/* Copyright (C) 2000-2003 MySQL AB This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ #include "mysql_priv.h" #ifdef HAVE_REPLICATION #include <mysql.h> #include <myisam.h> #include "slave.h" #include "sql_repl.h" #include "repl_failsafe.h" #include <thr_alarm.h> #include <my_dir.h> #include <sql_common.h> bool use_slave_mask = 0; MY_BITMAP slave_error_mask; typedef bool (*CHECK_KILLED_FUNC)(THD*,void*); volatile bool slave_sql_running = 0, slave_io_running = 0; char* slave_load_tmpdir = 0; MASTER_INFO *active_mi; volatile int active_mi_in_use = 0; HASH replicate_do_table, replicate_ignore_table; DYNAMIC_ARRAY replicate_wild_do_table, replicate_wild_ignore_table; bool do_table_inited = 0, ignore_table_inited = 0; bool wild_do_table_inited = 0, wild_ignore_table_inited = 0; bool table_rules_on = 0; ulonglong relay_log_space_limit = 0; /* When slave thread exits, we need to remember the temporary tables so we can re-use them on slave start. TODO: move the vars below under MASTER_INFO */ int disconnect_slave_event_count = 0, abort_slave_event_count = 0; int events_till_abort = -1; static int events_till_disconnect = -1; typedef enum { SLAVE_THD_IO, SLAVE_THD_SQL} SLAVE_THD_TYPE; static int process_io_rotate(MASTER_INFO* mi, Rotate_log_event* rev); static int process_io_create_file(MASTER_INFO* mi, Create_file_log_event* cev); static bool wait_for_relay_log_space(RELAY_LOG_INFO* rli); static inline bool io_slave_killed(THD* thd,MASTER_INFO* mi); static inline bool sql_slave_killed(THD* thd,RELAY_LOG_INFO* rli); static int count_relay_log_space(RELAY_LOG_INFO* rli); static int init_slave_thread(THD* thd, SLAVE_THD_TYPE thd_type); static int safe_connect(THD* thd, MYSQL* mysql, MASTER_INFO* mi); static int safe_reconnect(THD* thd, MYSQL* mysql, MASTER_INFO* mi, bool suppress_warnings); static int connect_to_master(THD* thd, MYSQL* mysql, MASTER_INFO* mi, bool reconnect, bool suppress_warnings); static int safe_sleep(THD* thd, int sec, CHECK_KILLED_FUNC thread_killed, void* thread_killed_arg); static int request_table_dump(MYSQL* mysql, const char* db, const char* table); static int create_table_from_dump(THD* thd, MYSQL *mysql, const char* db, const char* table_name); static int check_master_version(MYSQL* mysql, MASTER_INFO* mi); char* rewrite_db(char* db); /* Find out which replications threads are running SYNOPSIS init_thread_mask() mask Return value here mi master_info for slave inverse If set, returns which threads are not running IMPLEMENTATION Get a bit mask for which threads are running so that we can later restart these threads. RETURN mask If inverse == 0, running threads If inverse == 1, stopped threads */ void init_thread_mask(int* mask,MASTER_INFO* mi,bool inverse) { bool set_io = mi->slave_running, set_sql = mi->rli.slave_running; register int tmp_mask=0; if (set_io) tmp_mask |= SLAVE_IO; if (set_sql) tmp_mask |= SLAVE_SQL; if (inverse) tmp_mask^= (SLAVE_IO | SLAVE_SQL); *mask = tmp_mask; } /* lock_slave_threads() */ void lock_slave_threads(MASTER_INFO* mi) { //TODO: see if we can do this without dual mutex pthread_mutex_lock(&mi->run_lock); pthread_mutex_lock(&mi->rli.run_lock); } /* unlock_slave_threads() */ void unlock_slave_threads(MASTER_INFO* mi) { //TODO: see if we can do this without dual mutex pthread_mutex_unlock(&mi->rli.run_lock); pthread_mutex_unlock(&mi->run_lock); } /* Initialize slave structures */ int init_slave() { DBUG_ENTER("init_slave"); /* TODO: re-write this to interate through the list of files for multi-master */ active_mi= new MASTER_INFO; /* If master_host is not specified, try to read it from the master_info file. If master_host is specified, create the master_info file if it doesn't exists. */ if (!active_mi || init_master_info(active_mi,master_info_file,relay_log_info_file, !master_host)) { sql_print_error("Note: Failed to initialized master info"); goto err; } /* make sure slave thread gets started if server_id is set, valid master.info is present, and master_host has not been specified */ if (server_id && !master_host && active_mi->host[0]) master_host= active_mi->host; if (master_host && !opt_skip_slave_start) { if (start_slave_threads(1 /* need mutex */, 0 /* no wait for start*/, active_mi, master_info_file, relay_log_info_file, SLAVE_IO | SLAVE_SQL)) { sql_print_error("Warning: Can't create threads to handle slave"); goto err; } } DBUG_RETURN(0); err: DBUG_RETURN(1); } static void free_table_ent(TABLE_RULE_ENT* e) { my_free((gptr) e, MYF(0)); } static byte* get_table_key(TABLE_RULE_ENT* e, uint* len, my_bool not_used __attribute__((unused))) { *len = e->key_len; return (byte*)e->db; } /* Open the given relay log SYNOPSIS init_relay_log_pos() rli Relay information (will be initialized) log Name of relay log file to read from. NULL = First log pos Position in relay log file need_data_lock Set to 1 if this functions should do mutex locks errmsg Store pointer to error message here DESCRIPTION - Close old open relay log files. - If we are using the same relay log as the running IO-thread, then set rli->cur_log to point to the same IO_CACHE entry. - If not, open the 'log' binary file. TODO - check proper initialization of group_master_log_name/group_master_log_pos RETURN VALUES 0 ok 1 error. errmsg is set to point to the error message */ int init_relay_log_pos(RELAY_LOG_INFO* rli,const char* log, ulonglong pos, bool need_data_lock, const char** errmsg) { DBUG_ENTER("init_relay_log_pos"); *errmsg=0; pthread_mutex_t *log_lock=rli->relay_log.get_log_lock(); pthread_mutex_lock(log_lock); if (need_data_lock) pthread_mutex_lock(&rli->data_lock); /* Close log file and free buffers if it's already open */ if (rli->cur_log_fd >= 0) { end_io_cache(&rli->cache_buf); my_close(rli->cur_log_fd, MYF(MY_WME)); rli->cur_log_fd = -1; } rli->group_relay_log_pos = rli->event_relay_log_pos = pos; /* Test to see if the previous run was with the skip of purging If yes, we do not purge when we restart */ if (rli->relay_log.find_log_pos(&rli->linfo, NullS, 1)) { *errmsg="Could not find first log during relay log initialization"; goto err; } if (log && rli->relay_log.find_log_pos(&rli->linfo, log, 1)) { *errmsg="Could not find target log during relay log initialization"; goto err; } strmake(rli->group_relay_log_name,rli->linfo.log_file_name, sizeof(rli->group_relay_log_name)-1); strmake(rli->event_relay_log_name,rli->linfo.log_file_name, sizeof(rli->event_relay_log_name)-1); if (rli->relay_log.is_active(rli->linfo.log_file_name)) { /* The IO thread is using this log file. In this case, we will use the same IO_CACHE pointer to read data as the IO thread is using to write data. */ if (my_b_tell((rli->cur_log=rli->relay_log.get_log_file())) == 0 && check_binlog_magic(rli->cur_log,errmsg)) goto err; rli->cur_log_old_open_count=rli->relay_log.get_open_count(); } else { /* Open the relay log and set rli->cur_log to point at this one */ if ((rli->cur_log_fd=open_binlog(&rli->cache_buf, rli->linfo.log_file_name,errmsg)) < 0) goto err; rli->cur_log = &rli->cache_buf; } if (pos >= BIN_LOG_HEADER_SIZE) my_b_seek(rli->cur_log,(off_t)pos); err: /* If we don't purge, we can't honour relay_log_space_limit ; silently discard it */ if (!relay_log_purge) rli->log_space_limit= 0; pthread_cond_broadcast(&rli->data_cond); if (need_data_lock) pthread_mutex_unlock(&rli->data_lock); pthread_mutex_unlock(log_lock); DBUG_RETURN ((*errmsg) ? 1 : 0); } /* Init functio to set up array for errors that should be skipped for slave SYNOPSIS init_slave_skip_errors() arg List of errors numbers to skip, separated with ',' NOTES Called from get_options() in mysqld.cc on start-up */ void init_slave_skip_errors(const char* arg) { const char *p; if (bitmap_init(&slave_error_mask,MAX_SLAVE_ERROR,0)) { fprintf(stderr, "Badly out of memory, please check your system status\n"); exit(1); } use_slave_mask = 1; for (;my_isspace(system_charset_info,*arg);++arg) /* empty */; if (!my_strnncoll(system_charset_info,(uchar*)arg,4,(const uchar*)"all",4)) { bitmap_set_all(&slave_error_mask); return; } for (p= arg ; *p; ) { long err_code; if (!(p= str2int(p, 10, 0, LONG_MAX, &err_code))) break; if (err_code < MAX_SLAVE_ERROR) bitmap_set_bit(&slave_error_mask,(uint)err_code); while (!my_isdigit(system_charset_info,*p) && *p) p++; } } /* purge_relay_logs() NOTES Assumes to have a run lock on rli and that no slave thread are running. */ int purge_relay_logs(RELAY_LOG_INFO* rli, THD *thd, bool just_reset, const char** errmsg) { int error=0; DBUG_ENTER("purge_relay_logs"); /* Even if rli->inited==0, we still try to empty rli->master_log_* variables. Indeed, rli->inited==0 does not imply that they already are empty. It could be that slave's info initialization partly succeeded : for example if relay-log.info existed but *relay-bin*.* have been manually removed, init_relay_log_info reads the old relay-log.info and fills rli->master_log_*, then init_relay_log_info checks for the existence of the relay log, this fails and init_relay_log_info leaves rli->inited to 0. In that pathological case, rli->master_log_pos* will be properly reinited at the next START SLAVE (as RESET SLAVE or CHANGE MASTER, the callers of purge_relay_logs, will delete bogus *.info files or replace them with correct files), however if the user does SHOW SLAVE STATUS before START SLAVE, he will see old, confusing rli->master_log_*. In other words, we reinit rli->master_log_* for SHOW SLAVE STATUS to display fine in any case. */ rli->group_master_log_name[0]= 0; rli->group_master_log_pos= 0; if (!rli->inited) DBUG_RETURN(0); DBUG_ASSERT(rli->slave_running == 0); DBUG_ASSERT(rli->mi->slave_running == 0); rli->slave_skip_counter=0; pthread_mutex_lock(&rli->data_lock); if (rli->relay_log.reset_logs(thd)) { *errmsg = "Failed during log reset"; error=1; goto err; } /* Save name of used relay log file */ strmake(rli->group_relay_log_name, rli->relay_log.get_log_fname(), sizeof(rli->group_relay_log_name)-1); strmake(rli->event_relay_log_name, rli->relay_log.get_log_fname(), sizeof(rli->event_relay_log_name)-1); // Just first log with magic number and nothing else rli->log_space_total= BIN_LOG_HEADER_SIZE; rli->group_relay_log_pos= rli->event_relay_log_pos= BIN_LOG_HEADER_SIZE; rli->relay_log.reset_bytes_written(); if (!just_reset) error= init_relay_log_pos(rli, rli->group_relay_log_name, rli->group_relay_log_pos, 0 /* do not need data lock */, errmsg); err: #ifndef DBUG_OFF char buf[22]; #endif DBUG_PRINT("info",("log_space_total: %s",llstr(rli->log_space_total,buf))); pthread_mutex_unlock(&rli->data_lock); DBUG_RETURN(error); } int terminate_slave_threads(MASTER_INFO* mi,int thread_mask,bool skip_lock) { if (!mi->inited) return 0; /* successfully do nothing */ int error,force_all = (thread_mask & SLAVE_FORCE_ALL); pthread_mutex_t *sql_lock = &mi->rli.run_lock, *io_lock = &mi->run_lock; pthread_mutex_t *sql_cond_lock,*io_cond_lock; DBUG_ENTER("terminate_slave_threads"); sql_cond_lock=sql_lock; io_cond_lock=io_lock; if (skip_lock) { sql_lock = io_lock = 0; } if ((thread_mask & (SLAVE_IO|SLAVE_FORCE_ALL)) && mi->slave_running) { DBUG_PRINT("info",("Terminating IO thread")); mi->abort_slave=1; if ((error=terminate_slave_thread(mi->io_thd,io_lock, io_cond_lock, &mi->stop_cond, &mi->slave_running)) && !force_all) DBUG_RETURN(error); } if ((thread_mask & (SLAVE_SQL|SLAVE_FORCE_ALL)) && mi->rli.slave_running) { DBUG_PRINT("info",("Terminating SQL thread")); DBUG_ASSERT(mi->rli.sql_thd != 0) ; mi->rli.abort_slave=1; if ((error=terminate_slave_thread(mi->rli.sql_thd,sql_lock, sql_cond_lock, &mi->rli.stop_cond, &mi->rli.slave_running)) && !force_all) DBUG_RETURN(error); } DBUG_RETURN(0); } int terminate_slave_thread(THD* thd, pthread_mutex_t* term_lock, pthread_mutex_t *cond_lock, pthread_cond_t* term_cond, volatile bool* slave_running) { if (term_lock) { pthread_mutex_lock(term_lock); if (!