Merge branch 'mysql-8.0' into mysql-8.4

Change-Id: I2177201a6a55d24ba3d76ca350d3836ef8cef149

Author: Justin Jose

mysql-test/suite/rpl_gtid/r/rpl_replica_transaction_retry_with_non_standared_relay_log_name.result

@@ -0,0 +1,39 @@
+include/only_with_option.inc [GLOBAL.replica_transaction_retries > 1]
+#
+# 1. Create source-replica topology
+include/rpl/init_source_replica.inc
+Warnings:
+Note	####	Sending passwords in plain text without SSL/TLS is extremely insecure.
+Note	####	Storing MySQL user name or password information in the connection metadata repository is not secure and is therefore not recommended. Please consider using the USER and PASSWORD connection options for START REPLICA; see the 'START REPLICA Syntax' in the MySQL Manual for more information.
+[connection master]
+[connection slave]
+CALL mtr.add_suppression('A replica with the same server_uuid/server_id as this replica has connected to the source.*');
+#
+# 2. Copy test relay-log files to replica data directory
+#
+# 3. Create a new channel on replica
+CHANGE REPLICATION SOURCE TO RELAY_LOG_FILE='replicate-relay-bin.000001', RELAY_LOG_POS=4, SOURCE_USER='root', SOURCE_HOST='127.0.0.1', SOURCE_PORT=MASTER_PORT FOR CHANNEL 'PITR';
+Warnings:
+Note	1759	Sending passwords in plain text without SSL/TLS is extremely insecure.
+Note	1760	Storing MySQL user name or password information in the connection metadata repository is not secure and is therefore not recommended. Please consider using the USER and PASSWORD connection options for START REPLICA; see the 'START REPLICA Syntax' in the MySQL Manual for more information.
+#
+# 4. Add a debug point to simulate replica retry transaction
+# Adding debug point 'simulate_error_ha_delete_row_lock_wait_timeout' to @@GLOBAL.debug
+#
+# 5. Start replica
+include/rpl/start_replica.inc [FOR CHANNEL 'PITR']
+include/rpl/assert_replica_running.inc [FOR CHANNEL 'PITR']
+include/rpl/assert_replica_no_error.inc [FOR CHANNEL 'PITR']
+#
+# 6. Wait till replica apply all txns successfully
+include/rpl/wait_for_replica_status.inc [Replica_SQL_Running_State FOR CHANNEL 'PITR']
+#
+# 7. Verify that there is a retry transaction warning
+include/assert_grep.inc [Found error message regarding retry transaction]
+#
+# 8. Cleanup
+# Removing debug point 'simulate_error_ha_delete_row_lock_wait_timeout' from @@GLOBAL.debug
+include/rpl/stop_replica.inc [FOR CHANNEL 'PITR']
+RESET REPLICA ALL FOR CHANNEL 'PITR';
+include/rpl/start_replica.inc
+include/rpl/deinit.inc

mysql-test/suite/rpl_gtid/t/rpl_replica_transaction_retry_with_non_standared_relay_log_name.test

