LOG_POS_OFFSET+1, 0); |
|
561 |
/*
|
|
562 |
if reconnect master sends FD event with `created' as 0
|
|
563 |
to avoid destroying temp tables.
|
|
564 |
*/
|
|
565 |
int4store((char*) packet->ptr()+LOG_EVENT_MINIMAL_HEADER_LEN+ |
|
566 |
ST_CREATED_OFFSET+1, (ulong) 0); |
|
567 |
/* send it */
|
|
568 |
if (my_net_write(net, (uchar*) packet->ptr(), packet->length())) |
|
569 |
{
|
|
570 |
errmsg = "Failed on my_net_write()"; |
|
571 |
my_errno= ER_UNKNOWN_ERROR; |
|
572 |
goto err; |
|
573 |
}
|
|
574 |
||
575 |
/*
|
|
576 |
No need to save this event. We are only doing simple reads
|
|
577 |
(no real parsing of the events) so we don't need it. And so
|
|
578 |
we don't need the artificial Format_description_log_event of
|
|
579 |
3.23&4.x.
|
|
580 |
*/
|
|
581 |
}
|
|
582 |
}
|
|
583 |
else
|
|
584 |
{
|
|
585 |
if (test_for_non_eof_log_read_errors(error, &errmsg)) |
|
586 |
goto err; |
|
587 |
/*
|
|
588 |
It's EOF, nothing to do, go on reading next events, the
|
|
589 |
Format_description_log_event will be found naturally if it is written.
|
|
590 |
*/
|
|
591 |
}
|
|
592 |
/* reset the packet as we wrote to it in any case */
|
|
593 |
packet->set("\0", 1, &my_charset_bin); |
|
594 |
} /* end of if (pos > BIN_LOG_HEADER_SIZE); */ |
|
595 |
else
|
|
596 |
{
|
|
597 |
/* The Format_description_log_event event will be found naturally. */
|
|
598 |
}
|
|
599 |
||
600 |
/* seek to the requested position, to start the requested dump */
|
|
601 |
my_b_seek(&log, pos); // Seek will done on next read |
|
602 |
||
603 |
while (!net->error && net->vio != 0 && !thd->killed) |
|
604 |
{
|
|
605 |
while (!(error = Log_event::read_log_event(&log, packet, log_lock))) |
|
606 |
{
|
|
607 |
#ifndef DBUG_OFF
|
|
608 |
if (max_binlog_dump_events && !left_events--) |
|
609 |
{
|
|
610 |
net_flush(net); |
|
611 |
errmsg = "Debugging binlog dump abort"; |
|
612 |
my_errno= ER_UNKNOWN_ERROR; |
|
613 |
goto err; |
|
614 |
}
|
|
615 |
#endif
|
|
616 |
/*
|
|
617 |
log's filename does not change while it's active
|
|
618 |
*/
|
|
619 |
if (coord) |
|
620 |
coord->pos= uint4korr(packet->ptr() + 1 + LOG_POS_OFFSET); |
|
621 |
||
622 |
if ((*packet)[EVENT_TYPE_OFFSET+1] == FORMAT_DESCRIPTION_EVENT) |
|
623 |
{
|
|
624 |
binlog_can_be_corrupted= test((*packet)[FLAGS_OFFSET+1] & |
|
625 |
LOG_EVENT_BINLOG_IN_USE_F); |
|
626 |
(*packet)[FLAGS_OFFSET+1] &= ~LOG_EVENT_BINLOG_IN_USE_F; |
|
627 |
}
|
|
628 |
else if ((*packet)[EVENT_TYPE_OFFSET+1] == STOP_EVENT) |
|
629 |
binlog_can_be_corrupted= FALSE; |
|
630 |
||
631 |
if (my_net_write(net, (uchar*) packet->ptr(), packet->length())) |
|
632 |
{
|
|
633 |
errmsg = "Failed on my_net_write()"; |
|
634 |
my_errno= ER_UNKNOWN_ERROR; |
|
635 |
goto err; |
|
636 |
}
|
|
637 |
||
638 |
if ((*packet)[LOG_EVENT_OFFSET+1] == LOAD_EVENT) |
|
639 |
{
|
|
640 |
if (send_file(thd)) |
|
641 |
{
|
|
642 |
errmsg = "failed in send_file()"; |
|
643 |
my_errno= ER_UNKNOWN_ERROR; |
|
644 |
goto err; |
|
645 |
}
|
|
646 |
}
|
|
647 |
packet->set("\0", 1, &my_charset_bin); |
|
648 |
}
|
|
649 |
||
650 |
/*
|
|
651 |
here we were reading binlog that was not closed properly (as a result
|
|
652 |
of a crash ?). treat any corruption as EOF
|
|
653 |
*/
|
|
654 |
if (binlog_can_be_corrupted && error != LOG_READ_MEM) |
|
655 |
error=LOG_READ_EOF; |
|
656 |
/*
|
|
657 |
TODO: now that we are logging the offset, check to make sure
|
|
658 |
the recorded offset and the actual match.
|
|
659 |
Guilhem 2003-06: this is not true if this master is a slave
|
|
660 |
<4.0.15 running with --log-slave-updates, because then log_pos may
|
|
661 |
be the offset in the-master-of-this-master's binlog.
|
|
662 |
*/
|
|
663 |
if (test_for_non_eof_log_read_errors(error, &errmsg)) |
|
664 |
goto err; |
|
665 |
||
666 |
if (!(flags & BINLOG_DUMP_NON_BLOCK) && |
|
667 |
mysql_bin_log.is_active(log_file_name)) |
|
668 |
{
|
|
669 |
/*
|
|
670 |
Block until there is more data in the log
|
|
671 |
*/
|
|
672 |
if (net_flush(net)) |
|
673 |
{
|
|
674 |
errmsg = "failed on net_flush()"; |
|
675 |
my_errno= ER_UNKNOWN_ERROR; |
|
676 |
goto err; |
|
677 |
}
|
|
678 |
||
679 |
/*
|
|
680 |
We may have missed the update broadcast from the log
|
|
681 |
that has just happened, let's try to catch it if it did.
|
|
682 |
If we did not miss anything, we just wait for other threads
|
|
683 |
to signal us.
|
|
684 |
*/
|
|
685 |
{
|
|
686 |
log.error=0; |
|
687 |
bool read_packet = 0, fatal_error = 0; |
|
688 |
||
689 |
#ifndef DBUG_OFF
|
|
690 |
if (max_binlog_dump_events && !left_events--) |
|
691 |
{
|
|
692 |
errmsg = "Debugging binlog dump abort"; |
|
693 |
my_errno= ER_UNKNOWN_ERROR; |
|
694 |
goto err; |
|
695 |
}
|
|
696 |
#endif
|
|
697 |
||
698 |
/*
|
|
699 |
No one will update the log while we are reading
|
|
700 |
now, but we'll be quick and just read one record
|
|
701 |
||
702 |
TODO:
|
|
703 |
Add an counter that is incremented for each time we update the
|
|
704 |
binary log. We can avoid the following read if the counter
|
|
705 |
has not been updated since last read.