*slave_running) { pthread_mutex_unlock(term_lock); return ER_SLAVE_NOT_RUNNING; } } DBUG_ASSERT(thd != 0); /* Is is criticate to test if the slave is running. Otherwise, we might be referening freed memory trying to kick it */ THD_CHECK_SENTRY(thd); while (*slave_running) // Should always be true { KICK_SLAVE(thd); /* There is a small chance that slave thread might miss the first alarm. To protect againts it, resend the signal until it reacts */ struct timespec abstime; set_timespec(abstime,2); pthread_cond_timedwait(term_cond, cond_lock, &abstime); } if (term_lock) pthread_mutex_unlock(term_lock); return 0; } int start_slave_thread(pthread_handler h_func, pthread_mutex_t *start_lock, pthread_mutex_t *cond_lock, pthread_cond_t *start_cond, volatile bool *slave_running, volatile ulong *slave_run_id, MASTER_INFO* mi, bool high_priority) { pthread_t th; ulong start_id; DBUG_ASSERT(mi->inited); DBUG_ENTER("start_slave_thread"); if (start_lock) pthread_mutex_lock(start_lock); if (!server_id) { if (start_cond) pthread_cond_broadcast(start_cond); if (start_lock) pthread_mutex_unlock(start_lock); sql_print_error("Server id not set, will not start slave"); DBUG_RETURN(ER_BAD_SLAVE); } if (*slave_running) { if (start_cond) pthread_cond_broadcast(start_cond); if (start_lock) pthread_mutex_unlock(start_lock); DBUG_RETURN(ER_SLAVE_MUST_STOP); } start_id= *slave_run_id; DBUG_PRINT("info",("Creating new slave thread")); if (high_priority) my_pthread_attr_setprio(&connection_attrib,CONNECT_PRIOR); if (pthread_create(&th, &connection_attrib, h_func, (void*)mi)) { if (start_lock) pthread_mutex_unlock(start_lock); DBUG_RETURN(ER_SLAVE_THREAD); } if (start_cond && cond_lock) { THD* thd = current_thd; while (start_id == *slave_run_id) { DBUG_PRINT("sleep",("Waiting for slave thread to start")); const char* old_msg = thd->enter_cond(start_cond,cond_lock, "Waiting for slave thread to start"); pthread_cond_wait(start_cond,cond_lock); thd->exit_cond(old_msg); if (thd->killed) { pthread_mutex_unlock(cond_lock); DBUG_RETURN(ER_SERVER_SHUTDOWN); } } } if (start_lock) pthread_mutex_unlock(start_lock); DBUG_RETURN(0); } /* start_slave_threads() NOTES SLAVE_FORCE_ALL is not implemented here on purpose since it does not make sense to do that for starting a slave--we always care if it actually started the threads that were not previously running */ int start_slave_threads(bool need_slave_mutex, bool wait_for_start, MASTER_INFO* mi, const char* master_info_fname, const char* slave_info_fname, int thread_mask) { pthread_mutex_t *lock_io=0,*lock_sql=0,*lock_cond_io=0,*lock_cond_sql=0; pthread_cond_t* cond_io=0,*cond_sql=0; int error=0; DBUG_ENTER("start_slave_threads"); if (need_slave_mutex) { lock_io = &mi->run_lock; lock_sql = &mi->rli.run_lock; } if (wait_for_start) { cond_io = &mi->start_cond; cond_sql = &mi->rli.start_cond; lock_cond_io = &mi->run_lock; lock_cond_sql = &mi->rli.run_lock; } if (thread_mask & SLAVE_IO) error=start_slave_thread(handle_slave_io,lock_io,lock_cond_io, cond_io, &mi->slave_running, &mi->slave_run_id, mi, 1); //high priority, to read the most possible if (!error && (thread_mask & SLAVE_SQL)) { error=start_slave_thread(handle_slave_sql,lock_sql,lock_cond_sql, cond_sql, &mi->rli.slave_running, &mi->rli.slave_run_id, mi, 0); if (error) terminate_slave_threads(mi, thread_mask & SLAVE_IO, 0); } DBUG_RETURN(error); } void init_table_rule_hash(HASH* h, bool* h_inited) { hash_init(h, system_charset_info,TABLE_RULE_HASH_SIZE,0,0, (hash_get_key) get_table_key, (hash_free_key) free_table_ent, 0); *h_inited = 1; } void init_table_rule_array(DYNAMIC_ARRAY* a, bool* a_inited) { my_init_dynamic_array(a, sizeof(TABLE_RULE_ENT*), TABLE_RULE_ARR_SIZE, TABLE_RULE_ARR_SIZE); *a_inited = 1; } static TABLE_RULE_ENT* find_wild(DYNAMIC_ARRAY *a, const char* key, int len) { uint i; const char* key_end = key + len; for (i = 0; i < a->elements; i++) { TABLE_RULE_ENT* e ; get_dynamic(a, (gptr)&e, i); if (!my_wildcmp(system_charset_info, key, key_end, (const char*)e->db, (const char*)(e->db + e->key_len), '\\',wild_one,wild_many)) return e; } return 0; } /* Checks whether tables match some (wild_)do_table and (wild_)ignore_table rules (for replication) SYNOPSIS tables_ok() thd thread (SQL slave thread normally) tables list of tables to check NOTES Note that changing the order of the tables in the list can lead to different results. Note also the order of precedence of the do/ignore rules (see code below). For that reason, users should not set conflicting rules because they may get unpredicted results. RETURN VALUES 0 should not be logged/replicated 1 should be logged/replicated */ int tables_ok(THD* thd, TABLE_LIST* tables) { DBUG_ENTER("tables_ok"); for (; tables; tables = tables->next) { char hash_key[2*NAME_LEN+2]; char *end; uint len; if (!tables->updating) continue; end= strmov(hash_key, tables->db ? tables->db : thd->db); *end++= '.'; len= (uint) (strmov(end, tables->real_name) - hash_key); if (do_table_inited) // if there are any do's { if (hash_search(&replicate_do_table, (byte*) hash_key, len)) DBUG_RETURN(1); } if (ignore_table_inited) // if there are any ignores { if (hash_search(&replicate_ignore_table, (byte*) hash_key, len)) DBUG_RETURN(0); } if (wild_do_table_inited && find_wild(&replicate_wild_do_table, hash_key, len)) DBUG_RETURN(1); if (wild_ignore_table_inited && find_wild(&replicate_wild_ignore_table, hash_key, len)) DBUG_RETURN(0); } /* If no explicit rule found and there was a do list, do not replicate. If there was no do list, go ahead */ DBUG_RETURN(!do_table_inited && !wild_do_table_inited); } /* Checks whether a db matches wild_do_table and wild_ignore_table rules (for replication) SYNOPSIS db_ok_with_wild_table() db name of the db to check. Is tested with check_db_name() before calling this function. NOTES Here is the reason for this function. We advise users who want to exclude a database 'db1' safely to do it with replicate_wild_ignore_table='db1.%' instead of binlog_ignore_db or replicate_ignore_db because the two lasts only check for the selected db, which won't work in that case: USE db2; UPDATE db1.t SET ... #this will be replicated and should not whereas replicate_wild_ignore_table will work in all cases. With replicate_wild_ignore_table, we only check tables. When one does 'DROP DATABASE db1', tables are not involved and the statement will be replicated, while users could expect it would not (as it rougly means 'DROP db1.first_table, DROP db1.second_table...'). In other words, we want to interpret 'db1.%' as "everything touching db1". That is why we want to match 'db1' against 'db1.%' wild table rules. RETURN VALUES 0 should not be logged/replicated 1 should be logged/replicated */ int db_ok_with_wild_table(const char *db) { char hash_key[NAME_LEN+2]; char *end; int len; end= strmov(hash_key, db); *end++= '.'; len= end - hash_key ; if (wild_do_table_inited && find_wild(&replicate_wild_do_table, hash_key, len)) return 1; if (wild_ignore_table_inited && find_wild(&replicate_wild_ignore_table, hash_key, len)) return 0; /* If no explicit rule found and there was a do list, do not replicate. If there was no do list, go ahead */ return !wild_do_table_inited; } int add_table_rule(HASH* h, const char* table_spec) { const char* dot = strchr(table_spec, '.'); if (!dot) return 1; // len is always > 0 because we know the there exists a '.' uint len = (uint)strlen(table_spec); TABLE_RULE_ENT* e = (TABLE_RULE_ENT*)my_malloc(sizeof(TABLE_RULE_ENT) + len, MYF(MY_WME)); if (!e) return 1; e->db = (char*)e + sizeof(TABLE_RULE_ENT); e->tbl_name = e->db + (dot - table_spec) + 1; e->key_len = len; memcpy(e->db, table_spec, len); (void)hash_insert(h, (byte*)e); return 0; } /* Add table expression with wildcards to dynamic array */ int add_wild_table_rule(DYNAMIC_ARRAY* a, const char* table_spec) { const char* dot = strchr(table_spec, '.'); if (!dot) return 1; uint len = (uint)strlen(table_spec); TABLE_RULE_ENT* e = (TABLE_RULE_ENT*)my_malloc(sizeof(TABLE_RULE_ENT) + len, MYF(MY_WME)); if (!e) return 1; e->db = (char*)e + sizeof(TABLE_RULE_ENT); e->tbl_name = e->db + (dot - table_spec) + 1; e->key_len = len; memcpy(e->db, table_spec, len); insert_dynamic(a, (gptr)&e); return 0; } static void free_string_array(DYNAMIC_ARRAY *a) { uint i; for (i = 0; i < a->elements; i++) { char* p; get_dynamic(a, (gptr) &p, i); my_free(p, MYF(MY_WME)); } delete_dynamic(a); } #ifdef NOT_USED_YET static int end_slave_on_walk(MASTER_INFO* mi, gptr /*unused*/) { end_master_info(mi); return 0; } #endif /* Free all resources used by slave SYNOPSIS end_slave() */ void end_slave() { if (active_mi) { /* TODO: replace the line below with list_walk(&master_list, (list_walk_action)end_slave_on_walk,0); once multi-master code is ready. */ terminate_slave_threads(active_mi,SLAVE_FORCE_ALL); end_master_info(active_mi); if (do_table_inited) hash_free(&replicate_do_table); if (ignore_table_inited) hash_free(&replicate_ignore_table); if (wild_do_table_inited) free_string_array(&replicate_wild_do_table); if (wild_ignore_table_inited) free_string_array(&replicate_wild_ignore_table); delete active_mi; active_mi= 0; } } static bool io_slave_killed(THD* thd, MASTER_INFO* mi) { DBUG_ASSERT(mi->io_thd == thd); DBUG_ASSERT(mi->slave_running == 1); // tracking buffer overrun return mi->abort_slave || abort_loop || thd->killed; } static bool sql_slave_killed(THD* thd, RELAY_LOG_INFO* rli) { DBUG_ASSERT(rli->sql_thd == thd); DBUG_ASSERT(rli->slave_running == 1);// tracking buffer overrun return rli->abort_slave || abort_loop || thd->killed; } void slave_print_error(RELAY_LOG_INFO* rli, int err_code, const char* msg, ...) { va_list args; va_start(args,msg); my_vsnprintf(rli->last_slave_error, sizeof(rli->last_slave_error), msg, args); sql_print_error("Slave: %s, error_code=%d", rli->last_slave_error, err_code); rli->last_slave_errno = err_code; } /* skip_load_data_infile() NOTES This is used to tell a 3.23 master to break send_file() */ void skip_load_data_infile(NET *net) { (void)net_request_file(net, "/dev/null"); (void)my_net_read(net); // discard response (void)net_write_command(net, 0, "", 0, "", 0); // Send ok } bool net_request_file(NET* net, const char* fname) { DBUG_ENTER("net_request_file"); DBUG_RETURN(net_write_command(net, 