@@ -0,0 +1,95 @@
+##############################################################################
+# ==== Purpose ====
+# The purpose of this test is to verify that, when an issue occurs in the
+# applier and it identifies the transaction can be retried, applier uses
+# relay log file name mentioned in the index file instead of calculating
+# the file name based on the configured server option(--relay-log).
+#
+# ==== Requirement ====
+# When an issue occurs in the applier and it identifies the transaction can
+# be retired, the applier should use the relaylog file name mentioned in the
+# index file instead of calculating the file name from the configured
+# server option.
+#
+# ==== Implementation ====
+# 1. Create source-replica topology
+# 2. Copy test relay-log files to replica data directory
+# 3. Create a new channel on replica
+# 4. Add a debug point to simulate replica retry transaction
+# 5. Start replica
+# 6. Wait till replica apply all txns successfully
+# 7. Verify that there is a retry transaction warning
+# 8. Clean up
+#
+# ==== References ====
+# BUG#36395631:Replica applier retries ignore the relay log index file
+###############################################################################
+
+
+--source include/have_debug.inc
+--source include/have_binlog_format_row.inc
+--let $option_name = replica_transaction_retries
+--let $option_operator = >
+--let $option_value = 1
+--source include/only_with_option.inc
+
+--echo #
+--echo # 1. Create source-replica topology
+--let $rpl_skip_start_slave=1
+--source include/rpl/init_source_replica.inc
+
+--source include/rpl/connection_replica.inc
+CALL mtr.add_suppression('A replica with the same server_uuid/server_id as this replica has connected to the source.*');
+
+--echo #
+--echo # 2. Copy test relay-log files to replica data directory
+--let $datadir= `SELECT @@datadir`
+--copy_file $MYSQL_TEST_DIR/std_data/replicated-bin.000001 $datadir/replicate-relay-bin.000001
+--copy_file $MYSQL_TEST_DIR/std_data/replicated-bin.000002 $datadir/replicate-relay-bin.000002
+--copy_file $MYSQL_TEST_DIR/std_data/replicated-bin.index $datadir/slave-relay-bin-pitr.index
+
+--echo #
+--echo # 3. Create a new channel on replica
+--replace_result $MASTER_MYPORT MASTER_PORT
+--eval CHANGE REPLICATION SOURCE TO RELAY_LOG_FILE='replicate-relay-bin.000001', RELAY_LOG_POS=4, SOURCE_USER='root', SOURCE_HOST='127.0.0.1', SOURCE_PORT=$MASTER_MYPORT FOR CHANNEL 'PITR'
+
+--echo #
+--echo # 4. Add a debug point to simulate replica retry transaction
+--let $debug_point= simulate_error_ha_delete_row_lock_wait_timeout
+--source include/add_debug_point.inc
+
+--echo #
+--echo # 5. Start replica
+--let $rpl_channel_name= 'PITR'
+--source include/rpl/start_replica.inc
+--source include/rpl/assert_replica_running.inc
+--source include/rpl/assert_replica_no_error.inc
+
+--echo #
+--echo # 6. Wait till replica apply all txns successfully
+--let $slave_param= Replica_SQL_Running_State
+--let $slave_param_value= Replica has read all relay log; waiting for more updates
+--source include/rpl/wait_for_replica_status.inc
+
+--echo #
+--echo # 7. Verify that there is a retry transaction warning
+--let $assert_file= $MYSQLTEST_VARDIR/log/mysqld.2.err
+--let $assert_match= Error 'Lock wait timeout exceeded; try restarting transaction' on query
+--let $assert_select= Replica SQL
+--let $assert_text= Found error message regarding retry transaction
+--source include/assert_grep.inc
+
+--echo #
+--echo # 8. Cleanup
+--let $debug_point= simulate_error_ha_delete_row_lock_wait_timeout
+--source include/remove_debug_point.inc
+
+--let $rpl_channel_name= 'PITR'
+--source include/rpl/stop_replica.inc
+
+RESET REPLICA ALL FOR CHANNEL 'PITR';
+
+--let $rpl_channel_name=
+source include/rpl/start_replica.inc;
+
+--source include/rpl/deinit.inc

sql/handler.cc

@@ -8113,6 +8113,10 @@ int handler::ha_delete_row(const uchar *buf) {
   */
   assert(buf == table->record[0] || buf == table->record[1]);
   DBUG_EXECUTE_IF("inject_error_ha_delete_row", return HA_ERR_INTERNAL_ERROR;);
+  DBUG_EXECUTE_IF("simulate_error_ha_delete_row_lock_wait_timeout", {
+    DBUG_SET("-d,simulate_error_ha_delete_row_lock_wait_timeout");
+    return HA_ERR_LOCK_WAIT_TIMEOUT;
+  });
 
   DBUG_EXECUTE_IF(
       "handler_crashed_table_on_usage",

sql/log_event.cc

@@ -2628,8 +2628,10 @@ Slave_worker *Log_event::get_slave_worker(Relay_log_info *rli) {
              !is_mts_db_partitioned(rli));
 
       if (is_s_event || is_any_gtid_event(this)) {
-        Slave_job_item job_item = {this, rli->get_event_relay_log_number(),
-                                   rli->get_event_start_pos()};
+        Slave_job_item job_item = {this, rli->get_event_start_pos(), {'\0'}};
+        if (rli->get_event_relay_log_name())
+          strcpy(job_item.event_relay_log_name,
+                 rli->get_event_relay_log_name());
         // B-event is appended to the Deferred Array associated with GCAP
         rli->curr_group_da.push_back(job_item);
 