|
|
706 |
*/
|
|
707 |
||
708 |
pthread_mutex_lock(log_lock); |
|
709 |
switch (Log_event::read_log_event(&log, packet, (pthread_mutex_t*)0)) { |
|
710 |
case 0: |
|
711 |
/* we read successfully, so we'll need to send it to the slave */
|
|
712 |
pthread_mutex_unlock(log_lock); |
|
713 |
read_packet = 1; |
|
714 |
if (coord) |
|
715 |
coord->pos= uint4korr(packet->ptr() + 1 + LOG_POS_OFFSET); |
|
716 |
break; |
|
717 |
||
718 |
case LOG_READ_EOF: |
|
719 |
{
|
|
720 |
int ret; |
|
721 |
DBUG_PRINT("wait",("waiting for data in binary log")); |
|
722 |
if (thd->server_id==0) // for mysqlbinlog (mysqlbinlog.server_id==0) |
|
723 |
{
|
|
724 |
pthread_mutex_unlock(log_lock); |
|
725 |
goto end; |
|
726 |
}
|
|
727 |
||
728 |
#ifndef DBUG_OFF
|
|
729 |
ulong hb_info_counter= 0; |
|
730 |
#endif
|
|
731 |
do
|
|
732 |
{
|
|
733 |
if (coord) |
|
734 |
{
|
|
80.1.1
by Brian Aker
LL() cleanup |
735 |
DBUG_ASSERT(heartbeat_ts && heartbeat_period != 0LL); |
1
by brian
clean slate |
736 |
set_timespec_nsec(*heartbeat_ts, heartbeat_period); |
737 |
}
|
|
738 |
ret= mysql_bin_log.wait_for_update_bin_log(thd, heartbeat_ts); |
|
80.1.1
by Brian Aker
LL() cleanup |
739 |
DBUG_ASSERT(ret == 0 || heartbeat_period != 0LL && coord != NULL); |
1
by brian
clean slate |
740 |
if (ret == ETIMEDOUT || ret == ETIME) |
741 |
{
|
|
742 |
#ifndef DBUG_OFF
|
|
743 |
if (hb_info_counter < 3) |
|
744 |
{
|
|
745 |
sql_print_information("master sends heartbeat message"); |
|
746 |
hb_info_counter++; |
|
747 |
if (hb_info_counter == 3) |
|
748 |
sql_print_information("the rest of heartbeat info skipped ..."); |
|
749 |
}
|
|
750 |
#endif
|
|
751 |
if (send_heartbeat_event(net, packet, coord)) |
|
752 |
{
|
|
753 |
errmsg = "Failed on my_net_write()"; |
|
754 |
my_errno= ER_UNKNOWN_ERROR; |
|
755 |
pthread_mutex_unlock(log_lock); |
|
756 |
goto err; |
|
757 |
}
|
|
758 |
}
|
|
759 |
else
|
|
760 |
{
|
|
761 |
DBUG_ASSERT(ret == 0); |
|
762 |
DBUG_PRINT("wait",("binary log received update")); |
|
763 |
}
|
|
764 |
} while (ret != 0 && coord != NULL && !thd->killed); |
|
765 |
pthread_mutex_unlock(log_lock); |
|
766 |
}
|
|
767 |
break; |
|
768 |
||
769 |
default: |
|
770 |
pthread_mutex_unlock(log_lock); |
|
771 |
fatal_error = 1; |
|
772 |
break; |
|
773 |
}
|
|
774 |
||
775 |
if (read_packet) |
|
776 |
{
|
|
777 |
thd_proc_info(thd, "Sending binlog event to slave"); |
|
778 |
if (my_net_write(net, (uchar*) packet->ptr(), packet->length()) ) |
|
779 |
{
|
|
780 |
errmsg = "Failed on my_net_write()"; |
|
781 |
my_errno= ER_UNKNOWN_ERROR; |
|
782 |
goto err; |
|
783 |
}
|
|
784 |
||
785 |
if ((*packet)[LOG_EVENT_OFFSET+1] == LOAD_EVENT) |
|
786 |
{
|
|
787 |
if (send_file(thd)) |
|
788 |
{
|
|
789 |
errmsg = "failed in send_file()"; |
|
790 |
my_errno= ER_UNKNOWN_ERROR; |
|
791 |
goto err; |
|
792 |
}
|
|
793 |
}
|
|
794 |
packet->set("\0", 1, &my_charset_bin); |
|
795 |
/*
|
|
796 |
No need to net_flush because we will get to flush later when
|
|
797 |
we hit EOF pretty quick
|
|
798 |
*/
|
|
799 |
}
|
|
800 |
||
801 |
if (fatal_error) |
|
802 |
{
|
|
803 |
errmsg = "error reading log entry"; |
|
804 |
my_errno= ER_MASTER_FATAL_ERROR_READING_BINLOG; |
|
805 |
goto err; |
|
806 |
}
|
|
807 |
log.error=0; |
|
808 |
}
|
|
809 |
}
|
|
810 |
else
|
|
811 |
{
|
|
812 |
bool loop_breaker = 0; |
|
813 |
/* need this to break out of the for loop from switch */
|
|
814 |
||
815 |
thd_proc_info(thd, "Finished reading one binlog; switching to next binlog"); |
|
816 |
switch (mysql_bin_log.find_next_log(&linfo, 1)) { |
|
817 |
case LOG_INFO_EOF: |
|
818 |
loop_breaker = (flags & BINLOG_DUMP_NON_BLOCK); |
|
819 |
break; |
|
820 |
case 0: |
|
821 |
break; |
|
822 |
default: |
|
823 |
errmsg = "could not find next log"; |
|
824 |
my_errno= ER_MASTER_FATAL_ERROR_READING_BINLOG; |
|
825 |
goto err; |
|
826 |
}
|
|
827 |
||
828 |
if (loop_breaker) |
|
829 |
break; |
|
830 |
||
831 |
end_io_cache(&log); |
|
832 |
(void) my_close(file, MYF(MY_WME)); |
|
833 |
||
834 |
/*
|
|
835 |
Call fake_rotate_event() in case the previous log (the one which
|
|
836 |
we have just finished reading) did not contain a Rotate event
|
|
837 |
(for example (I don't know any other example) the previous log
|
|
838 |
was the last one before the master was shutdown & restarted).
|
|
839 |
This way we tell the slave about the new log's name and
|
|
840 |
position. If the binlog is 5.0, the next event we are going to
|
|
841 |
read and send is Format_description_log_event.
|
|
842 |
*/
|
|
843 |
if ((file=open_binlog(&log, log_file_name, &errmsg)) < 0 || |
|
844 |
fake_rotate_event(net, packet, log_file_name, BIN_LOG_HEADER_SIZE, |
|
845 |
&errmsg)) |
|
846 |
{
|
|
847 |
my_errno= ER_MASTER_FATAL_ERROR_READING_BINLOG; |
|
848 |
goto err; |
|
849 |
}
|
|
850 |
||
851 |
packet->length(0); |
|
852 |
packet->append('\0'); |
|
853 |
if (coord) |
|
854 |
coord->file_name= log_file_name; // reset to the next |
|
855 |
}
|
|
856 |
}
|
|
857 |
||
858 |
end: |
|
859 |
end_io_cache(&log); |
|
860 |
(void)my_close(file, MYF(MY_WME)); |
|
861 |
||
862 |
my_eof(thd); |
|
863 |
thd_proc_info(thd, "Waiting to finalize termination"); |
|
864 |
pthread_mutex_lock(&LOCK_thread_count); |
|
865 |
thd->current_linfo = 0; |
|
866 |
pthread_mutex_unlock(&LOCK_thread_count); |
|
867 |
DBUG_VOID_RETURN; |
|
868 |
||
869 |
err: |
|
870 |
thd_proc_info(thd, "Waiting to finalize termination"); |
|
871 |
end_io_cache(&log); |
|
872 |
/*
|
|
873 |
Exclude iteration through thread list
|
|
874 |
this is needed for purge_logs() - it will iterate through
|
|
875 |
thread list and update thd->current_linfo->index_file_offset
|
|
876 |
this mutex will make sure that it never tried to update our linfo
|
|
877 |
after we return from this stack frame
|
|
878 |
*/
|
|
879 |
pthread_mutex_lock(&LOCK_thread_count); |
|
880 |
thd->current_linfo = 0; |
|
881 |
pthread_mutex_unlock(&LOCK_thread_count); |
|
882 |
if (file >= 0) |
|
883 |
(void) my_close(file, MYF(MY_WME)); |
|
884 |
||
885 |
my_message(my_errno, errmsg, MYF(0)); |
|
886 |
DBUG_VOID_RETURN; |
|
887 |
}
|
|
888 |
||
889 |
int start_slave(THD* thd , Master_info* mi, bool net_report) |
|
890 |
{
|
|
891 |
int slave_errno= 0; |
|
892 |
int thread_mask; |
|
893 |
DBUG_ENTER("start_slave"); |
|
894 |
||
895 |
lock_slave_threads(mi); // this allows us to cleanly read slave_running |
|
896 |
// Get a mask of _stopped_ threads
|
|
897 |
init_thread_mask(&thread_mask,mi,1 /* inverse */); |
|
898 |
/*
|
|
899 |
Below we will start all stopped threads. But if the user wants to
|
|
900 |
start only one thread, do as if the other thread was running (as we
|
|
901 |
don't wan't to touch the other thread), so set the bit to 0 for the
|
|
902 |
other thread
|
|
903 |
*/
|
|
904 |
if (thd->lex->slave_thd_opt) |
|
905 |
thread_mask&= thd->lex->slave_thd_opt; |
|
906 |
if (thread_mask) //some threads are stopped, start them |
|
907 |
{
|
|
908 |
if (init_master_info(mi,master_info_file,relay_log_info_file, 0, |
|
909 |
thread_mask)) |
|
910 |
slave_errno=ER_MASTER_INFO; |
|
911 |
else if (server_id_supplied && *mi->host) |
|
912 |
{
|
|
913 |
/*
|
|
914 |
If we will start SQL thread we will care about UNTIL options If
|
|
915 |
not and they are specified we will ignore them and warn user
|
|
916 |
about this fact.