251, fname, strlen(fname), "", 0)); } char* rewrite_db(char* db) { if (replicate_rewrite_db.is_empty() || !db) return db; I_List_iterator<i_string_pair> it(replicate_rewrite_db); i_string_pair* tmp; while ((tmp=it++)) { if (!strcmp(tmp->key, db)) return tmp->val; } return db; } /* Checks whether a db matches some do_db and ignore_db rules (for logging or replication) SYNOPSIS db_ok() db name of the db to check do_list either binlog_do_db or replicate_do_db ignore_list either binlog_ignore_db or replicate_ignore_db RETURN VALUES 0 should not be logged/replicated 1 should be logged/replicated */ int db_ok(const char* db, I_List<i_string> &do_list, I_List<i_string> &ignore_list ) { if (do_list.is_empty() && ignore_list.is_empty()) return 1; // ok to replicate if the user puts no constraints /* If the user has specified restrictions on which databases to replicate and db was not selected, do not replicate. */ if (!db) return 0; if (!do_list.is_empty()) // if the do's are not empty { I_List_iterator<i_string> it(do_list); i_string* tmp; while ((tmp=it++)) { if (!strcmp(tmp->ptr, db)) return 1; // match } return 0; } else // there are some elements in the don't, otherwise we cannot get here { I_List_iterator<i_string> it(ignore_list); i_string* tmp; while ((tmp=it++)) { if (!strcmp(tmp->ptr, db)) return 0; // match } return 1; } } static int init_strvar_from_file(char *var, int max_size, IO_CACHE *f, const char *default_val) { uint length; if ((length=my_b_gets(f,var, max_size))) { char* last_p = var + length -1; if (*last_p == '\n') *last_p = 0; // if we stopped on newline, kill it else { /* If we truncated a line or stopped on last char, remove all chars up to and including newline. */ int c; while (((c=my_b_get(f)) != '\n' && c != my_b_EOF)); } return 0; } else if (default_val) { strmake(var, default_val, max_size-1); return 0; } return 1; } static int init_intvar_from_file(int* var, IO_CACHE* f, int default_val) { char buf[32]; if (my_b_gets(f, buf, sizeof(buf))) { *var = atoi(buf); return 0; } else if (default_val) { *var = default_val; return 0; } return 1; } static int check_master_version(MYSQL* mysql, MASTER_INFO* mi) { const char* errmsg= 0; switch (*mysql->server_version) { case '3': mi->old_format = 1; break; case '4': case '5': mi->old_format = 0; break; default: errmsg = "Master reported unrecognized MySQL version"; break; } if (errmsg) { sql_print_error(errmsg); return 1; } return 0; } static int create_table_from_dump(THD* thd, MYSQL *mysql, const char* db, const char* table_name) { ulong packet_len; char *query; Vio* save_vio; HA_CHECK_OPT check_opt; TABLE_LIST tables; int error= 1; handler *file; ulong save_options; NET *net= &mysql->net; packet_len= my_net_read(net); // read create table statement if (packet_len == packet_error) { send_error(thd, ER_MASTER_NET_READ); return 1; } if (net->read_pos[0] == 255) // error from master { char *err_msg; err_msg= (char*) net->read_pos + ((mysql->server_capabilities & CLIENT_PROTOCOL_41) ? 3+SQLSTATE_LENGTH+1 : 3); net_printf(thd, ER_MASTER, err_msg); return 1; } thd->command = COM_TABLE_DUMP; thd->query_length= packet_len; /* Note that we should not set thd->query until the area is initalized */ if (!(query = thd->strmake((char*) net->read_pos, packet_len))) { sql_print_error("create_table_from_dump: out of memory"); net_printf(thd, ER_GET_ERRNO, "Out of memory"); return 1; } thd->query= query; thd->current_tablenr = 0; thd->query_error = 0; thd->net.no_send_ok = 1; /* we do not want to log create table statement */ save_options = thd->options; thd->options &= ~(ulong) (OPTION_BIN_LOG); thd->proc_info = "Creating table from master dump"; // save old db in case we are creating in a different database char* save_db = thd->db; thd->db = (char*)db; mysql_parse(thd, thd->query, packet_len); // run create table thd->db = save_db; // leave things the way the were before thd->options = save_options; if (thd->query_error) goto err; // mysql_parse took care of the error send bzero((char*) &tables,sizeof(tables)); tables.db = (char*)db; tables.alias= tables.real_name= (char*)table_name; tables.lock_type = TL_WRITE; thd->proc_info = "Opening master dump table"; if (!open_ltable(thd, &tables, TL_WRITE)) { send_error(thd,0,0); // Send error from open_ltable sql_print_error("create_table_from_dump: could not open created table"); goto err; } file = tables.table->file; thd->proc_info = "Reading master dump table data"; if (file->net_read_dump(net)) { net_printf(thd, ER_MASTER_NET_READ); sql_print_error("create_table_from_dump::failed in\ handler::net_read_dump()"); goto err; } check_opt.init(); check_opt.flags|= T_VERY_SILENT | T_CALC_CHECKSUM | T_QUICK; thd->proc_info = "Rebuilding the index on master dump table"; /* We do not want repair() to spam us with messages just send them to the error log, and report the failure in case of problems. */ save_vio = thd->net.vio; thd->net.vio = 0; error=file->repair(thd,&check_opt) != 0; thd->net.vio = save_vio; if (error) net_printf(thd, ER_INDEX_REBUILD,tables.table->real_name); err: close_thread_tables(thd); thd->net.no_send_ok = 0; return error; } int fetch_master_table(THD *thd, const char *db_name, const char *table_name, MASTER_INFO *mi, MYSQL *mysql) { int error= 1; const char *errmsg=0; bool called_connected= (mysql != NULL); DBUG_ENTER("fetch_master_table"); DBUG_PRINT("enter", ("db_name: '%s' table_name: '%s'", db_name,table_name)); if (!called_connected) { if (!(mysql = mysql_init(NULL))) { send_error(thd); // EOM DBUG_RETURN(1); } if (connect_to_master(thd, mysql, mi)) { net_printf(thd, ER_CONNECT_TO_MASTER, mysql_error(mysql)); mysql_close(mysql); DBUG_RETURN(1); } if (thd->killed) goto err; } if (request_table_dump(mysql, db_name, table_name)) { error= ER_UNKNOWN_ERROR; errmsg= "Failed on table dump request"; goto err; } if (create_table_from_dump(thd, mysql, db_name, table_name)) goto err; // create_table_from_dump will have sent the error already error = 0; err: thd->net.no_send_ok = 0; // Clear up garbage after create_table_from_dump if (!called_connected) mysql_close(mysql); if (errmsg && thd->net.vio) send_error(thd, error, errmsg); DBUG_RETURN(test(error)); // Return 1 on error } void end_master_info(MASTER_INFO* mi) { DBUG_ENTER("end_master_info"); if (!mi->inited) DBUG_VOID_RETURN; end_relay_log_info(&mi->rli); if (mi->fd >= 0) { end_io_cache(&mi->file); (void)my_close(mi->fd, MYF(MY_WME)); mi->fd = -1; } mi->inited = 0; DBUG_VOID_RETURN; } int init_relay_log_info(RELAY_LOG_INFO* rli, const char* info_fname) { char fname[FN_REFLEN+128]; int info_fd; const char* msg = 0; int error = 0; DBUG_ENTER("init_relay_log_info"); if (rli->inited) // Set if this function called DBUG_RETURN(0); fn_format(fname, info_fname, mysql_data_home, "", 4+32); pthread_mutex_lock(&rli->data_lock); info_fd = rli->info_fd; rli->cur_log_fd = -1; rli->slave_skip_counter=0; rli->abort_pos_wait=0; rli->log_space_limit= relay_log_space_limit; rli->log_space_total= 0; // TODO: make this work with multi-master if (!opt_relay_logname) { char tmp[FN_REFLEN]; /* TODO: The following should be using fn_format(); We just need to first change fn_format() to cut the file name if it's too long. */ strmake(tmp,glob_hostname,FN_REFLEN-5); strmov(strcend(tmp,'.'),"-relay-bin"); opt_relay_logname=my_strdup(tmp,MYF(MY_WME)); } if (open_log(&rli->relay_log, glob_hostname, opt_relay_logname, "-relay-bin", opt_relaylog_index_name, LOG_BIN, 1 /* read_append cache */, 1 /* no auto events */)) DBUG_RETURN(1); /* if file does not exist */ if (access(fname,F_OK)) { /* If someone removed the file from underneath our feet, just close the old descriptor and re-create the old file */ if (info_fd >= 0) my_close(info_fd, MYF(MY_WME)); if ((info_fd = my_open(fname, O_CREAT|O_RDWR|O_BINARY, MYF(MY_WME))) < 0 || init_io_cache(&rli->info_file, info_fd, IO_SIZE*2, READ_CACHE, 0L,0, MYF(MY_WME))) { msg= current_thd->net.last_error; goto err; } /* Init relay log with first entry in the relay index file */ if (init_relay_log_pos(rli,NullS,BIN_LOG_HEADER_SIZE,0 /* no data lock */, &msg)) goto err; rli->group_master_log_name[0]= 0; rli->group_master_log_pos= 0; rli->info_fd= info_fd; } else // file exists { if (info_fd >= 0) reinit_io_cache(&rli->info_file, READ_CACHE, 0L,0,0); else if ((info_fd = my_open(fname, O_RDWR|O_BINARY, MYF(MY_WME))) < 0 || init_io_cache(&rli->info_file, info_fd, IO_SIZE*2, READ_CACHE, 0L, 0, MYF(MY_WME))) { if (info_fd >= 0) my_close(info_fd, MYF(0)); rli->info_fd= -1; rli->relay_log.close(1); pthread_mutex_unlock(&rli->data_lock); DBUG_RETURN(1); } rli->info_fd = info_fd; int relay_log_pos, master_log_pos; if (init_strvar_from_file(rli->group_relay_log_name, sizeof(rli->group_relay_log_name), &rli->info_file, "") || init_intvar_from_file(&relay_log_pos, &rli->info_file, BIN_LOG_HEADER_SIZE) || init_strvar_from_file(rli->group_master_log_name, sizeof(rli->group_master_log_name), &rli->info_file, "") || init_intvar_from_file(&master_log_pos, &rli->info_file, 0)) { msg="Error reading slave log configuration"; goto err; } strmake(rli->event_relay_log_name,rli->group_relay_log_name, sizeof(rli->event_relay_log_name)-1); rli->group_relay_log_pos= rli->event_relay_log_pos= relay_log_pos; rli->group_master_log_pos= master_log_pos; if (init_relay_log_pos(rli, rli->group_relay_log_name, rli->group_relay_log_pos, 0 /* no data lock*/, &msg)) goto err; } DBUG_ASSERT(rli->event_relay_log_pos >= BIN_LOG_HEADER_SIZE); DBUG_ASSERT(my_b_tell(rli->cur_log) == rli->event_relay_log_pos); /* Now change the cache from READ to WRITE - must do this before flush_relay_log_info */ reinit_io_cache(&rli->info_file, WRITE_CACHE,0L,0,1); error= flush_relay_log_info(rli); if (count_relay_log_space(rli)) { msg="Error counting relay log space"; goto err; } rli->inited= 1; pthread_mutex_unlock(&rli->data_lock); DBUG_RETURN(error); err: sql_print_error(msg); end_io_cache(&rli->info_file); if (info_fd >= 0) my_close(info_fd, MYF(0)); rli->info_fd= -1; rli->relay_log.close(1); pthread_mutex_unlock(&rli->data_lock); DBUG_RETURN(1); } static inline int add_relay_log(RELAY_LOG_INFO* rli,LOG_INFO* linfo) { MY_STAT s; DBUG_ENTER("add_relay_log"); if (!my_stat(linfo->log_file_name,&s,MYF(0))) { sql_print_error("log %s listed in the index, but failed to stat", linfo->log_file_name); DBUG_RETURN(1); } rli->log_space_total += s.st_size; #ifndef DBUG_OFF char buf[22]; DBUG_PRINT("info",("log_space_total: %s", llstr(rli->log_space_total,buf))); #endif DBUG_RETURN(0); } static bool wait_for_relay_log_space(RELAY_LOG_INFO* rli) { bool slave_killed=0; MASTER_INFO* mi = rli->mi; const char* save_proc_info; THD* thd = mi->io_thd; DBUG_ENTER("wait_for_relay_log_space"); pthread_mutex_lock(&rli->log_space_lock); save_proc_info = thd->proc_info; thd->proc_info = "Waiting for relay log space to free"; while (rli->log_space_limit < rli->log_space_total && !