|
|
917 |
*/
|
|
918 |
if (thread_mask & SLAVE_SQL) |
|
919 |
{
|
|
920 |
pthread_mutex_lock(&mi->rli.data_lock); |
|
921 |
||
922 |
if (thd->lex->mi.pos) |
|
923 |
{
|
|
924 |
mi->rli.until_condition= Relay_log_info::UNTIL_MASTER_POS; |
|
925 |
mi->rli.until_log_pos= thd->lex->mi.pos; |
|
926 |
/*
|
|
927 |
We don't check thd->lex->mi.log_file_name for NULL here
|
|
928 |
since it is checked in sql_yacc.yy
|
|
929 |
*/
|
|
930 |
strmake(mi->rli.until_log_name, thd->lex->mi.log_file_name, |
|
931 |
sizeof(mi->rli.until_log_name)-1); |
|
932 |
}
|
|
933 |
else if (thd->lex->mi.relay_log_pos) |
|
934 |
{
|
|
935 |
mi->rli.until_condition= Relay_log_info::UNTIL_RELAY_POS; |
|
936 |
mi->rli.until_log_pos= thd->lex->mi.relay_log_pos; |
|
937 |
strmake(mi->rli.until_log_name, thd->lex->mi.relay_log_name, |
|
938 |
sizeof(mi->rli.until_log_name)-1); |
|
939 |
}
|
|
940 |
else
|
|
941 |
mi->rli.clear_until_condition(); |
|
942 |
||
943 |
if (mi->rli.until_condition != Relay_log_info::UNTIL_NONE) |
|
944 |
{
|
|
945 |
/* Preparing members for effective until condition checking */
|
|
946 |
const char *p= fn_ext(mi->rli.until_log_name); |
|
947 |
char *p_end; |
|
948 |
if (*p) |
|
949 |
{
|
|
950 |
//p points to '.'
|
|
951 |
mi->rli.until_log_name_extension= strtoul(++p,&p_end, 10); |
|
952 |
/*
|
|
953 |
p_end points to the first invalid character. If it equals
|
|
954 |
to p, no digits were found, error. If it contains '\0' it
|
|
955 |
means conversion went ok.
|
|
956 |
*/
|
|
957 |
if (p_end==p || *p_end) |
|
958 |
slave_errno=ER_BAD_SLAVE_UNTIL_COND; |
|
959 |
}
|
|
960 |
else
|
|
961 |
slave_errno=ER_BAD_SLAVE_UNTIL_COND; |
|
962 |
||
963 |
/* mark the cached result of the UNTIL comparison as "undefined" */
|
|
964 |
mi->rli.until_log_names_cmp_result= |
|
965 |
Relay_log_info::UNTIL_LOG_NAMES_CMP_UNKNOWN; |
|
966 |
||
967 |
/* Issuing warning then started without --skip-slave-start */
|
|
968 |
if (!opt_skip_slave_start) |
|
969 |
push_warning(thd, MYSQL_ERROR::WARN_LEVEL_NOTE, |
|
970 |
ER_MISSING_SKIP_SLAVE, |
|
971 |
ER(ER_MISSING_SKIP_SLAVE)); |
|
972 |
}
|
|
973 |
||
974 |
pthread_mutex_unlock(&mi->rli.data_lock); |
|
975 |
}
|
|
976 |
else if (thd->lex->mi.pos || thd->lex->mi.relay_log_pos) |
|
977 |
push_warning(thd, MYSQL_ERROR::WARN_LEVEL_NOTE, ER_UNTIL_COND_IGNORED, |
|
978 |
ER(ER_UNTIL_COND_IGNORED)); |
|
979 |
||
980 |
if (!slave_errno) |
|
981 |
slave_errno = start_slave_threads(0 /*no mutex */, |
|
982 |
1 /* wait for start */, |
|
983 |
mi, |
|
984 |
master_info_file,relay_log_info_file, |
|
985 |
thread_mask); |
|
986 |
}
|
|
987 |
else
|
|
988 |
slave_errno = ER_BAD_SLAVE; |
|
989 |
}
|
|
990 |
else
|
|
991 |
{
|
|
992 |
/* no error if all threads are already started, only a warning */
|
|
993 |
push_warning(thd, MYSQL_ERROR::WARN_LEVEL_NOTE, ER_SLAVE_WAS_RUNNING, |
|
994 |
ER(ER_SLAVE_WAS_RUNNING)); |
|
995 |
}
|
|
996 |
||
997 |
unlock_slave_threads(mi); |
|
998 |
||
999 |
if (slave_errno) |
|
1000 |
{
|
|
1001 |
if (net_report) |
|
1002 |
my_message(slave_errno, ER(slave_errno), MYF(0)); |
|
1003 |
DBUG_RETURN(1); |
|
1004 |
}
|
|
1005 |
else if (net_report) |
|
1006 |
my_ok(thd); |
|
1007 |
||
1008 |
DBUG_RETURN(0); |
|
1009 |
}
|
|
1010 |
||
1011 |
||
1012 |
int stop_slave(THD* thd, Master_info* mi, bool net_report ) |
|
1013 |
{
|
|
1014 |
DBUG_ENTER("stop_slave"); |
|
1015 |
||
1016 |
int slave_errno; |
|
1017 |
if (!thd) |
|
1018 |
thd = current_thd; |
|
1019 |
||
1020 |
thd_proc_info(thd, "Killing slave"); |
|
1021 |
int thread_mask; |
|
1022 |
lock_slave_threads(mi); |
|
1023 |
// Get a mask of _running_ threads
|
|
1024 |
init_thread_mask(&thread_mask,mi,0 /* not inverse*/); |
|
1025 |
/*
|
|
1026 |
Below we will stop all running threads.
|
|
1027 |
But if the user wants to stop only one thread, do as if the other thread
|
|
1028 |
was stopped (as we don't wan't to touch the other thread), so set the
|
|
1029 |
bit to 0 for the other thread
|
|
1030 |
*/
|
|
1031 |
if (thd->lex->slave_thd_opt) |
|
1032 |
thread_mask &= thd->lex->slave_thd_opt; |
|
1033 |
||
1034 |
if (thread_mask) |
|
1035 |
{
|
|
1036 |
slave_errno= terminate_slave_threads(mi,thread_mask, |
|
1037 |
1 /*skip lock */); |
|
1038 |
}
|
|
1039 |
else
|
|
1040 |
{
|
|
1041 |
//no error if both threads are already stopped, only a warning
|
|
1042 |
slave_errno= 0; |
|
1043 |
push_warning(thd, MYSQL_ERROR::WARN_LEVEL_NOTE, ER_SLAVE_WAS_NOT_RUNNING, |
|
1044 |
ER(ER_SLAVE_WAS_NOT_RUNNING)); |
|
1045 |
}
|
|
1046 |
unlock_slave_threads(mi); |
|
1047 |
thd_proc_info(thd, 0); |
|
1048 |
||
1049 |
if (slave_errno) |
|
1050 |
{
|
|
1051 |
if (net_report) |
|
1052 |
my_message(slave_errno, ER(slave_errno), MYF(0)); |
|
1053 |
DBUG_RETURN(1); |
|
1054 |
}
|
|
1055 |
else if (net_report) |
|
1056 |
my_ok(thd); |
|
1057 |
||
1058 |
DBUG_RETURN(0); |
|
1059 |
}
|
|
1060 |
||
1061 |
||
1062 |
/*
|
|
1063 |
Remove all relay logs and start replication from the start
|
|
1064 |
||
1065 |
SYNOPSIS
|
|
1066 |
reset_slave()
|
|
1067 |
thd Thread handler
|
|
1068 |
mi Master info for the slave
|
|
1069 |
||
1070 |
RETURN
|
|
1071 |
0 ok
|
|
1072 |
1 error
|
|
1073 |
*/
|
|
1074 |
||
1075 |
||
1076 |
int reset_slave(THD *thd, Master_info* mi) |
|
1077 |
{
|
|
15
by brian
Fix for stat, NETWARE removal |
1078 |
struct stat stat_area; |
1
by brian
clean slate |
1079 |
char fname[FN_REFLEN]; |
1080 |
int thread_mask= 0, error= 0; |
|
1081 |
uint sql_errno=0; |
|
1082 |
const char* errmsg=0; |
|
1083 |
DBUG_ENTER("reset_slave"); |
|
1084 |
||
1085 |
lock_slave_threads(mi); |
|
1086 |
init_thread_mask(&thread_mask,mi,0 /* not inverse */); |
|
1087 |
if (thread_mask) // We refuse if any slave thread is running |
|
1088 |
{
|
|
1089 |
sql_errno= ER_SLAVE_MUST_STOP; |
|
1090 |
error=1; |
|
1091 |
goto err; |
|
1092 |
}
|
|
1093 |
||
1094 |
ha_reset_slave(thd); |
|
1095 |
||
1096 |
// delete relay logs, clear relay log coordinates
|
|
1097 |
if ((error= purge_relay_logs(&mi->rli, thd, |
|
1098 |
1 /* just reset */, |
|
1099 |
&errmsg))) |
|
1100 |
goto err; |
|
1101 |
||
1102 |
/* Clear master's log coordinates */
|
|
1103 |
init_master_log_pos(mi); |
|
1104 |
/*
|
|
1105 |
Reset errors (the idea is that we forget about the
|
|
1106 |
old master).