(slave_killed=io_slave_killed(thd,mi)) && !rli->ignore_log_space_limit) pthread_cond_wait(&rli->log_space_cond, &rli->log_space_lock); thd->proc_info = save_proc_info; pthread_mutex_unlock(&rli->log_space_lock); DBUG_RETURN(slave_killed); } static int count_relay_log_space(RELAY_LOG_INFO* rli) { LOG_INFO linfo; DBUG_ENTER("count_relay_log_space"); rli->log_space_total= 0; if (rli->relay_log.find_log_pos(&linfo, NullS, 1)) { sql_print_error("Could not find first log while counting relay log space"); DBUG_RETURN(1); } do { if (add_relay_log(rli,&linfo)) DBUG_RETURN(1); } while (!rli->relay_log.find_next_log(&linfo, 1)); DBUG_RETURN(0); } int init_master_info(MASTER_INFO* mi, const char* master_info_fname, const char* slave_info_fname, bool abort_if_no_master_info_file) { int fd,error; char fname[FN_REFLEN+128]; DBUG_ENTER("init_master_info"); if (mi->inited) DBUG_RETURN(0); mi->mysql=0; mi->file_id=1; fn_format(fname, master_info_fname, mysql_data_home, "", 4+32); /* We need a mutex while we are changing master info parameters to keep other threads from reading bogus info */ pthread_mutex_lock(&mi->data_lock); fd = mi->fd; if (access(fname,F_OK)) { if (abort_if_no_master_info_file) { pthread_mutex_unlock(&mi->data_lock); DBUG_RETURN(0); } /* if someone removed the file from underneath our feet, just close the old descriptor and re-create the old file */ if (fd >= 0) my_close(fd, MYF(MY_WME)); if ((fd = my_open(fname, O_CREAT|O_RDWR|O_BINARY, MYF(MY_WME))) < 0 || init_io_cache(&mi->file, fd, IO_SIZE*2, READ_CACHE, 0L,0, MYF(MY_WME))) goto err; mi->master_log_name[0] = 0; mi->master_log_pos = BIN_LOG_HEADER_SIZE; // skip magic number mi->fd = fd; if (master_host) strmake(mi->host, master_host, sizeof(mi->host) - 1); if (master_user) strmake(mi->user, master_user, sizeof(mi->user) - 1); if (master_password) strmake(mi->password, master_password, SCRAMBLED_PASSWORD_CHAR_LENGTH); mi->port = master_port; mi->connect_retry = master_connect_retry; } else // file exists { if (fd >= 0) reinit_io_cache(&mi->file, READ_CACHE, 0L,0,0); else if ((fd = my_open(fname, O_RDWR|O_BINARY, MYF(MY_WME))) < 0 || init_io_cache(&mi->file, fd, IO_SIZE*2, READ_CACHE, 0L, 0, MYF(MY_WME))) goto err; mi->fd = fd; int port, connect_retry, master_log_pos; if (init_strvar_from_file(mi->master_log_name, sizeof(mi->master_log_name), &mi->file, "") || init_intvar_from_file(&master_log_pos, &mi->file, 4) || init_strvar_from_file(mi->host, sizeof(mi->host), &mi->file, master_host) || init_strvar_from_file(mi->user, sizeof(mi->user), &mi->file, master_user) || init_strvar_from_file(mi->password, SCRAMBLED_PASSWORD_CHAR_LENGTH+1, &mi->file, master_password) || init_intvar_from_file(&port, &mi->file, master_port) || init_intvar_from_file(&connect_retry, &mi->file, master_connect_retry)) { sql_print_error("Error reading master configuration"); goto err; } /* This has to be handled here as init_intvar_from_file can't handle my_off_t types */ mi->master_log_pos= (my_off_t) master_log_pos; mi->port= (uint) port; mi->connect_retry= (uint) connect_retry; } DBUG_PRINT("master_info",("log_file_name: %s position: %ld", mi->master_log_name, (ulong) mi->master_log_pos)); if (init_relay_log_info(&mi->rli, slave_info_fname)) goto err; mi->rli.mi = mi; mi->inited = 1; // now change cache READ -> WRITE - must do this before flush_master_info reinit_io_cache(&mi->file, WRITE_CACHE,0L,0,1); error=test(flush_master_info(mi)); pthread_mutex_unlock(&mi->data_lock); DBUG_RETURN(error); err: if (fd >= 0) { my_close(fd, MYF(0)); end_io_cache(&mi->file); } mi->fd= -1; pthread_mutex_unlock(&mi->data_lock); DBUG_RETURN(1); } int register_slave_on_master(MYSQL* mysql) { char buf[1024], *pos= buf; uint report_host_len, report_user_len=0, report_password_len=0; if (!report_host) return 0; report_host_len= strlen(report_host); if (report_user) report_user_len= strlen(report_user); if (report_password) report_password_len= strlen(report_password); /* 30 is a good safety margin */ if (report_host_len + report_user_len + report_password_len + 30 > sizeof(buf)) return 0; // safety int4store(pos, server_id); pos+= 4; pos= net_store_data(pos, report_host, report_host_len); pos= net_store_data(pos, report_user, report_user_len); pos= net_store_data(pos, report_password, report_password_len); int2store(pos, (uint16) report_port); pos+= 2; int4store(pos, rpl_recovery_rank); pos+= 4; /* The master will fill in master_id */ int4store(pos, 0); pos+= 4; if (simple_command(mysql, COM_REGISTER_SLAVE, (char*) buf, (uint) (pos- buf), 0)) { sql_print_error("Error on COM_REGISTER_SLAVE: %d '%s'", mysql_errno(mysql), mysql_error(mysql)); return 1; } return 0; } int show_master_info(THD* thd, MASTER_INFO* mi) { // TODO: fix this for multi-master List<Item> field_list; Protocol *protocol= thd->protocol; DBUG_ENTER("show_master_info"); field_list.push_back(new Item_empty_string("Master_Host", sizeof(mi->host))); field_list.push_back(new Item_empty_string("Master_User", sizeof(mi->user))); field_list.push_back(new Item_return_int("Master_Port", 7, MYSQL_TYPE_LONG)); field_list.push_back(new Item_return_int("Connect_retry", 10, MYSQL_TYPE_LONG)); field_list.push_back(new Item_empty_string("Master_Log_File", FN_REFLEN)); field_list.push_back(new Item_return_int("Read_Master_Log_Pos", 10, MYSQL_TYPE_LONGLONG)); field_list.push_back(new Item_empty_string("Relay_Log_File", FN_REFLEN)); field_list.push_back(new Item_return_int("Relay_Log_Pos", 10, MYSQL_TYPE_LONGLONG)); field_list.push_back(new Item_empty_string("Relay_Master_Log_File", FN_REFLEN)); field_list.push_back(new Item_empty_string("Slave_IO_Running", 3)); field_list.push_back(new Item_empty_string("Slave_SQL_Running", 3)); field_list.push_back(new Item_empty_string("Replicate_do_db", 20)); field_list.push_back(new Item_empty_string("Replicate_ignore_db", 20)); field_list.push_back(new Item_return_int("Last_errno", 4, MYSQL_TYPE_LONG)); field_list.push_back(new Item_empty_string("Last_error", 20)); field_list.push_back(new Item_return_int("Skip_counter", 10, MYSQL_TYPE_LONG)); field_list.push_back(new Item_return_int("Exec_master_log_pos", 10, MYSQL_TYPE_LONGLONG)); field_list.push_back(new Item_return_int("Relay_log_space", 10, MYSQL_TYPE_LONGLONG)); if (protocol->send_fields(&field_list, 1)) DBUG_RETURN(-1); if (mi->host[0]) { String *packet= &thd->packet; protocol->prepare_for_resend(); pthread_mutex_lock(&mi->data_lock); pthread_mutex_lock(&mi->rli.data_lock); protocol->store(mi->host, &my_charset_bin); protocol->store(mi->user, &my_charset_bin); protocol->store((uint32) mi->port); protocol->store((uint32) mi->connect_retry); protocol->store(mi->master_log_name, &my_charset_bin); protocol->store((ulonglong) mi->master_log_pos); protocol->store(mi->rli.group_relay_log_name + dirname_length(mi->rli.group_relay_log_name), &my_charset_bin); protocol->store((ulonglong) mi->rli.group_relay_log_pos); protocol->store(mi->rli.group_master_log_name, &my_charset_bin); protocol->store(mi->slave_running ? "Yes":"No", &my_charset_bin); protocol->store(mi->rli.slave_running ? "Yes":"No", &my_charset_bin); protocol->store(&replicate_do_db); protocol->store(&replicate_ignore_db); protocol->store((uint32) mi->rli.last_slave_errno); protocol->store(mi->rli.last_slave_error, &my_charset_bin); protocol->store((uint32) mi->rli.slave_skip_counter); protocol->store((ulonglong) mi->rli.group_master_log_pos); protocol->store((ulonglong) mi->rli.log_space_total); pthread_mutex_unlock(&mi->rli.data_lock); pthread_mutex_unlock(&mi->data_lock); if (my_net_write(&thd->net, (char*)thd->packet.ptr(), packet->length())) DBUG_RETURN(-1); } send_eof(thd); DBUG_RETURN(0); } bool flush_master_info(MASTER_INFO* mi) { IO_CACHE* file = &mi->file; char lbuf[22]; DBUG_ENTER("flush_master_info"); DBUG_PRINT("enter",("master_pos: %ld", (long) mi->master_log_pos)); my_b_seek(file, 0L); my_b_printf(file, "%s\n%s\n%s\n%s\n%s\n%d\n%d\n", mi->master_log_name, llstr(mi->master_log_pos, lbuf), mi->host, mi->user, mi->password, mi->port, mi->connect_retry); flush_io_cache(file); DBUG_RETURN(0); } st_relay_log_info::st_relay_log_info() :info_fd(-1), cur_log_fd(-1), save_temporary_tables(0), cur_log_old_open_count(0), group_master_log_pos(0), log_space_total(0), ignore_log_space_limit(0), slave_skip_counter(0), abort_pos_wait(0), slave_run_id(0), sql_thd(0), last_slave_errno(0), inited(0), abort_slave(0), slave_running(0) { group_relay_log_name[0]= event_relay_log_name[0]= group_master_log_name[0]= 0; last_slave_error[0]=0; bzero(&info_file,sizeof(info_file)); bzero(&cache_buf, sizeof(cache_buf)); pthread_mutex_init(&run_lock, MY_MUTEX_INIT_FAST); pthread_mutex_init(&data_lock, MY_MUTEX_INIT_FAST); pthread_mutex_init(&log_space_lock, MY_MUTEX_INIT_FAST); pthread_cond_init(&data_cond, NULL); pthread_cond_init(&start_cond, NULL); pthread_cond_init(&stop_cond, NULL); pthread_cond_init(&log_space_cond, NULL); } st_relay_log_info::~st_relay_log_info() { pthread_mutex_destroy(&run_lock); pthread_mutex_destroy(&data_lock); pthread_mutex_destroy(&log_space_lock); pthread_cond_destroy(&data_cond); pthread_cond_destroy(&start_cond); pthread_cond_destroy(&stop_cond); pthread_cond_destroy(&log_space_cond); } /* Waits until the SQL thread reaches (has executed up to) the log/position or timed out. SYNOPSIS wait_for_pos() thd client thread that sent SELECT MASTER_POS_WAIT log_name log name to wait for log_pos position to wait for timeout timeout in seconds before giving up waiting NOTES timeout is longlong whereas it should be ulong ; but this is to catch if the user submitted a negative timeout. RETURN VALUES -2 improper arguments (log_pos<0) or slave not running, or master info changed during the function's execution, or client thread killed. -2 is translated to NULL by caller -1 timed out >=0 number of log events the function had to wait before reaching the desired log/position */ int st_relay_log_info::wait_for_pos(THD* thd, String* log_name, longlong log_pos, longlong timeout) { if (!inited) return -1; int event_count = 0; ulong init_abort_pos_wait; int error=0; struct timespec abstime; // for timeout checking set_timespec(abstime,timeout); DBUG_ENTER("wait_for_pos"); DBUG_PRINT("enter",("group_master_log_name: '%s' pos: %lu timeout: %ld", group_master_log_name, (ulong) group_master_log_pos, (long) timeout)); pthread_mutex_lock(&data_lock); /* This function will abort when it notices that some CHANGE MASTER or RESET MASTER has changed the master info. To catch this, these commands modify abort_pos_wait ; we just monitor abort_pos_wait and see if it has changed. */ init_abort_pos_wait= abort_pos_wait; /* We'll need to handle all possible log names comparisons (e.g. 999 vs 1000). We use ulong for string->number conversion ; this is no stronger limitation than in find_uniq_filename in sql/log.cc */ ulong log_name_extension; char log_name_tmp[FN_REFLEN]; //make a char[] from String char *end= strmake(log_name_tmp, log_name->ptr(), min(log_name->length(), FN_REFLEN-1)); char *p= fn_ext(log_name_tmp); char *p_end; if (!*p || log_pos<0) { error= -2; //means improper arguments goto err; } //p points to '.' log_name_extension= strtoul(++p, &p_end, 10); /* p_end points to the first invalid character. If it equals to p, no digits were found, error. If it contains '\0' it means conversion went ok. */ if (p_end==p || *p_end) { error= -2; goto err; } //"compare and wait" main loop while (!thd->killed && init_abort_pos_wait == abort_pos_wait && mi->slave_running) { bool pos_reached; int cmp_result= 0; DBUG_ASSERT(*group_master_log_name || group_master_log_pos == 0); if (*group_master_log_name) { char *basename= group_master_log_name + dirname_length(group_master_log_name); /* First compare the parts before the extension. Find the dot in the master's log basename, and protect against user's input error : if the names do not match up to '.' included, return error */ char *q= (char*)(fn_ext(basename)+1); if (strncmp(basename, log_name_tmp, (int)(q-basename))) { error= -2; break; } // Now compare extensions. char *q_end; ulong group_master_log_name_extension= strtoul(q, &q_end, 10); if (group_master_log_name_extension < log_name_extension) cmp_result = -1 ; else cmp_result= (group_master_log_name_extension > log_name_extension) ? 1 : 0 ; } pos_reached = ((!cmp_result && group_master_log_pos >= (ulonglong)log_pos) || cmp_result > 0); if (pos_reached || thd->killed) break; //wait for master update, with optional timeout. DBUG_PRINT("info",("Waiting for master update")); const char* msg = thd->enter_cond(&data_cond, &data_lock, "Waiting for master update"); if (timeout > 0) { /* Note that pthread_cond_timedwait checks for the timeout before for the condition ; i.e. it returns ETIMEDOUT if the system time equals or exceeds the time specified by abstime before the condition variable is signaled or broadcast, _or_ if the absolute time specified by abstime has already passed at the time of the call. For that reason, pthread_cond_timedwait will do the "timeoutting" job even if its condition is always immediately signaled (case of a loaded master). */ error=pthread_cond_timedwait(&data_cond, &data_lock, &abstime); } else pthread_cond_wait(&data_cond, &data_lock); thd->exit_cond(msg); if (error == ETIMEDOUT || error == ETIME) { error= -1; break; } error=0; event_count++; } err: pthread_mutex_unlock(&data_lock); DBUG_PRINT("exit",("killed: %d abort: %d slave_running: %d \ improper_arguments: %d timed_out: %d", (int) thd->killed, (int) (init_abort_pos_wait != abort_pos_wait), (int) mi->slave_running, (int) (error == -2), (int) (error == -1))); if (thd->killed || init_abort_pos_wait != abort_pos_wait || !mi->slave_running) { error= -2; } DBUG_RETURN( error ? error : event_count ); } /* init_slave_thread() */ static int init_slave_thread(THD* thd, SLAVE_THD_TYPE thd_type) { DBUG_ENTER("init_slave_thread"); thd->system_thread = thd->bootstrap = 1; thd->host_or_ip= ""; thd->client_capabilities = 0; my_net_init(&thd->net, 0); thd->net.read_timeout = slave_net_timeout; thd->master_access= ~0; thd->priv_user = 0; thd->slave_thread = 1; thd->options = (((opt_log_slave_updates) ? OPTION_BIN_LOG:0) | OPTION_AUTO_IS_NULL) ; thd->client_capabilities = CLIENT_LOCAL_FILES; thd->real_id=pthread_self(); pthread_mutex_lock(&LOCK_thread_count); thd->thread_id = thread_id++; pthread_mutex_unlock(&LOCK_thread_count); if (init_thr_lock() || thd->store_globals()) { thd->cleanup(); delete thd; DBUG_RETURN(-1); } #if !defined(__WIN__) && !defined(OS2) && !defined(__NETWARE__) sigset_t set; VOID(sigemptyset(&set)); // Get mask in use VOID(pthread_sigmask(SIG_UNBLOCK,&set,&thd->block_signals)); #endif if (thd->variables.max_join_size == HA_POS_ERROR) thd->options |= OPTION_BIG_SELECTS; if (thd_type == SLAVE_THD_SQL) thd->proc_info= "Waiting for the next event in slave queue"; else thd->proc_info= "Waiting for master update"; thd->version=refresh_version; thd->set_time(); DBUG_RETURN(0); } static int safe_sleep(THD* thd, int sec, CHECK_KILLED_FUNC thread_killed, void* thread_killed_arg) { int nap_time; thr_alarm_t alarmed; thr_alarm_init(&alarmed); time_t start_time= time((time_t*) 0); time_t end_time= start_time+sec; while ((nap_time= (int) (end_time - start_time)) > 0) { ALARM alarm_buff; /* The only reason we are asking for alarm is so that we will be woken up in case of murder, so if we do not get killed, set the alarm so it goes off after we wake up naturally */ thr_alarm(&alarmed, 2 * nap_time, &alarm_buff); sleep(nap_time); thr_end_alarm(&alarmed); if ((*thread_killed)(thd,thread_killed_arg)) return 1; start_time=time((time_t*) 0); } return 0; } static int request_dump(MYSQL* mysql, MASTER_INFO* mi, bool *suppress_warnings) { char buf[FN_REFLEN + 10]; int len; int binlog_flags = 0; // for now char* logname = mi->master_log_name; DBUG_ENTER("request_dump"); // TODO if big log files: Change next to int8store() int4store(buf, (longlong) mi->master_log_pos); int2store(buf + 4, binlog_flags); int4store(buf + 6, server_id); len = (uint) strlen(logname); memcpy(buf + 10, logname,len); if (simple_command(mysql, COM_BINLOG_DUMP, buf, len + 10, 1)) { /* Something went wrong, so we will just reconnect and retry later in the future, we should do a better error analysis, but for now we just fill up the error log :-) */ if (mysql_errno(mysql) == ER_NET_READ_INTERRUPTED) *suppress_warnings= 1; // Suppress reconnect warning else sql_print_error("Error on COM_BINLOG_DUMP: %d %s, will retry in %d secs", mysql_errno(mysql), mysql_error(mysql), master_connect_retry); DBUG_RETURN(1); } DBUG_RETURN(0); } static int request_table_dump(MYSQL* mysql, const char* db, const char* table) { char buf[1024]; char * p = buf; uint table_len = (uint) strlen(table); uint db_len = (uint) strlen(db); if (table_len + db_len > sizeof(buf) - 2) { sql_print_error("request_table_dump: Buffer overrun"); return 1; } *p++ = db_len; memcpy(p, db, db_len); p += db_len; *p++ = table_len; memcpy(p, table, table_len); if (simple_command(mysql, COM_TABLE_DUMP, buf, p - buf + table_len, 1)) { sql_print_error("request_table_dump: Error sending the table dump \ command"); return 1; } return 0; } /* Read one event from the master SYNOPSIS read_event() mysql MySQL connection mi Master connection information suppress_warnings TRUE when a normal net read timeout has caused us to try a reconnect. We do not want to print anything to the error log in this case because this a anormal event in an idle server. RETURN VALUES 'packet_error' Error number Length of packet */ static ulong read_event(MYSQL* mysql, MASTER_INFO *mi, bool* suppress_warnings) { ulong len; *suppress_warnings= 0; /* my_real_read() will time us out We check if we were told to die, and if not, try reading again TODO: Move 'events_till_disconnect' to the MASTER_INFO structure */ #ifndef DBUG_OFF if (disconnect_slave_event_count && !(events_till_disconnect--)) return packet_error; #endif len = net_safe_read(mysql); if (len == packet_error || (long) len < 1) { if (mysql_errno(mysql) == ER_NET_READ_INTERRUPTED) { /* We are trying a normal reconnect after a read timeout; we suppress prints to .err file as long as the reconnect happens without problems */ *suppress_warnings= TRUE; } else sql_print_error("Error reading packet from server: %s (\ server_errno=%d)", mysql_error(mysql), mysql_errno(mysql)); return packet_error; } if (len == 1) { sql_print_error("Slave: received 0 length packet from server, apparent\ master shutdown: %s", mysql_error(mysql)); return packet_error; } DBUG_PRINT("info",( "len=%u, net->read_pos[4] = %d\n", len, mysql->net.read_pos[4])); return len - 1; } int check_expected_error(THD* thd, RELAY_LOG_INFO* rli, int expected_error) { switch (expected_error) { case ER_NET_READ_ERROR: case ER_NET_ERROR_ON_WRITE: case ER_SERVER_SHUTDOWN: case ER_NEW_ABORTING_CONNECTION: my_snprintf(rli->last_slave_error, sizeof(rli->last_slave_error), "Slave: query '%s' partially completed on the master \ and was aborted. There is a chance that your master is inconsistent at this \ point. If you are sure that your master is ok, run this query manually on the\ slave and then restart the slave with SET GLOBAL SQL_SLAVE_SKIP_COUNTER=1;\ SLAVE START;", thd->query); rli->last_slave_errno = expected_error; sql_print_error("%s",rli->last_slave_error); return 1; default: return 0; } } static int exec_relay_log_event(THD* thd, RELAY_LOG_INFO* rli) { Log_event * ev = next_event(rli); DBUG_ASSERT(rli->sql_thd==thd); if (sql_slave_killed(thd,rli)) { /* do not forget to free ev ! */ if (ev) delete ev; return 1; } if (ev) { int type_code = ev->get_type_code(); int exec_res; pthread_mutex_lock(&rli->data_lock); /* Skip queries originating from this server or number of queries specified by the user in slave_skip_counter We can't however skip event's that has something to do with the log files themselves. */ if (ev->server_id == (uint32) ::server_id || (rli->slave_skip_counter && type_code != ROTATE_EVENT)) { /* TODO: I/O thread should not even log events with the same server id */ rli->inc_group_relay_log_pos(ev->get_event_len(), type_code != STOP_EVENT ? ev->log_pos : LL(0), 1/* skip lock*/); flush_relay_log_info(rli); /* Protect against common user error of setting the counter to 1 instead of 2 while recovering from an failed auto-increment insert */ if (rli->slave_skip_counter && !((type_code == INTVAR_EVENT || type_code == STOP_EVENT) && rli->slave_skip_counter == 1)) --rli->slave_skip_counter; pthread_mutex_unlock(&rli->data_lock); delete ev; return 0; // avoid infinite update loops } pthread_mutex_unlock(&rli->data_lock); thd->server_id = ev->server_id; // use the original server id for logging thd->set_time(); // time the query if (!ev->when) ev->when = time(NULL); ev->thd = thd; thd->log_pos = ev->log_pos; exec_res = ev->exec_event(rli); DBUG_ASSERT(rli->sql_thd==thd); delete ev; return exec_res; } else { sql_print_error("\ Could not parse relay log event entry. The possible reasons are: the master's \ binary log is corrupted (you can check this by running 'mysqlbinlog' on the \ binary log), the slave's relay log is corrupted (you can check this by running \ 'mysqlbinlog' on the relay log), a network problem, or a bug in the master's \ or slave's MySQL code. If you want to check the master's binary log or slave's \ relay log, you will be able to know their names by issuing 'SHOW SLAVE STATUS' \ on this slave.\ "); return 1; } } /* Slave I/O Thread entry point */ extern "C" pthread_handler_decl(handle_slave_io,arg) { THD *thd; // needs to be first for thread_stack MYSQL *mysql; MASTER_INFO *mi = (MASTER_INFO*)arg; char llbuff[22]; uint retry_count; // needs to call my_thread_init(), otherwise we get a coredump in DBUG_ stuff my_thread_init(); DBUG_ENTER("handle_slave_io"); #ifndef DBUG_OFF slave_begin: #endif DBUG_ASSERT(mi->inited); mysql= NULL ; retry_count= 0; pthread_mutex_lock(&mi->run_lock); /* Inform waiting threads that slave has started */ mi->slave_run_id++; #ifndef DBUG_OFF mi->events_till_abort = abort_slave_event_count; #endif thd= new THD; // note that contructor of THD uses DBUG_ ! THD_CHECK_SENTRY(thd); pthread_detach_this_thread(); if (init_slave_thread(thd, SLAVE_THD_IO)) { pthread_cond_broadcast(&mi->start_cond); pthread_mutex_unlock(&mi->run_lock); sql_print_error("Failed during slave I/O thread initialization"); goto err; } mi->io_thd = thd; thd->thread_stack = (char*)&thd; // remember where our stack is pthread_mutex_lock(&LOCK_thread_count); threads.append(thd); pthread_mutex_unlock(&LOCK_thread_count); mi->slave_running = 1; mi->abort_slave = 0; pthread_mutex_unlock(&mi->run_lock); pthread_cond_broadcast(&mi->start_cond); DBUG_PRINT("master_info",("log_file_name: '%s' position: %s", mi->master_log_name, llstr(mi->master_log_pos,llbuff))); if (!