|
|
1107 |
*/
|
|
1108 |
mi->rli.clear_error(); |
|
1109 |
mi->rli.clear_until_condition(); |
|
1110 |
||
1111 |
// close master_info_file, relay_log_info_file, set mi->inited=rli->inited=0
|
|
1112 |
end_master_info(mi); |
|
1113 |
// and delete these two files
|
|
1114 |
fn_format(fname, master_info_file, mysql_data_home, "", 4+32); |
|
15
by brian
Fix for stat, NETWARE removal |
1115 |
if (!stat(fname, &stat_area) && my_delete(fname, MYF(MY_WME))) |
1
by brian
clean slate |
1116 |
{
|
1117 |
error=1; |
|
1118 |
goto err; |
|
1119 |
}
|
|
1120 |
// delete relay_log_info_file
|
|
1121 |
fn_format(fname, relay_log_info_file, mysql_data_home, "", 4+32); |
|
15
by brian
Fix for stat, NETWARE removal |
1122 |
if (!stat(fname, &stat_area) && my_delete(fname, MYF(MY_WME))) |
1
by brian
clean slate |
1123 |
{
|
1124 |
error=1; |
|
1125 |
goto err; |
|
1126 |
}
|
|
1127 |
||
1128 |
err: |
|
1129 |
unlock_slave_threads(mi); |
|
1130 |
if (error) |
|
1131 |
my_error(sql_errno, MYF(0), errmsg); |
|
1132 |
DBUG_RETURN(error); |
|
1133 |
}
|
|
1134 |
||
1135 |
/*
|
|
1136 |
||
1137 |
Kill all Binlog_dump threads which previously talked to the same slave
|
|
1138 |
("same" means with the same server id). Indeed, if the slave stops, if the
|
|
1139 |
Binlog_dump thread is waiting (pthread_cond_wait) for binlog update, then it
|
|
1140 |
will keep existing until a query is written to the binlog. If the master is
|
|
1141 |
idle, then this could last long, and if the slave reconnects, we could have 2
|
|
1142 |
Binlog_dump threads in SHOW PROCESSLIST, until a query is written to the
|
|
1143 |
binlog. To avoid this, when the slave reconnects and sends COM_BINLOG_DUMP,
|
|
1144 |
the master kills any existing thread with the slave's server id (if this id is
|
|
1145 |
not zero; it will be true for real slaves, but false for mysqlbinlog when it
|
|
1146 |
sends COM_BINLOG_DUMP to get a remote binlog dump).
|
|
1147 |
||
1148 |
SYNOPSIS
|
|
1149 |
kill_zombie_dump_threads()
|
|
1150 |
slave_server_id the slave's server id
|
|
1151 |
||
1152 |
*/
|
|
1153 |
||
1154 |
||
1155 |
void kill_zombie_dump_threads(uint32 slave_server_id) |
|
1156 |
{
|
|
1157 |
pthread_mutex_lock(&LOCK_thread_count); |
|
1158 |
I_List_iterator<THD> it(threads); |
|
1159 |
THD *tmp; |
|
1160 |
||
1161 |
while ((tmp=it++)) |
|
1162 |
{
|
|
1163 |
if (tmp->command == COM_BINLOG_DUMP && |
|
1164 |
tmp->server_id == slave_server_id) |
|
1165 |
{
|
|
1166 |
pthread_mutex_lock(&tmp->LOCK_delete); // Lock from delete |
|
1167 |
break; |
|
1168 |
}
|
|
1169 |
}
|
|
1170 |
pthread_mutex_unlock(&LOCK_thread_count); |
|
1171 |
if (tmp) |
|
1172 |
{
|
|
1173 |
/*
|
|
1174 |
Here we do not call kill_one_thread() as
|
|
1175 |
it will be slow because it will iterate through the list
|
|
1176 |
again. We just to do kill the thread ourselves.
|
|
1177 |
*/
|
|
1178 |
tmp->awake(THD::KILL_QUERY); |
|
1179 |
pthread_mutex_unlock(&tmp->LOCK_delete); |
|
1180 |
}
|
|
1181 |
}
|
|
1182 |
||
1183 |
||
1184 |
bool change_master(THD* thd, Master_info* mi) |
|
1185 |
{
|
|
1186 |
int thread_mask; |
|
1187 |
const char* errmsg= 0; |
|
1188 |
bool need_relay_log_purge= 1; |
|
1189 |
DBUG_ENTER("change_master"); |
|
1190 |
||
1191 |
lock_slave_threads(mi); |
|
1192 |
init_thread_mask(&thread_mask,mi,0 /*not inverse*/); |
|
1193 |
if (thread_mask) // We refuse if any slave thread is running |
|
1194 |
{
|
|
1195 |
my_message(ER_SLAVE_MUST_STOP, ER(ER_SLAVE_MUST_STOP), MYF(0)); |
|
1196 |
unlock_slave_threads(mi); |
|
1197 |
DBUG_RETURN(TRUE); |
|
1198 |
}
|
|
1199 |
||
1200 |
thd_proc_info(thd, "Changing master"); |
|
1201 |
LEX_MASTER_INFO* lex_mi= &thd->lex->mi; |
|
1202 |
// TODO: see if needs re-write
|
|
1203 |
if (init_master_info(mi, master_info_file, relay_log_info_file, 0, |
|
1204 |
thread_mask)) |
|
1205 |
{
|
|
1206 |
my_message(ER_MASTER_INFO, ER(ER_MASTER_INFO), MYF(0)); |
|
1207 |
unlock_slave_threads(mi); |
|
1208 |
DBUG_RETURN(TRUE); |
|
1209 |
}
|
|
1210 |
||
1211 |
/*
|
|
1212 |
Data lock not needed since we have already stopped the running threads,
|
|
1213 |
and we have the hold on the run locks which will keep all threads that
|
|
1214 |
could possibly modify the data structures from running
|
|
1215 |
*/
|
|
1216 |
||
1217 |
/*
|
|
1218 |
If the user specified host or port without binlog or position,
|
|
1219 |
reset binlog's name to FIRST and position to 4.
|
|
1220 |
*/
|
|
1221 |
||
1222 |
if ((lex_mi->host || lex_mi->port) && !lex_mi->log_file_name && !lex_mi->pos) |
|
1223 |
{
|
|
1224 |
mi->master_log_name[0] = 0; |
|
1225 |
mi->master_log_pos= BIN_LOG_HEADER_SIZE; |
|
1226 |
}
|
|
1227 |
||
1228 |
if (lex_mi->log_file_name) |
|
1229 |
strmake(mi->master_log_name, lex_mi->log_file_name, |
|
1230 |
sizeof(mi->master_log_name)-1); |
|
1231 |
if (lex_mi->pos) |
|
1232 |
{
|
|
1233 |
mi->master_log_pos= lex_mi->pos; |
|
1234 |
}
|
|
1235 |
DBUG_PRINT("info", ("master_log_pos: %lu", (ulong) mi->master_log_pos)); |
|
1236 |
||
1237 |
if (lex_mi->host) |
|
1238 |
strmake(mi->host, lex_mi->host, sizeof(mi->host)-1); |
|
1239 |
if (lex_mi->user) |
|
1240 |
strmake(mi->user, lex_mi->user, sizeof(mi->user)-1); |
|
1241 |
if (lex_mi->password) |
|
1242 |
strmake(mi->password, lex_mi->password, sizeof(mi->password)-1); |
|
1243 |
if (lex_mi->port) |
|
1244 |
mi->port = lex_mi->port; |
|
1245 |
if (lex_mi->connect_retry) |
|
1246 |
mi->connect_retry = lex_mi->connect_retry; |
|
1247 |
if (lex_mi->heartbeat_opt != LEX_MASTER_INFO::LEX_MI_UNCHANGED) |
|
1248 |
mi->heartbeat_period = lex_mi->heartbeat_period; |
|
1249 |
else
|
|
1250 |
mi->heartbeat_period= (float) min(SLAVE_MAX_HEARTBEAT_PERIOD, |
|
1251 |
(slave_net_timeout/2.0)); |
|
80.1.1
by Brian Aker
LL() cleanup |
1252 |
mi->received_heartbeats= 0LL; // counter lives until master is CHANGEd |
1
by brian
clean slate |
1253 |
if (lex_mi->ssl != LEX_MASTER_INFO::LEX_MI_UNCHANGED) |
1254 |
mi->ssl= (lex_mi->ssl == LEX_MASTER_INFO::LEX_MI_ENABLE); |
|
1255 |
||
1256 |
if (lex_mi->ssl_verify_server_cert != LEX_MASTER_INFO::LEX_MI_UNCHANGED) |
|
1257 |
mi->ssl_verify_server_cert= |
|
1258 |
(lex_mi->ssl_verify_server_cert == LEX_MASTER_INFO::LEX_MI_ENABLE); |
|
1259 |
||
1260 |
if (lex_mi->ssl_ca) |
|
1261 |
strmake(mi->ssl_ca, lex_mi->ssl_ca, sizeof(mi->ssl_ca)-1); |
|
1262 |
if (lex_mi->ssl_capath) |
|
1263 |
strmake(mi->ssl_capath, lex_mi->ssl_capath, sizeof(mi->ssl_capath)-1); |
|
1264 |
if (lex_mi->ssl_cert) |
|
1265 |
strmake(mi->ssl_cert, lex_mi->ssl_cert, sizeof(mi->ssl_cert)-1); |
|
1266 |
if (lex_mi->ssl_cipher) |
|
1267 |
strmake(mi->ssl_cipher, lex_mi->ssl_cipher, sizeof(mi->ssl_cipher)-1); |
|
1268 |
if (lex_mi->ssl_key) |
|
1269 |
strmake(mi->ssl_key, lex_mi->ssl_key, sizeof(mi->ssl_key)-1); |
|
1270 |
if (lex_mi->ssl || lex_mi->ssl_ca || lex_mi->ssl_capath || |
|
1271 |
lex_mi->ssl_cert || lex_mi->ssl_cipher || lex_mi->ssl_key || |
|
1272 |
lex_mi->ssl_verify_server_cert ) |
|
1273 |
push_warning(thd, MYSQL_ERROR::WARN_LEVEL_NOTE, |
|
1274 |
ER_SLAVE_IGNORED_SSL_PARAMS, ER(ER_SLAVE_IGNORED_SSL_PARAMS)); |
|
1275 |
||
1276 |
if (lex_mi->relay_log_name) |
|
1277 |
{
|
|
1278 |
need_relay_log_purge= 0; |
|
1279 |
strmake(mi->rli.group_relay_log_name,lex_mi->relay_log_name, |
|
1280 |
sizeof(mi->rli.group_relay_log_name)-1); |
|
1281 |
strmake(mi->rli.event_relay_log_name,lex_mi->relay_log_name, |
|
1282 |
sizeof(mi->rli.event_relay_log_name)-1); |
|
1283 |
}
|
|
1284 |
||
1285 |
if (lex_mi->relay_log_pos) |
|
1286 |
{
|
|
1287 |
need_relay_log_purge= 0; |
|
1288 |
mi->rli.group_relay_log_pos= mi->rli.event_relay_log_pos= lex_mi->relay_log_pos; |
|
1289 |
}
|
|
1290 |
||
1291 |
/*
|
|
1292 |
If user did specify neither host nor port nor any log name nor any log
|
|
1293 |
pos, i.e. he specified only user/password/master_connect_retry, he probably
|
|
1294 |
wants replication to resume from where it had left, i.e. from the
|
|
1295 |
coordinates of the **SQL** thread (imagine the case where the I/O is ahead
|
|
1296 |
of the SQL; restarting from the coordinates of the I/O would lose some
|
|
1297 |
events which is probably unwanted when you are just doing minor changes
|
|
1298 |
like changing master_connect_retry).