(mi->mysql = mysql = mysql_init(NULL))) { sql_print_error("Slave I/O thread: error in mysql_init()"); goto err; } thd->proc_info = "connecting to master"; // we can get killed during safe_connect if (!safe_connect(thd, mysql, mi)) sql_print_error("Slave I/O thread: connected to master '%s@%s:%d',\ replication started in log '%s' at position %s", mi->user, mi->host, mi->port, IO_RPL_LOG_NAME, llstr(mi->master_log_pos,llbuff)); else { sql_print_error("Slave I/O thread killed while connecting to master"); goto err; } connected: thd->slave_net = &mysql->net; thd->proc_info = "Checking master version"; if (check_master_version(mysql, mi)) goto err; if (!mi->old_format) { /* Register ourselves with the master. If fails, this is not fatal - we just print the error message and go on with life. */ thd->proc_info = "Registering slave on master"; if (register_slave_on_master(mysql) || update_slave_list(mysql, mi)) goto err; } DBUG_PRINT("info",("Starting reading binary log from master")); while (!io_slave_killed(thd,mi)) { bool suppress_warnings= 0; thd->proc_info = "Requesting binlog dump"; if (request_dump(mysql, mi, &suppress_warnings)) { sql_print_error("Failed on request_dump()"); if (io_slave_killed(thd,mi)) { sql_print_error("Slave I/O thread killed while requesting master \ dump"); goto err; } thd->proc_info = "Waiting to reconnect after a failed dump request"; end_server(mysql); /* First time retry immediately, assuming that we can recover right away - if first time fails, sleep between re-tries hopefuly the admin can fix the problem sometime */ if (retry_count++) { if (retry_count > master_retry_count) goto err; // Don't retry forever safe_sleep(thd,mi->connect_retry,(CHECK_KILLED_FUNC)io_slave_killed, (void*)mi); } if (io_slave_killed(thd,mi)) { sql_print_error("Slave I/O thread killed while retrying master \ dump"); goto err; } thd->proc_info = "Reconnecting after a failed dump request"; if (!suppress_warnings) sql_print_error("Slave I/O thread: failed dump request, \ reconnecting to try again, log '%s' at postion %s", IO_RPL_LOG_NAME, llstr(mi->master_log_pos,llbuff)); if (safe_reconnect(thd, mysql, mi, suppress_warnings) || io_slave_killed(thd,mi)) { sql_print_error("Slave I/O thread killed during or \ after reconnect"); goto err; } goto connected; } while (!io_slave_killed(thd,mi)) { bool suppress_warnings= 0; thd->proc_info = "Reading master update"; ulong event_len = read_event(mysql, mi, &suppress_warnings); if (io_slave_killed(thd,mi)) { if (global_system_variables.log_warnings) sql_print_error("Slave I/O thread killed while reading event"); goto err; } if (event_len == packet_error) { uint mysql_error_number= mysql_errno(mysql); if (mysql_error_number == ER_NET_PACKET_TOO_LARGE) { sql_print_error("\ Log entry on master is longer than max_allowed_packet (%ld) on \ slave. If the entry is correct, restart the server with a higher value of \ max_allowed_packet", thd->variables.max_allowed_packet); goto err; } if (mysql_error_number == ER_MASTER_FATAL_ERROR_READING_BINLOG) { sql_print_error(ER(mysql_error_number), mysql_error_number, mysql_error(mysql)); goto err; } thd->proc_info = "Waiting to reconnect after a failed read"; end_server(mysql); if (retry_count++) { if (retry_count > master_retry_count) goto err; // Don't retry forever safe_sleep(thd,mi->connect_retry,(CHECK_KILLED_FUNC)io_slave_killed, (void*) mi); } if (io_slave_killed(thd,mi)) { if (global_system_variables.log_warnings) sql_print_error("Slave I/O thread killed while waiting to \ reconnect after a failed read"); goto err; } thd->proc_info = "Reconnecting after a failed read"; if (!suppress_warnings) sql_print_error("Slave I/O thread: Failed reading log event, \ reconnecting to retry, log '%s' position %s", IO_RPL_LOG_NAME, llstr(mi->master_log_pos, llbuff)); if (safe_reconnect(thd, mysql, mi, suppress_warnings) || io_slave_killed(thd,mi)) { if (global_system_variables.log_warnings) sql_print_error("Slave I/O thread killed during or after a \ reconnect done to recover from failed read"); goto err; } goto connected; } // if (event_len == packet_error) retry_count=0; // ok event, reset retry counter thd->proc_info = "Queueing event from master"; if (queue_event(mi,(const char*)mysql->net.read_pos + 1, event_len)) { sql_print_error("Slave I/O thread could not queue event from master"); goto err; } flush_master_info(mi); /* See if the relay logs take too much space. We don't lock mi->rli.log_space_lock here; this dirty read saves time and does not introduce any problem: - if mi->rli.ignore_log_space_limit is 1 but becomes 0 just after (so the clean value is 0), then we are reading only one more event as we should, and we'll block only at the next event. No big deal. - if mi->rli.ignore_log_space_limit is 0 but becomes 1 just after (so the clean value is 1), then we are going into wait_for_relay_log_space() for no reason, but this function will do a clean read, notice the clean value and exit immediately. */ if (mi->rli.log_space_limit && mi->rli.log_space_limit < mi->rli.log_space_total && !mi->rli.ignore_log_space_limit) if (wait_for_relay_log_space(&mi->rli)) { sql_print_error("Slave I/O thread aborted while waiting for relay \ log space"); goto err; } // TODO: check debugging abort code #ifndef DBUG_OFF if (abort_slave_event_count && !--events_till_abort) { sql_print_error("Slave I/O thread: debugging abort"); goto err; } #endif } } // error = 0; err: // print the current replication position sql_print_error("Slave I/O thread exiting, read up to log '%s', position %s", IO_RPL_LOG_NAME, llstr(mi->master_log_pos,llbuff)); VOID(pthread_mutex_lock(&LOCK_thread_count)); thd->query = thd->db = 0; // extra safety VOID(pthread_mutex_unlock(&LOCK_thread_count)); if (mysql) { mysql_close(mysql); mi->mysql=0; } thd->proc_info = "Waiting for slave mutex on exit"; pthread_mutex_lock(&mi->run_lock); mi->slave_running = 0; mi->io_thd = 0; // TODO: make rpl_status part of MASTER_INFO change_rpl_status(RPL_ACTIVE_SLAVE,RPL_IDLE_SLAVE); mi->abort_slave = 0; // TODO: check if this is needed DBUG_ASSERT(thd->net.buff != 0); net_end(&thd->net); // destructor will not free it, because net.vio is 0 pthread_mutex_lock(&LOCK_thread_count); THD_CHECK_SENTRY(thd); delete thd; pthread_mutex_unlock(&LOCK_thread_count); pthread_cond_broadcast(&mi->stop_cond); // tell the world we are done pthread_mutex_unlock(&mi->run_lock); #ifndef DBUG_OFF if (abort_slave_event_count && !events_till_abort) goto slave_begin; #endif my_thread_end(); pthread_exit(0); DBUG_RETURN(0); // Can't return anything here } /* Slave SQL Thread entry point */ extern "C" pthread_handler_decl(handle_slave_sql,arg) { THD *thd; /* needs to be first for thread_stack */ char llbuff[22],llbuff1[22]; RELAY_LOG_INFO* rli = &((MASTER_INFO*)arg)->rli; const char *errmsg; // needs to call my_thread_init(), otherwise we get a coredump in DBUG_ stuff my_thread_init(); DBUG_ENTER("handle_slave_sql"); #ifndef DBUG_OFF slave_begin: #endif DBUG_ASSERT(rli->inited); pthread_mutex_lock(&rli->run_lock); DBUG_ASSERT(!rli->slave_running); errmsg= 0; #ifndef DBUG_OFF rli->events_till_abort = abort_slave_event_count; #endif thd = new THD; // note that contructor of THD uses DBUG_ ! thd->thread_stack = (char*)&thd; // remember where our stack is /* Inform waiting threads that slave has started */ rli->slave_run_id++; pthread_detach_this_thread(); if (init_slave_thread(thd, SLAVE_THD_SQL)) { /* TODO: this is currently broken - slave start and change master will be stuck if we fail here */ pthread_cond_broadcast(&rli->start_cond); pthread_mutex_unlock(&rli->run_lock); sql_print_error("Failed during slave thread initialization"); goto err; } thd->init_for_queries(); rli->sql_thd= thd; thd->temporary_tables = rli->save_temporary_tables; // restore temp tables pthread_mutex_lock(&LOCK_thread_count); threads.append(thd); pthread_mutex_unlock(&LOCK_thread_count); rli->slave_running = 1; rli->abort_slave = 0; pthread_mutex_unlock(&rli->run_lock); pthread_cond_broadcast(&rli->start_cond); //tell the I/O thread to take relay_log_space_limit into account from now on pthread_mutex_lock(&rli->log_space_lock); rli->ignore_log_space_limit= 0; pthread_mutex_unlock(&rli->log_space_lock); if (init_relay_log_pos(rli, rli->group_relay_log_name, rli->group_relay_log_pos, 1 /*need data lock*/, &errmsg)) { sql_print_error("Error initializing relay log position: %s", errmsg); goto err; } THD_CHECK_SENTRY(thd); DBUG_ASSERT(rli->event_relay_log_pos >= BIN_LOG_HEADER_SIZE); DBUG_ASSERT(my_b_tell(rli->cur_log) == rli->event_relay_log_pos); DBUG_ASSERT(rli->sql_thd == thd); DBUG_PRINT("master_info",("log_file_name: %s position: %s", rli->group_master_log_name, llstr(rli->group_master_log_pos,llbuff))); if (global_system_variables.log_warnings) sql_print_error("Slave SQL thread initialized, starting replication in \ log '%s' at position %s, relay log '%s' position: %s", RPL_LOG_NAME, llstr(rli->group_master_log_pos,llbuff),rli->group_relay_log_name, llstr(rli->group_relay_log_pos,llbuff1)); /* Read queries from the IO/THREAD until this thread is killed */ while (!sql_slave_killed(thd,rli)) { thd->proc_info = "Processing master log event"; DBUG_ASSERT(rli->sql_thd == thd); THD_CHECK_SENTRY(thd); if (exec_relay_log_event(thd,rli)) { // do not scare the user if SQL thread was simply killed or stopped if (!sql_slave_killed(thd,rli)) sql_print_error("\ Error running query, slave SQL thread aborted. Fix the problem, and restart \ the slave SQL thread with \"SLAVE START\". We stopped at log \ '%s' position %s", RPL_LOG_NAME, llstr(rli->group_master_log_pos, llbuff)); goto err; } } /* Thread stopped. Print the current replication position to the log */ sql_print_error("Slave SQL thread exiting, replication stopped in log \ '%s' at position %s", RPL_LOG_NAME, llstr(rli->group_master_log_pos,llbuff)); err: VOID(pthread_mutex_lock(&LOCK_thread_count)); thd->query = thd->db = 0; // extra safety VOID(pthread_mutex_unlock(&LOCK_thread_count)); thd->proc_info = "Waiting for slave mutex on exit"; pthread_mutex_lock(&rli->run_lock); DBUG_ASSERT(rli->slave_running == 1); // tracking buffer overrun rli->slave_running = 0; rli->save_temporary_tables = thd->temporary_tables; /* TODO: see if we can do this conditionally in next_event() instead to avoid unneeded position re-init */ thd->temporary_tables = 0; // remove tempation from destructor to close them DBUG_ASSERT(thd->net.buff != 0); net_end(&thd->net); // destructor will not free it, because we are weird DBUG_ASSERT(rli->sql_thd == thd); THD_CHECK_SENTRY(thd); rli->sql_thd= 0; pthread_mutex_lock(&LOCK_thread_count); THD_CHECK_SENTRY(thd); delete thd; pthread_mutex_unlock(&LOCK_thread_count); pthread_cond_broadcast(&rli->stop_cond); // tell the world we are done pthread_mutex_unlock(&rli->run_lock); #ifndef DBUG_OFF // TODO: reconsider the code below if (abort_slave_event_count && !rli->events_till_abort) goto slave_begin; #endif my_thread_end(); pthread_exit(0); DBUG_RETURN(0); // Can't return anything here } /* process_io_create_file() */ static int process_io_create_file(MASTER_INFO* mi, Create_file_log_event* cev) { int error = 1; ulong num_bytes; bool cev_not_written; THD *thd = mi->io_thd; NET *net = &mi->mysql->net; DBUG_ENTER("process_io_create_file"); if (unlikely(!cev->is_valid())) DBUG_RETURN(1); /* TODO: fix to honor table rules, not only db rules */ if (!db_ok(cev->db, replicate_do_db, replicate_ignore_db)) { skip_load_data_infile(net); DBUG_RETURN(0); } DBUG_ASSERT(cev->inited_from_old); thd->file_id = cev->file_id = mi->file_id++; thd->server_id = cev->server_id; cev_not_written = 1; if (unlikely(net_request_file(net,cev->fname))) { sql_print_error("Slave I/O: failed requesting download of '%s'", cev->fname); goto err; } /* This dummy block is so we could instantiate Append_block_log_event once and then modify it slightly instead of doing it multiple times in the loop */ { Append_block_log_event aev(thd,0,0,0); for (;;) { if (unlikely((num_bytes=my_net_read(net)) == packet_error)) { sql_print_error("Network read error downloading '%s' from master", cev->fname); goto err; } if (unlikely(!num_bytes)) /* eof */ { net_write_command(net, 0, "", 0, "", 0);/* 3.23 master wants it */ Execute_load_log_event xev(thd,0); xev.log_pos = mi->master_log_pos; if (unlikely(mi->rli.relay_log.append(&xev))) { sql_print_error("Slave I/O: error writing Exec_load event to \ relay log"); goto err; } mi->rli.relay_log.harvest_bytes_written(&mi->rli.log_space_total); break; } if (unlikely(cev_not_written)) { cev->block = (char*)net->read_pos; cev->block_len = num_bytes; cev->log_pos = mi->master_log_pos; if (unlikely(mi->rli.relay_log.append(cev))) { sql_print_error("Slave I/O: error writing Create_file event to \ relay log"); goto err; } cev_not_written=0; mi->rli.relay_log.harvest_bytes_written(&mi->rli.log_space_total); } else { aev.block = (char*)net->read_pos; aev.block_len = num_bytes; aev.log_pos = mi->master_log_pos; if (unlikely(mi->rli.relay_log.append(&aev))) { sql_print_error("Slave I/O: error writing Append_block event to \ relay log"); goto err; } mi->rli.relay_log.harvest_bytes_written(&mi->rli.log_space_total) ; } } } error=0; err: DBUG_RETURN(error); } /* Start using a new binary log on the master SYNOPSIS process_io_rotate() mi master_info for the slave rev The rotate log event read from the binary log DESCRIPTION Updates the master info with the place in the next binary log where we should start reading. NOTES We assume we already locked mi->data_lock RETURN VALUES 0 ok 1 Log event is illegal */ static int process_io_rotate(MASTER_INFO *mi, Rotate_log_event *rev) { int return_val= 1; DBUG_ENTER("process_io_rotate"); safe_mutex_assert_owner(&mi->data_lock); if (unlikely(!rev->is_valid())) DBUG_RETURN(1); memcpy(mi->master_log_name, rev->new_log_ident, rev->ident_len+1); mi->master_log_pos= rev->pos; DBUG_PRINT("info", ("master_log_pos: '%s' %d", mi->master_log_name, (ulong) mi->master_log_pos)); #ifndef DBUG_OFF /* If we do not do this, we will be getting the first rotate event forever, so we need to not disconnect after one. */ if (disconnect_slave_event_count) events_till_disconnect++; #endif DBUG_RETURN(0); } /* queue_old_event() Writes a 3.23 event to the relay log. TODO: Test this code before release - it has to be tested on a separate setup with 3.23 master */ static int queue_old_event(MASTER_INFO *mi, const char *buf, ulong event_len) { const char *errmsg = 0; ulong inc_pos; bool ignore_event= 0; char *tmp_buf = 0; RELAY_LOG_INFO *rli= &mi->rli; DBUG_ENTER("queue_old_event"); /* If we get Load event, we need to pass a non-reusable buffer to read_log_event, so we do a trick */ if (buf[EVENT_TYPE_OFFSET] == LOAD_EVENT) { if (unlikely(!(tmp_buf=(char*)my_malloc(event_len+1,MYF(MY_WME))))) { sql_print_error("Slave I/O: out of memory for Load event"); DBUG_RETURN(1); } memcpy(tmp_buf,buf,event_len); tmp_buf[event_len]=0; // Create_file constructor wants null-term buffer buf = (const char*)tmp_buf; } Log_event *ev = Log_event::read_log_event(buf,event_len, &errmsg, 1 /*old format*/ ); if (unlikely(!ev)) { sql_print_error("Read invalid event from master: '%s',\ master could be corrupt but a more likely cause of this is a bug", errmsg); my_free((char*) tmp_buf, MYF(MY_ALLOW_ZERO_PTR)); DBUG_RETURN(1); } pthread_mutex_lock(&mi->data_lock); ev->log_pos = mi->master_log_pos; switch (ev->get_type_code()) { case STOP_EVENT: ignore_event= 1; inc_pos= event_len; break; case ROTATE_EVENT: if (unlikely(process_io_rotate(mi,(Rotate_log_event*)ev))) { delete ev; pthread_mutex_unlock(&mi->data_lock); DBUG_RETURN(1); } inc_pos= 0; break; case CREATE_FILE_EVENT: { /* We come here when and only when tmp_buf != 0 */ DBUG_ASSERT(tmp_buf); int error = process_io_create_file(mi,(Create_file_log_event*)ev); delete ev; mi->master_log_pos += event_len; DBUG_PRINT("info", ("master_log_pos: %d", (ulong) mi->master_log_pos)); pthread_mutex_unlock(&mi->data_lock); my_free((char*)tmp_buf, MYF(0)); DBUG_RETURN(error); } default: inc_pos= event_len; break; } if (likely(!ignore_event)) { if (unlikely(rli->relay_log.append(ev))) { delete ev; pthread_mutex_unlock(&mi->data_lock); DBUG_RETURN(1); } rli->relay_log.harvest_bytes_written(&rli->log_space_total); } delete ev; mi->master_log_pos+= inc_pos; DBUG_PRINT("info", ("master_log_pos: %d", (ulong) mi->master_log_pos)); pthread_mutex_unlock(&mi->data_lock); DBUG_RETURN(0); } /* queue_event() */ int queue_event(MASTER_INFO* mi,const char* buf, ulong event_len) { int error= 0; ulong inc_pos; RELAY_LOG_INFO *rli= &mi->rli; DBUG_ENTER("queue_event"); if (mi->old_format) DBUG_RETURN(queue_old_event(mi,buf,event_len)); pthread_mutex_lock(&mi->data_lock); /* TODO: figure out if other events in addition to Rotate require special processing. Guilhem 2003-06 : I don't think so. */ switch (buf[EVENT_TYPE_OFFSET]) { case STOP_EVENT: /* We needn't write this event to the relay log. Indeed, it just indicates a master server shutdown. The only thing this does is cleaning. But cleaning is already done on a per-master-thread basis (as the master server is shutting down cleanly, it has written all DROP TEMPORARY TABLE and DO RELEASE_LOCK; prepared statements' deletion are TODO). We don't even increment mi->master_log_pos, because we may be just after a Rotate event. Btw, in a few milliseconds we are going to have a Start event from the next binlog (unless the master is presently running without --log-bin). */ goto err; case ROTATE_EVENT: { Rotate_log_event rev(buf,event_len,0); if (unlikely(process_io_rotate(mi,&rev))) { error= 1; goto err; } /* Now the I/O thread has just changed its mi->master_log_name, so incrementing mi->master_log_pos is nonsense. */ inc_pos= 0; break; } default: inc_pos= event_len; break; } /* If this event is originating from this server, don't queue it. We don't check this for 3.23 events because it's simpler like this; 3.23 will be filtered anyway by the SQL slave thread which also tests the server id (we must also keep this test in the SQL thread, in case somebody upgrades a 4.0 slave which has a not-filtered relay log). ANY event coming from ourselves can be ignored: it is obvious for queries; for STOP_EVENT/ROTATE_EVENT/START_EVENT: these cannot come from ourselves (--log-slave-updates would not log that) unless this slave is also its direct master (an unsupported, useless setup!). */ if (uint4korr(buf + SERVER_ID_OFFSET) == ::server_id) { /* Do not write it to the relay log. We still want to increment, so that we won't re-read this event from the master if the slave IO thread is now stopped/restarted (more efficient if the events we are ignoring are big LOAD DATA INFILE). */ mi->master_log_pos+= inc_pos; DBUG_PRINT("info", ("master_log_pos: %d, event originating from the same server, ignored", (ulong) mi->master_log_pos)); } else /* write the event to the relay log */ if (likely(!(error= rli->relay_log.appendv(buf,event_len,0)))) { mi->master_log_pos+= inc_pos; DBUG_PRINT("info", ("master_log_pos: %d", (ulong) mi->master_log_pos)); rli->relay_log.harvest_bytes_written(&rli->log_space_total); } err: pthread_mutex_unlock(&mi->data_lock); DBUG_RETURN(error); } void end_relay_log_info(RELAY_LOG_INFO* rli) { DBUG_ENTER("end_relay_log_info"); if (!rli->inited) DBUG_VOID_RETURN; if (rli->info_fd >= 0) { end_io_cache(&rli->info_file); (void) my_close(rli->info_fd, MYF(MY_WME)); rli->info_fd = -1; } if (rli->cur_log_fd >= 0) { end_io_cache(&rli->cache_buf); (void)my_close(rli->cur_log_fd, MYF(MY_WME)); rli->cur_log_fd = -1; } rli->inited = 0; rli->relay_log.close(1); DBUG_VOID_RETURN; } /* Try to connect until successful or slave killed SYNPOSIS safe_connect() thd Thread handler for slave mysql MySQL connection handle mi Replication handle RETURN 0 ok # Error */ static int safe_connect(THD* thd, MYSQL* mysql, MASTER_INFO* mi) { return connect_to_master(thd, mysql, mi, 0, 0); } /* SYNPOSIS connect_to_master() IMPLEMENTATION Try to connect until successful or slave killed or we have retried master_retry_count times */ static int connect_to_master(THD* thd, MYSQL* mysql, MASTER_INFO* mi, bool reconnect, bool suppress_warnings) { int slave_was_killed; int last_errno= -2; // impossible error ulong err_count=0; char llbuff[22]; DBUG_ENTER("connect_to_master"); #ifndef DBUG_OFF events_till_disconnect = disconnect_slave_event_count; #endif ulong client_flag= CLIENT_REMEMBER_OPTIONS; if (opt_slave_compressed_protocol) client_flag=CLIENT_COMPRESS; /* We will use compression */ mysql_options(mysql, MYSQL_OPT_CONNECT_TIMEOUT, (char *) &slave_net_timeout); mysql_options(mysql, MYSQL_OPT_READ_TIMEOUT, (char *) &slave_net_timeout); mysql_options(mysql, MYSQL_SET_CHARSET_NAME, default_charset_info->csname); /* This one is not strictly needed but we have it here for completeness */ mysql_options(mysql, MYSQL_SET_CHARSET_DIR, (char *) charsets_dir); while (!