|
|
1299 |
A side-effect is that if only the I/O thread was started, this thread may
|
|
1300 |
restart from ''/4 after the CHANGE MASTER. That's a minor problem (it is a
|
|
1301 |
much more unlikely situation than the one we are fixing here).
|
|
1302 |
Note: coordinates of the SQL thread must be read here, before the
|
|
1303 |
'if (need_relay_log_purge)' block which resets them.
|
|
1304 |
*/
|
|
1305 |
if (!lex_mi->host && !lex_mi->port && |
|
1306 |
!lex_mi->log_file_name && !lex_mi->pos && |
|
1307 |
need_relay_log_purge) |
|
1308 |
{
|
|
1309 |
/*
|
|
1310 |
Sometimes mi->rli.master_log_pos == 0 (it happens when the SQL thread is
|
|
1311 |
not initialized), so we use a max().
|
|
1312 |
What happens to mi->rli.master_log_pos during the initialization stages
|
|
1313 |
of replication is not 100% clear, so we guard against problems using
|
|
1314 |
max().
|
|
1315 |
*/
|
|
1316 |
mi->master_log_pos = max(BIN_LOG_HEADER_SIZE, |
|
1317 |
mi->rli.group_master_log_pos); |
|
1318 |
strmake(mi->master_log_name, mi->rli.group_master_log_name, |
|
1319 |
sizeof(mi->master_log_name)-1); |
|
1320 |
}
|
|
1321 |
/*
|
|
1322 |
Relay log's IO_CACHE may not be inited, if rli->inited==0 (server was never
|
|
1323 |
a slave before).
|
|
1324 |
*/
|
|
1325 |
if (flush_master_info(mi, 0)) |
|
1326 |
{
|
|
1327 |
my_error(ER_RELAY_LOG_INIT, MYF(0), "Failed to flush master info file"); |
|
1328 |
unlock_slave_threads(mi); |
|
1329 |
DBUG_RETURN(TRUE); |
|
1330 |
}
|
|
1331 |
if (need_relay_log_purge) |
|
1332 |
{
|
|
1333 |
relay_log_purge= 1; |
|
1334 |
thd_proc_info(thd, "Purging old relay logs"); |
|
1335 |
if (purge_relay_logs(&mi->rli, thd, |
|
1336 |
0 /* not only reset, but also reinit */, |
|
1337 |
&errmsg)) |
|
1338 |
{
|
|
1339 |
my_error(ER_RELAY_LOG_FAIL, MYF(0), errmsg); |
|
1340 |
unlock_slave_threads(mi); |
|
1341 |
DBUG_RETURN(TRUE); |
|
1342 |
}
|
|
1343 |
}
|
|
1344 |
else
|
|
1345 |
{
|
|
1346 |
const char* msg; |
|
1347 |
relay_log_purge= 0; |
|
1348 |
/* Relay log is already initialized */
|
|
1349 |
if (init_relay_log_pos(&mi->rli, |
|
1350 |
mi->rli.group_relay_log_name, |
|
1351 |
mi->rli.group_relay_log_pos, |
|
1352 |
0 /*no data lock*/, |
|
1353 |
&msg, 0)) |
|
1354 |
{
|
|
1355 |
my_error(ER_RELAY_LOG_INIT, MYF(0), msg); |
|
1356 |
unlock_slave_threads(mi); |
|
1357 |
DBUG_RETURN(TRUE); |
|
1358 |
}
|
|
1359 |
}
|
|
1360 |
/*
|
|
1361 |
Coordinates in rli were spoilt by the 'if (need_relay_log_purge)' block,
|
|
1362 |
so restore them to good values. If we left them to ''/0, that would work;
|
|
1363 |
but that would fail in the case of 2 successive CHANGE MASTER (without a
|
|
1364 |
START SLAVE in between): because first one would set the coords in mi to
|
|
1365 |
the good values of those in rli, the set those in rli to ''/0, then
|
|
1366 |
second CHANGE MASTER would set the coords in mi to those of rli, i.e. to
|
|
1367 |
''/0: we have lost all copies of the original good coordinates.
|
|
1368 |
That's why we always save good coords in rli.
|
|
1369 |
*/
|
|
1370 |
mi->rli.group_master_log_pos= mi->master_log_pos; |
|
1371 |
DBUG_PRINT("info", ("master_log_pos: %lu", (ulong) mi->master_log_pos)); |
|
1372 |
strmake(mi->rli.group_master_log_name,mi->master_log_name, |
|
1373 |
sizeof(mi->rli.group_master_log_name)-1); |
|
1374 |
||
1375 |
if (!mi->rli.group_master_log_name[0]) // uninitialized case |
|
1376 |
mi->rli.group_master_log_pos=0; |
|
1377 |
||
1378 |
pthread_mutex_lock(&mi->rli.data_lock); |
|
1379 |
mi->rli.abort_pos_wait++; /* for MASTER_POS_WAIT() to abort */ |
|
1380 |
/* Clear the errors, for a clean start */
|
|
1381 |
mi->rli.clear_error(); |
|
1382 |
mi->rli.clear_until_condition(); |
|
1383 |
/*
|
|
1384 |
If we don't write new coordinates to disk now, then old will remain in
|
|
1385 |
relay-log.info until START SLAVE is issued; but if mysqld is shutdown
|
|
1386 |
before START SLAVE, then old will remain in relay-log.info, and will be the
|
|
1387 |
in-memory value at restart (thus causing errors, as the old relay log does
|
|
1388 |
not exist anymore).
|
|
1389 |
*/
|
|
1390 |
flush_relay_log_info(&mi->rli); |
|
1391 |
pthread_cond_broadcast(&mi->data_cond); |
|
1392 |
pthread_mutex_unlock(&mi->rli.data_lock); |
|
1393 |
||
1394 |
unlock_slave_threads(mi); |
|
1395 |
thd_proc_info(thd, 0); |
|
1396 |
my_ok(thd); |
|
1397 |
DBUG_RETURN(FALSE); |
|
1398 |
}
|
|
1399 |
||
1400 |
int reset_master(THD* thd) |
|
1401 |
{
|
|
1402 |
if (!mysql_bin_log.is_open()) |
|
1403 |
{
|
|
1404 |
my_message(ER_FLUSH_MASTER_BINLOG_CLOSED, |
|
1405 |
ER(ER_FLUSH_MASTER_BINLOG_CLOSED), MYF(ME_BELL+ME_WAITTANG)); |
|
1406 |
return 1; |
|
1407 |
}
|
|
1408 |
return mysql_bin_log.reset_logs(thd); |
|
1409 |
}
|
|
1410 |
||
1411 |
int cmp_master_pos(const char* log_file_name1, ulonglong log_pos1, |
|
1412 |
const char* log_file_name2, ulonglong log_pos2) |
|
1413 |
{
|
|
1414 |
int res; |
|
1415 |
uint log_file_name1_len= strlen(log_file_name1); |
|
1416 |
uint log_file_name2_len= strlen(log_file_name2); |
|
1417 |
||
1418 |
// We assume that both log names match up to '.'
|
|
1419 |
if (log_file_name1_len == log_file_name2_len) |
|
1420 |
{
|
|
1421 |
if ((res= strcmp(log_file_name1, log_file_name2))) |
|
1422 |
return res; |
|
1423 |
return (log_pos1 < log_pos2) ? -1 : (log_pos1 == log_pos2) ? 0 : 1; |
|
1424 |
}
|
|
1425 |
return ((log_file_name1_len < log_file_name2_len) ? -1 : 1); |
|
1426 |
}
|
|
1427 |
||
1428 |
||
1429 |
bool mysql_show_binlog_events(THD* thd) |
|
1430 |
{
|
|
1431 |
Protocol *protocol= thd->protocol; |
|
1432 |
List<Item> field_list; |
|
1433 |
const char *errmsg = 0; |
|
1434 |
bool ret = TRUE; |
|
1435 |
IO_CACHE log; |
|
1436 |
File file = -1; |
|
1437 |
DBUG_ENTER("mysql_show_binlog_events"); |
|
1438 |
||
1439 |
Log_event::init_show_field_list(&field_list); |
|
1440 |
if (protocol->send_fields(&field_list, |
|
1441 |
Protocol::SEND_NUM_ROWS | Protocol::SEND_EOF)) |
|
1442 |
DBUG_RETURN(TRUE); |
|
1443 |
||
1444 |
Format_description_log_event *description_event= new |
|
1445 |
Format_description_log_event(3); /* MySQL 4.0 by default */ |
|
1446 |
||
1447 |
/*
|
|
1448 |
Wait for handlers to insert any pending information
|
|
1449 |
into the binlog. For e.g. ndb which updates the binlog asynchronously
|
|
1450 |
this is needed so that the uses sees all its own commands in the binlog
|
|
1451 |
*/
|
|
1452 |
ha_binlog_wait(thd); |
|
1453 |
||
1454 |
if (mysql_bin_log.is_open()) |
|
1455 |
{
|
|
1456 |
LEX_MASTER_INFO *lex_mi= &thd->lex->mi; |
|
1457 |
SELECT_LEX_UNIT *unit= &thd->lex->unit; |
|
1458 |
ha_rows event_count, limit_start, limit_end; |
|
1459 |
my_off_t pos = max(BIN_LOG_HEADER_SIZE, lex_mi->pos); // user-friendly |
|
1460 |
char search_file_name[FN_REFLEN], *name; |
|
1461 |
const char *log_file_name = lex_mi->log_file_name; |
|
1462 |
pthread_mutex_t *log_lock = mysql_bin_log.get_log_lock(); |
|
1463 |
LOG_INFO linfo; |
|
1464 |
Log_event* ev; |
|
1465 |
||
1466 |
unit->set_limit(thd->lex->current_select); |
|
1467 |
limit_start= unit->offset_limit_cnt; |
|
1468 |
limit_end= unit->select_limit_cnt; |
|
1469 |
||
1470 |
name= search_file_name; |
|
1471 |
if (log_file_name) |
|
1472 |
mysql_bin_log.make_log_name(search_file_name, log_file_name); |
|
1473 |
else
|
|
1474 |
name=0; // Find first log |
|
1475 |
||
1476 |
linfo.index_file_offset = 0; |
|
1477 |
||
1478 |
if (mysql_bin_log.find_log_pos(&linfo, name, 1)) |
|
1479 |
{
|
|
1480 |
errmsg = "Could not find target log"; |
|
1481 |
goto err; |
|
1482 |
}
|
|
1483 |
||
1484 |
pthread_mutex_lock(&LOCK_thread_count); |
|
1485 |
thd->current_linfo = &linfo; |
|
1486 |
pthread_mutex_unlock(&LOCK_thread_count); |
|
1487 |
||
1488 |
if ((file=open_binlog(&log, linfo.log_file_name, &errmsg)) < 0) |
|
1489 |
goto err; |
|
1490 |
||
1491 |
/*
|
|
1492 |
to account binlog event header size
|
|
1493 |
*/
|
|
1494 |
thd->variables.max_allowed_packet += MAX_LOG_EVENT_HEADER; |
|
1495 |
||
1496 |
pthread_mutex_lock(log_lock); |
|
1497 |
||
1498 |
/*
|
|
1499 |
open_binlog() sought to position 4.
|
|
1500 |
Read the first event in case it's a Format_description_log_event, to
|
|
1501 |
know the format. If there's no such event, we are 3.23 or 4.x. This
|
|
1502 |
code, like before, can't read 3.23 binlogs.
|
|
1503 |
This code will fail on a mixed relay log (one which has Format_desc then
|
|
1504 |
Rotate then Format_desc).
|
|
1505 |
*/
|
|
1506 |
ev = Log_event::read_log_event(&log,(pthread_mutex_t*)0,description_event); |
|
1507 |
if (ev) |
|
1508 |
{
|
|
1509 |
if (ev->get_type_code() == FORMAT_DESCRIPTION_EVENT) |
|
1510 |
{
|
|
1511 |
delete description_event; |
|
1512 |
description_event= (Format_description_log_event*) ev; |
|
1513 |
}
|
|
1514 |
else
|
|
1515 |
delete ev; |
|
1516 |
}
|
|
1517 |
||
1518 |
my_b_seek(&log, pos); |
|
1519 |
||
1520 |
if (!description_event->is_valid()) |
|
1521 |
{
|
|
1522 |
errmsg="Invalid Format_description event; could be out of memory"; |
|
1523 |
goto err; |
|
1524 |
}
|
|
1525 |
||
1526 |
for (event_count = 0; |
|
1527 |
(ev = Log_event::read_log_event(&log,(pthread_mutex_t*) 0, |
|
1528 |
description_event)); ) |
|
1529 |
{
|
|
1530 |
if (event_count >= limit_start && |
|
1531 |
ev->net_send(protocol, linfo.log_file_name, pos)) |
|
1532 |
{
|
|
1533 |
errmsg = "Net error"; |
|
1534 |
delete ev; |
|
1535 |
pthread_mutex_unlock(log_lock); |
|
1536 |
goto err; |
|
1537 |
}
|
|
1538 |
||
1539 |
pos = my_b_tell(&log); |
|
1540 |
delete ev; |
|
1541 |
||
1542 |
if (++event_count >= limit_end) |
|
1543 |
break; |
|
1544 |
}
|
|
1545 |
||
1546 |
if (event_count < limit_end && log.error) |
|
1547 |
{
|
|
1548 |
errmsg = "Wrong offset or I/O error"; |
|
1549 |
pthread_mutex_unlock(log_lock); |
|
1550 |
goto err; |
|
1551 |
}
|
|
1552 |
||
1553 |
pthread_mutex_unlock(log_lock); |
|
1554 |
}
|
|
1555 |
||
1556 |
ret= FALSE; |
|
1557 |
||
1558 |
err: |
|
1559 |
delete description_event; |
|
1560 |
if (file >= 0) |
|
1561 |
{
|
|
1562 |
end_io_cache(&log); |
|
1563 |
(void) my_close(file, MYF(MY_WME)); |
|
1564 |
}
|
|
1565 |
||
1566 |
if (errmsg) |
|
1567 |
my_error(ER_ERROR_WHEN_EXECUTING_COMMAND, MYF(0), |
|
1568 |
"SHOW BINLOG EVENTS", errmsg); |
|
1569 |
else
|
|
1570 |
my_eof(thd); |
|
1571 |
||
1572 |
pthread_mutex_lock(&LOCK_thread_count); |
|
1573 |
thd->current_linfo = 0; |
|
1574 |
pthread_mutex_unlock(&LOCK_thread_count); |
|
1575 |
DBUG_RETURN(ret); |
|
1576 |
}
|
|
1577 |
||
1578 |
||
1579 |
bool show_binlog_info(THD* thd) |
|
1580 |
{
|
|
1581 |
Protocol *protocol= thd->protocol; |
|
1582 |
DBUG_ENTER("show_binlog_info"); |
|
1583 |
List<Item> field_list; |
|
1584 |
field_list.push_back(new Item_empty_string("File", FN_REFLEN)); |
|
1585 |
field_list.push_back(new Item_return_int("Position",20, |
|
1586 |
MYSQL_TYPE_LONGLONG)); |
|
1587 |
field_list.push_back(new Item_empty_string("Binlog_Do_DB",255)); |
|
1588 |
field_list.push_back(new Item_empty_string("Binlog_Ignore_DB",255)); |
|
1589 |
||
1590 |
if (protocol->send_fields(&field_list, |
|
1591 |
Protocol::SEND_NUM_ROWS | Protocol::SEND_EOF)) |
|
1592 |
DBUG_RETURN(TRUE); |
|
1593 |
protocol->prepare_for_resend(); |
|
1594 |
||
1595 |
if (mysql_bin_log.is_open()) |
|
1596 |
{
|
|
1597 |
LOG_INFO li; |
|
1598 |
mysql_bin_log.get_current_log(&li); |
|
1599 |
int dir_len = dirname_length(li.log_file_name); |
|
1600 |
protocol->store(li.log_file_name + dir_len, &my_charset_bin); |
|
1601 |
protocol->store((ulonglong) li.pos); |
|
1602 |
protocol->store(binlog_filter->get_do_db()); |
|
1603 |
protocol->store(binlog_filter->get_ignore_db()); |
|
1604 |
if (protocol->write()) |
|
1605 |