(slave_was_killed = io_slave_killed(thd,mi)) && (reconnect ? mysql_reconnect(mysql) != 0 : mysql_real_connect(mysql, mi->host, mi->user, mi->password, 0, mi->port, 0, client_flag) == 0)) { /* Don't repeat last error */ if ((int)mysql_errno(mysql) != last_errno) { last_errno=mysql_errno(mysql); suppress_warnings= 0; sql_print_error("Slave I/O thread: error %s to master \ '%s@%s:%d': \ Error: '%s' errno: %d retry-time: %d retries: %d", (reconnect ? "reconnecting" : "connecting"), mi->user,mi->host,mi->port, mysql_error(mysql), last_errno, mi->connect_retry, master_retry_count); } /* By default we try forever. The reason is that failure will trigger master election, so if the user did not set master_retry_count we do not want to have election triggered on the first failure to connect */ if (++err_count == master_retry_count) { slave_was_killed=1; if (reconnect) change_rpl_status(RPL_ACTIVE_SLAVE,RPL_LOST_SOLDIER); break; } safe_sleep(thd,mi->connect_retry,(CHECK_KILLED_FUNC)io_slave_killed, (void*)mi); } if (!slave_was_killed) { if (reconnect) { if (!suppress_warnings && global_system_variables.log_warnings) sql_print_error("Slave: connected to master '%s@%s:%d',\ replication resumed in log '%s' at position %s", mi->user, mi->host, mi->port, IO_RPL_LOG_NAME, llstr(mi->master_log_pos,llbuff)); } else { change_rpl_status(RPL_IDLE_SLAVE,RPL_ACTIVE_SLAVE); mysql_log.write(thd, COM_CONNECT_OUT, "%s@%s:%d", mi->user, mi->host, mi->port); } #ifdef SIGNAL_WITH_VIO_CLOSE thd->set_active_vio(mysql->net.vio); #endif } DBUG_PRINT("exit",("slave_was_killed: %d", slave_was_killed)); DBUG_RETURN(slave_was_killed); } /* safe_reconnect() IMPLEMENTATION Try to connect until successful or slave killed or we have retried master_retry_count times */ static int safe_reconnect(THD* thd, MYSQL* mysql, MASTER_INFO* mi, bool suppress_warnings) { DBUG_ENTER("safe_reconnect"); DBUG_RETURN(connect_to_master(thd, mysql, mi, 1, suppress_warnings)); } /* Store the file and position where the execute-slave thread are in the relay log. SYNOPSIS flush_relay_log_info() rli Relay log information NOTES - As this is only called by the slave thread, we don't need to have a lock on this. - If there is an active transaction, then we don't update the position in the relay log. This is to ensure that we re-execute statements if we die in the middle of an transaction that was rolled back. - As a transaction never spans binary logs, we don't have to handle the case where we do a relay-log-rotation in the middle of the transaction. If this would not be the case, we would have to ensure that we don't delete the relay log file where the transaction started when we switch to a new relay log file. TODO - Change the log file information to a binary format to avoid calling longlong2str. RETURN VALUES 0 ok 1 write error */ bool flush_relay_log_info(RELAY_LOG_INFO* rli) { bool error=0; IO_CACHE *file = &rli->info_file; char buff[FN_REFLEN*2+22*2+4], *pos; my_b_seek(file, 0L); pos=strmov(buff, rli->group_relay_log_name); *pos++='\n'; pos=longlong2str(rli->group_relay_log_pos, pos, 10); *pos++='\n'; pos=strmov(pos, rli->group_master_log_name); *pos++='\n'; pos=longlong2str(rli->group_master_log_pos, pos, 10); *pos='\n'; if (my_b_write(file, (byte*) buff, (ulong) (pos-buff)+1)) error=1; if (flush_io_cache(file)) error=1; if (flush_io_cache(rli->cur_log)) // QQ Why this call ? error=1; return error; } /* Called when we notice that the current "hot" log got rotated under our feet. */ static IO_CACHE *reopen_relay_log(RELAY_LOG_INFO *rli, const char **errmsg) { DBUG_ASSERT(rli->cur_log != &rli->cache_buf); DBUG_ASSERT(rli->cur_log_fd == -1); DBUG_ENTER("reopen_relay_log"); IO_CACHE *cur_log = rli->cur_log=&rli->cache_buf; if ((rli->cur_log_fd=open_binlog(cur_log,rli->event_relay_log_name, errmsg)) <0) DBUG_RETURN(0); /* We want to start exactly where we was before: relay_log_pos Current log pos pending Number of bytes already processed from the event */ my_b_seek(cur_log,rli->event_relay_log_pos); DBUG_RETURN(cur_log); } Log_event* next_event(RELAY_LOG_INFO* rli) { Log_event* ev; IO_CACHE* cur_log = rli->cur_log; pthread_mutex_t *log_lock = rli->relay_log.get_log_lock(); const char* errmsg=0; THD* thd = rli->sql_thd; DBUG_ENTER("next_event"); DBUG_ASSERT(thd != 0); /* For most operations we need to protect rli members with data_lock, so we will hold it for the most of the loop below However, we will release it whenever it is worth the hassle, and in the cases when we go into a pthread_cond_wait() with the non-data_lock mutex */ pthread_mutex_lock(&rli->data_lock); while (!sql_slave_killed(thd,rli)) { /* We can have two kinds of log reading: hot_log: rli->cur_log points at the IO_CACHE of relay_log, which is actively being updated by the I/O thread. We need to be careful in this case and make sure that we are not looking at a stale log that has already been rotated. If it has been, we reopen the log. The other case is much simpler: We just have a read only log that nobody else will be updating. */ bool hot_log; if ((hot_log = (cur_log != &rli->cache_buf))) { DBUG_ASSERT(rli->cur_log_fd == -1); // foreign descriptor pthread_mutex_lock(log_lock); /* Reading xxx_file_id is safe because the log will only be rotated when we hold relay_log.LOCK_log */ if (rli->relay_log.get_open_count() != rli->cur_log_old_open_count) { // The master has switched to a new log file; Reopen the old log file cur_log=reopen_relay_log(rli, &errmsg); pthread_mutex_unlock(log_lock); if (!cur_log) // No more log files goto err; hot_log=0; // Using old binary log } } DBUG_ASSERT(my_b_tell(cur_log) >= BIN_LOG_HEADER_SIZE); DBUG_ASSERT(my_b_tell(cur_log) == rli->event_relay_log_pos); /* Relay log is always in new format - if the master is 3.23, the I/O thread will convert the format for us */ if ((ev=Log_event::read_log_event(cur_log,0,(bool)0 /* new format */))) { DBUG_ASSERT(thd==rli->sql_thd); if (hot_log) pthread_mutex_unlock(log_lock); pthread_mutex_unlock(&rli->data_lock); DBUG_RETURN(ev); } DBUG_ASSERT(thd==rli->sql_thd); if (opt_reckless_slave) // For mysql-test cur_log->error = 0; if (cur_log->error < 0) { errmsg = "slave SQL thread aborted because of I/O error"; if (hot_log) pthread_mutex_unlock(log_lock); goto err; } if (!cur_log->error) /* EOF */ { /* On a hot log, EOF means that there are no more updates to process and we must block until I/O thread adds some and signals us to continue */ if (hot_log) { DBUG_ASSERT(rli->relay_log.get_open_count() == rli->cur_log_old_open_count); /* We can, and should release data_lock while we are waiting for update. If we do not, show slave status will block */ pthread_mutex_unlock(&rli->data_lock); /* Possible deadlock : - the I/O thread has reached log_space_limit - the SQL thread has read all relay logs, but cannot purge for some reason: * it has already purged all logs except the current one * there are other logs than the current one but they're involved in a transaction that finishes in the current one (or is not finished) Solution : Wake up the possibly waiting I/O thread, and set a boolean asking the I/O thread to temporarily ignore the log_space_limit constraint, because we do not want the I/O thread to block because of space (it's ok if it blocks for any other reason (e.g. because the master does not send anything). Then the I/O thread stops waiting and reads more events. The SQL thread decides when the I/O thread should take log_space_limit into account again : ignore_log_space_limit is reset to 0 in purge_first_log (when the SQL thread purges the just-read relay log), and also when the SQL thread starts. We should also reset ignore_log_space_limit to 0 when the user does RESET SLAVE, but in fact, no need as RESET SLAVE requires that the slave be stopped, and when the SQL thread is later restarted ignore_log_space_limit will be reset to 0. */ pthread_mutex_lock(&rli->log_space_lock); // prevent the I/O thread from blocking next times rli->ignore_log_space_limit= 1; /* If the I/O thread is blocked, unblock it. Ok to broadcast after unlock, because the mutex is only destroyed in ~st_relay_log_info(), i.e. when rli is destroyed, and rli will not be destroyed before we exit the present function. */ pthread_mutex_unlock(&rli->log_space_lock); pthread_cond_broadcast(&rli->log_space_cond); // Note that wait_for_update unlocks lock_log ! rli->relay_log.wait_for_update(rli->sql_thd); // re-acquire data lock since we released it earlier pthread_mutex_lock(&rli->data_lock); continue; } /* If the log was not hot, we need to move to the next log in sequence. The next log could be hot or cold, we deal with both cases separately after doing some common initialization */ end_io_cache(cur_log); DBUG_ASSERT(rli->cur_log_fd >= 0); my_close(rli->cur_log_fd, MYF(MY_WME)); rli->cur_log_fd = -1; if (relay_log_purge) { /* purge_first_log will properly set up relay log coordinates in rli. If the group's coordinates are equal to the event's coordinates (i.e. the relay log was not rotated in the middle of a group), we can purge this relay log too. We do ulonglong and string comparisons, this may be slow but - purging the last relay log is nice (it can save 1GB of disk), so we like to detect the case where we can do it, and given this, - I see no better detection method - purge_first_log is not called that often */ if (rli->relay_log.purge_first_log (rli, rli->group_relay_log_pos == rli->event_relay_log_pos && !strcmp(rli->group_relay_log_name,rli->event_relay_log_name))) { errmsg = "Error purging processed logs"; goto err; } } else { /* If hot_log is set, then we already have a lock on LOCK_log. If not, we have to get the lock. According to Sasha, the only time this code will ever be executed is if we are recovering from a bug. */ if (rli->relay_log.find_next_log(&rli->linfo, !hot_log)) { errmsg = "error switching to the next log"; goto err; } rli->event_relay_log_pos = BIN_LOG_HEADER_SIZE; strmake(rli->event_relay_log_name,rli->linfo.log_file_name, sizeof(rli->event_relay_log_name)-1); flush_relay_log_info(rli); } // next log is hot if (rli->relay_log.is_active(rli->linfo.log_file_name)) { #ifdef EXTRA_DEBUG sql_print_error("next log '%s' is currently active", rli->linfo.log_file_name); #endif rli->cur_log= cur_log= rli->relay_log.get_log_file(); rli->cur_log_old_open_count= rli->relay_log.get_open_count(); DBUG_ASSERT(rli->cur_log_fd == -1); /* Read pointer has to be at the start since we are the only reader */ if (check_binlog_magic(cur_log,&errmsg)) goto err; continue; } /* if we get here, the log was not hot, so we will have to open it ourselves */ #ifdef EXTRA_DEBUG sql_print_error("next log '%s' is not active", rli->linfo.log_file_name); #endif // open_binlog() will check the magic header if ((rli->cur_log_fd=open_binlog(cur_log,rli->linfo.log_file_name, &errmsg)) <0) goto err; } else { /* Read failed with a non-EOF error. TODO: come up with something better to handle this error */ if (hot_log) pthread_mutex_unlock(log_lock); sql_print_error("Slave SQL thread: I/O error reading \ event(errno: %d cur_log->error: %d)", my_errno,cur_log->error); // set read position to the beginning of the event my_b_seek(cur_log,rli->event_relay_log_pos); /* otherwise, we have had a partial read */ errmsg = "Aborting slave SQL thread because of partial event read"; break; // To end of function } } if (!errmsg && global_system_variables.log_warnings) errmsg = "slave SQL thread was killed"; err: pthread_mutex_unlock(&rli->data_lock); if (errmsg) sql_print_error("Error reading relay log event: %s", errmsg); DBUG_RETURN(0); } #ifdef __GNUC__ template class I_List_iterator<i_string>; template class I_List_iterator<i_string_pair>; #endif #endif /* HAVE_REPLICATION */