DBUG_RETURN(TRUE); |
|
1606 |
}
|
|
1607 |
my_eof(thd); |
|
1608 |
DBUG_RETURN(FALSE); |
|
1609 |
}
|
|
1610 |
||
1611 |
||
1612 |
/*
|
|
1613 |
Send a list of all binary logs to client
|
|
1614 |
||
1615 |
SYNOPSIS
|
|
1616 |
show_binlogs()
|
|
1617 |
thd Thread specific variable
|
|
1618 |
||
1619 |
RETURN VALUES
|
|
1620 |
FALSE OK
|
|
1621 |
TRUE error
|
|
1622 |
*/
|
|
1623 |
||
1624 |
bool show_binlogs(THD* thd) |
|
1625 |
{
|
|
1626 |
IO_CACHE *index_file; |
|
1627 |
LOG_INFO cur; |
|
1628 |
File file; |
|
1629 |
char fname[FN_REFLEN]; |
|
1630 |
List<Item> field_list; |
|
1631 |
uint length; |
|
1632 |
int cur_dir_len; |
|
1633 |
Protocol *protocol= thd->protocol; |
|
1634 |
DBUG_ENTER("show_binlogs"); |
|
1635 |
||
1636 |
if (!mysql_bin_log.is_open()) |
|
1637 |
{
|
|
1638 |
my_message(ER_NO_BINARY_LOGGING, ER(ER_NO_BINARY_LOGGING), MYF(0)); |
|
1639 |
return 1; |
|
1640 |
}
|
|
1641 |
||
1642 |
field_list.push_back(new Item_empty_string("Log_name", 255)); |
|
1643 |
field_list.push_back(new Item_return_int("File_size", 20, |
|
1644 |
MYSQL_TYPE_LONGLONG)); |
|
1645 |
if (protocol->send_fields(&field_list, |
|
1646 |
Protocol::SEND_NUM_ROWS | Protocol::SEND_EOF)) |
|
1647 |
DBUG_RETURN(TRUE); |
|
1648 |
||
1649 |
pthread_mutex_lock(mysql_bin_log.get_log_lock()); |
|
1650 |
mysql_bin_log.lock_index(); |
|
1651 |
index_file=mysql_bin_log.get_index_file(); |
|
1652 |
||
1653 |
mysql_bin_log.raw_get_current_log(&cur); // dont take mutex |
|
1654 |
pthread_mutex_unlock(mysql_bin_log.get_log_lock()); // lockdep, OK |
|
1655 |
||
1656 |
cur_dir_len= dirname_length(cur.log_file_name); |
|
1657 |
||
1658 |
reinit_io_cache(index_file, READ_CACHE, (my_off_t) 0, 0, 0); |
|
1659 |
||
1660 |
/* The file ends with EOF or empty line */
|
|
1661 |
while ((length=my_b_gets(index_file, fname, sizeof(fname))) > 1) |
|
1662 |
{
|
|
1663 |
int dir_len; |
|
1664 |
ulonglong file_length= 0; // Length if open fails |
|
1665 |
fname[--length] = '\0'; // remove the newline |
|
1666 |
||
1667 |
protocol->prepare_for_resend(); |
|
1668 |
dir_len= dirname_length(fname); |
|
1669 |
length-= dir_len; |
|
1670 |
protocol->store(fname + dir_len, length, &my_charset_bin); |
|
1671 |
||
1672 |
if (!(strncmp(fname+dir_len, cur.log_file_name+cur_dir_len, length))) |
|
1673 |
file_length= cur.pos; /* The active log, use the active position */ |
|
1674 |
else
|
|
1675 |
{
|
|
1676 |
/* this is an old log, open it and find the size */
|
|
1677 |
if ((file= my_open(fname, O_RDONLY | O_SHARE | O_BINARY, |
|
1678 |
MYF(0))) >= 0) |
|
1679 |
{
|
|
1680 |
file_length= (ulonglong) my_seek(file, 0L, MY_SEEK_END, MYF(0)); |
|
1681 |
my_close(file, MYF(0)); |
|
1682 |
}
|
|
1683 |
}
|
|
1684 |
protocol->store(file_length); |
|
1685 |
if (protocol->write()) |
|
1686 |
goto err; |
|
1687 |
}
|
|
1688 |
mysql_bin_log.unlock_index(); |
|
1689 |
my_eof(thd); |
|
1690 |
DBUG_RETURN(FALSE); |
|
1691 |
||
1692 |
err: |
|
1693 |
mysql_bin_log.unlock_index(); |
|
1694 |
DBUG_RETURN(TRUE); |
|
1695 |
}
|
|
1696 |
||
1697 |
/**
|
|
1698 |
Load data's io cache specific hook to be executed
|
|
1699 |
before a chunk of data is being read into the cache's buffer
|
|
1700 |
The fuction instantianates and writes into the binlog
|
|
1701 |
replication events along LOAD DATA processing.
|
|
1702 |
|
|
1703 |
@param file pointer to io-cache
|
|
1704 |
@return 0
|
|
1705 |
*/
|
|
1706 |
int log_loaded_block(IO_CACHE* file) |
|
1707 |
{
|
|
1708 |
DBUG_ENTER("log_loaded_block"); |
|
1709 |
LOAD_FILE_INFO *lf_info; |
|
1710 |
uint block_len; |
|
1711 |
/* buffer contains position where we started last read */
|
|
1712 |
uchar* buffer= (uchar*) my_b_get_buffer_start(file); |
|
1713 |
uint max_event_size= current_thd->variables.max_allowed_packet; |
|
1714 |
lf_info= (LOAD_FILE_INFO*) file->arg; |
|
1715 |
if (lf_info->thd->current_stmt_binlog_row_based) |
|
1716 |
DBUG_RETURN(0); |
|
1717 |
if (lf_info->last_pos_in_file != HA_POS_ERROR && |
|
1718 |
lf_info->last_pos_in_file >= my_b_get_pos_in_file(file)) |
|
1719 |
DBUG_RETURN(0); |
|
1720 |
||
1721 |
for (block_len= my_b_get_bytes_in_buffer(file); block_len > 0; |
|
1722 |
buffer += min(block_len, max_event_size), |
|
1723 |
block_len -= min(block_len, max_event_size)) |
|
1724 |
{
|
|
1725 |
lf_info->last_pos_in_file= my_b_get_pos_in_file(file); |
|
1726 |
if (lf_info->wrote_create_file) |
|
1727 |
{
|
|
1728 |
Append_block_log_event a(lf_info->thd, lf_info->thd->db, buffer, |
|
1729 |
min(block_len, max_event_size), |
|
1730 |
lf_info->log_delayed); |
|
1731 |
mysql_bin_log.write(&a); |
|
1732 |
}
|
|
1733 |
else
|
|
1734 |
{
|
|
1735 |
Begin_load_query_log_event b(lf_info->thd, lf_info->thd->db, |
|
1736 |
buffer, |
|
1737 |
min(block_len, max_event_size), |
|
1738 |
lf_info->log_delayed); |
|
1739 |
mysql_bin_log.write(&b); |
|
1740 |
lf_info->wrote_create_file= 1; |
|
1741 |
DBUG_SYNC_POINT("debug_lock.created_file_event",10); |
|
1742 |
}
|
|
1743 |
}
|
|
1744 |
DBUG_RETURN(0); |
|
1745 |
}
|
|
1746 |
||
1747 |
/*
|
|
1748 |
Replication System Variables
|
|
1749 |
*/
|
|
1750 |
||
1751 |
class sys_var_slave_skip_counter :public sys_var |
|
1752 |
{
|
|
1753 |
public: |
|
1754 |
sys_var_slave_skip_counter(sys_var_chain *chain, const char *name_arg) |
|
1755 |
:sys_var(name_arg) |
|
1756 |
{ chain_sys_var(chain); } |
|
1757 |
bool check(THD *thd, set_var *var); |
|
1758 |
bool update(THD *thd, set_var *var); |
|
1759 |
bool check_type(enum_var_type type) { return type != OPT_GLOBAL; } |
|
1760 |
/*
|
|
1761 |
We can't retrieve the value of this, so we don't have to define
|
|
1762 |
type() or value_ptr()
|
|
1763 |
*/
|
|
1764 |
};
|
|
1765 |
||
1766 |
class sys_var_sync_binlog_period :public sys_var_long_ptr |
|
1767 |
{
|
|
1768 |
public: |
|
1769 |
sys_var_sync_binlog_period(sys_var_chain *chain, const char *name_arg, |
|
1770 |
ulong *value_ptr) |
|
1771 |
:sys_var_long_ptr(chain, name_arg,value_ptr) {} |
|
1772 |
bool update(THD *thd, set_var *var); |
|
1773 |
};
|
|
1774 |
||
1775 |
static void fix_slave_net_timeout(THD *thd, enum_var_type type) |
|
1776 |
{
|
|
1777 |
DBUG_ENTER("fix_slave_net_timeout"); |
|
1778 |
#ifdef HAVE_REPLICATION
|
|
1779 |
pthread_mutex_lock(&LOCK_active_mi); |
|
1780 |
if (active_mi && slave_net_timeout < active_mi->heartbeat_period) |
|
1781 |
push_warning_printf(thd, MYSQL_ERROR::WARN_LEVEL_WARN, |
|
1782 |
ER_SLAVE_HEARTBEAT_VALUE_OUT_OF_RANGE, |
|
1783 |
"The currect value for master_heartbeat_period"
|
|
1784 |
" exceeds the new value of `slave_net_timeout' sec."
|
|
1785 |
" A sensible value for the period should be"
|
|
1786 |
" less than the timeout."); |
|
1787 |
pthread_mutex_unlock(&LOCK_active_mi); |
|
1788 |
#endif
|
|
1789 |
DBUG_VOID_RETURN; |
|
1790 |
}
|
|
1791 |
||
1792 |
static sys_var_chain vars = { NULL, NULL }; |
|
1793 |
||
1794 |
static sys_var_bool_ptr sys_relay_log_purge(&vars, "relay_log_purge", |
|
1795 |
&relay_log_purge); |
|
1796 |
static sys_var_long_ptr sys_slave_net_timeout(&vars, "slave_net_timeout", |
|
1797 |
&slave_net_timeout, |
|
1798 |
fix_slave_net_timeout); |
|
1799 |
static sys_var_long_ptr sys_slave_trans_retries(&vars, "slave_transaction_retries", |
|
1800 |
&slave_trans_retries); |
|
1801 |
static sys_var_sync_binlog_period sys_sync_binlog_period(&vars, "sync_binlog", &sync_binlog_period); |
|
1802 |
static sys_var_slave_skip_counter sys_slave_skip_counter(&vars, "sql_slave_skip_counter"); |
|
1803 |
||
1804 |
static int show_slave_skip_errors(THD *thd, SHOW_VAR *var, char *buff); |
|
1805 |
||
1806 |
||
1807 |
static SHOW_VAR fixed_vars[]= { |
|
1808 |
{"log_slave_updates", (char*) &opt_log_slave_updates, SHOW_MY_BOOL}, |
|
1809 |
{"relay_log" , (char*) &opt_relay_logname, SHOW_CHAR_PTR}, |
|
1810 |
{"relay_log_index", (char*) &opt_relaylog_index_name, SHOW_CHAR_PTR}, |
|
1811 |
{"relay_log_info_file", (char*) &relay_log_info_file, SHOW_CHAR_PTR}, |
|
1812 |
{"relay_log_space_limit", (char*) &relay_log_space_limit, SHOW_LONGLONG}, |
|
1813 |
{"slave_load_tmpdir", (char*) &slave_load_tmpdir, SHOW_CHAR_PTR}, |
|
1814 |
{"slave_skip_errors", (char*) &show_slave_skip_errors, SHOW_FUNC}, |
|
1815 |
};
|
|
1816 |
||
1817 |
static int show_slave_skip_errors(THD *thd, SHOW_VAR *var, char *buff) |
|
1818 |
{
|
|
1819 |
var->type=SHOW_CHAR; |
|
1820 |
var->value= buff; |
|
1821 |
if (!use_slave_mask || bitmap_is_clear_all(&slave_error_mask)) |
|
1822 |
{
|
|
1823 |
var->value= const_cast<char *>("OFF"); |
|
1824 |
}
|
|
1825 |
else if (bitmap_is_set_all(&slave_error_mask)) |
|
1826 |
{
|
|
1827 |
var->value= const_cast<char *>("ALL"); |
|
1828 |
}
|
|
1829 |
else
|
|
1830 |
{
|
|
1831 |
/* 10 is enough assuming errors are max 4 digits */
|
|
1832 |
int i; |
|
1833 |
var->value= buff; |
|
1834 |
for (i= 1; |
|
1835 |
i < MAX_SLAVE_ERROR && |
|
1836 |
(buff - var->value) < SHOW_VAR_FUNC_BUFF_SIZE; |
|
1837 |
i++) |
|
1838 |
{
|
|
1839 |
if (bitmap_is_set(&slave_error_mask, i)) |
|
1840 |
{
|
|
1841 |
buff= int10_to_str(i, buff, 10); |
|
1842 |
*buff++= ','; |
|
1843 |
}
|
|
1844 |
}
|
|
1845 |
if (var->value != buff) |
|
1846 |
buff--; // Remove last ',' |
|
1847 |
if (i < MAX_SLAVE_ERROR) |
|
1848 |
buff= strmov(buff, "..."); // Couldn't show all errors |
|
1849 |
*buff=0; |
|
1850 |
}
|
|
1851 |
return 0; |
|
1852 |
}
|
|
1853 |
||
1854 |
bool sys_var_slave_skip_counter::check(THD *thd, set_var *var) |
|
1855 |
{
|
|
1856 |
int result= 0; |
|
1857 |
pthread_mutex_lock(&LOCK_active_mi); |
|
1858 |
pthread_mutex_lock(&active_mi->rli.run_lock); |
|
1859 |
if (active_mi->rli.slave_running) |
|
1860 |
{
|
|
1861 |
my_message(ER_SLAVE_MUST_STOP, ER(ER_SLAVE_MUST_STOP), MYF(0)); |
|
1862 |
result=1; |
|
1863 |
}
|
|
1864 |
pthread_mutex_unlock(&active_mi->rli.run_lock); |
|
1865 |
pthread_mutex_unlock(&LOCK_active_mi); |
|
1866 |
var->save_result.ulong_value= (ulong) var->value->val_int(); |
|
1867 |
return result; |
|
1868 |
}
|
|
1869 |
||
1870 |
||
1871 |
bool sys_var_slave_skip_counter::update(THD *thd, set_var *var) |
|
1872 |
{
|
|
1873 |
pthread_mutex_lock(&LOCK_active_mi); |
|
1874 |
pthread_mutex_lock(&active_mi->rli.run_lock); |
|
1875 |
/*
|
|
1876 |
The following test should normally never be true as we test this
|
|
1877 |
in the check function; To be safe against multiple
|
|
1878 |
SQL_SLAVE_SKIP_COUNTER request, we do the check anyway
|
|
1879 |
*/
|
|
1880 |
if (!active_mi->rli.slave_running) |
|
1881 |
{
|
|
1882 |
pthread_mutex_lock(&active_mi->rli.data_lock); |
|
1883 |
active_mi->rli.slave_skip_counter= var->save_result.ulong_value; |
|
1884 |
pthread_mutex_unlock(&active_mi->rli.data_lock); |
|
1885 |
}
|
|
1886 |
pthread_mutex_unlock(&active_mi->rli.run_lock); |
|
1887 |
pthread_mutex_unlock(&LOCK_active_mi); |
|
1888 |
return 0; |
|
1889 |
}
|
|
1890 |
||
1891 |
||
1892 |
bool sys_var_sync_binlog_period::update(THD *thd, set_var *var) |
|
1893 |
{
|
|
1894 |
sync_binlog_period= (ulong) var->save_result.ulonglong_value; |
|
1895 |
return 0; |
|
1896 |
}
|
|
1897 |
||
1898 |
int init_replication_sys_vars() |
|
1899 |
{
|
|
1900 |
mysql_append_static_vars(fixed_vars, sizeof(fixed_vars) / sizeof(SHOW_VAR)); |
|
1901 |
||
1902 |
if (mysql_add_sys_var_chain(vars.first, my_long_options)) |
|
1903 |
{
|
|
1904 |
/* should not happen */
|
|
1905 |
fprintf(stderr, "failed to initialize replication system variables"); |
|
1906 |
unireg_abort(1); |
|
1907 |
}
|
|
1908 |
return 0; |
|
1909 |
}
|
|
1910 |
||
1911 |
#endif /* HAVE_REPLICATION */ |
|
1912 |
||
1913 |