@@ -2664,8 +2666,9 @@ Slave_worker *Log_event::get_slave_worker(Relay_log_info *rli) {
        TODO: Make GITD event as B-event that is starts_group() to
        return true.
       */
-      Slave_job_item job_item = {this, rli->get_event_relay_log_number(),
-                                 rli->get_event_relay_log_pos()};
+      Slave_job_item job_item = {this, rli->get_event_relay_log_pos(), {'\0'}};
+      if (rli->get_event_relay_log_name())
+        strcpy(job_item.event_relay_log_name, rli->get_event_relay_log_name());
 
       // B-event is appended to the Deferred Array associated with GCAP
       rli->curr_group_da.push_back(job_item);
@@ -2693,8 +2696,9 @@ Slave_worker *Log_event::get_slave_worker(Relay_log_info *rli) {
         rli, &rli->workers, this);
     if (ret_worker == nullptr) {
       /* get_least_occupied_worker may return NULL if the thread is killed */
-      Slave_job_item job_item = {this, rli->get_event_relay_log_number(),
-                                 rli->get_event_start_pos()};
+      Slave_job_item job_item = {this, rli->get_event_start_pos(), {'\0'}};
+      if (rli->get_event_relay_log_name())
+        strcpy(job_item.event_relay_log_name, rli->get_event_relay_log_name());
       rli->curr_group_da.push_back(job_item);
 
       assert(thd->killed);
@@ -2845,8 +2849,9 @@ Slave_worker *Log_event::get_slave_worker(Relay_log_info *rli) {
         Their association with relay-log physical coordinates is provided
         by the same mechanism that applies to a regular event.
       */
-      Slave_job_item job_item = {this, rli->get_event_relay_log_number(),
-                                 rli->get_event_start_pos()};
+      Slave_job_item job_item = {this, rli->get_event_start_pos(), {'\0'}};
+      if (rli->get_event_relay_log_name())
+        strcpy(job_item.event_relay_log_name, rli->get_event_relay_log_name());
       rli->curr_group_da.push_back(job_item);
 
       assert(!ret_worker);

sql/rpl_replica.cc

@@ -4524,8 +4524,9 @@ apply_event_and_update_pos(Log_event **ptr_ev, THD *thd, Relay_log_info *rli) {
 
     if (!exec_res && (ev->worker != rli)) {
       if (ev->worker) {
-        Slave_job_item item = {ev, rli->get_event_relay_log_number(),
-                               rli->get_event_start_pos()};
+        Slave_job_item item = {ev, rli->get_event_start_pos(), {'\0'}};
+        if (rli->get_event_relay_log_name())
+          strcpy(item.event_relay_log_name, rli->get_event_relay_log_name());
         Slave_job_item *job_item = &item;
         Slave_worker *w = (Slave_worker *)ev->worker;
         // specially marked group typically with OVER_MAX_DBS_IN_EVENT_MTS db:s
@@ -6763,7 +6764,7 @@ static void slave_stop_workers(Relay_log_info *rli, bool *mts_inited) {
   if (!rli->workers.empty()) {
     for (int i = static_cast<int>(rli->workers.size()) - 1; i >= 0; i--) {
       Slave_worker *w = rli->workers[i];
-      struct slave_job_item item = {nullptr, 0, 0};
+      struct slave_job_item item = {nullptr, 0, {'\0'}};
       struct slave_job_item *job_item = &item;
       mysql_mutex_lock(&w->jobs_lock);
 

sql/rpl_rli.cc

@@ -140,7 +140,6 @@ Relay_log_info::Relay_log_info(bool is_slave_recovery,
       transaction_parser(mysql::binlog::event::Transaction_boundary_parser::
                              TRX_BOUNDARY_PARSER_APPLIER),
       group_relay_log_pos(0),
-      event_relay_log_number(0),
       event_relay_log_pos(0),
       group_source_log_seen_start_pos(false),
       group_source_log_start_pos(0),
@@ -2682,23 +2681,6 @@ ulong Relay_log_info::adapt_to_master_version_updown(ulong master_version,
   return master_version;
 }
 
-void Relay_log_info::relay_log_number_to_name(uint number,
-                                              char name[FN_REFLEN + 1]) {
-  char *str = nullptr;
-  char relay_bin_channel[FN_REFLEN + 1];
-  const char *relay_log_basename_channel = add_channel_to_relay_log_name(
-      relay_bin_channel, FN_REFLEN + 1, relay_log_basename);
-
-  /* str points to closing null of relay log basename channel */
-  str = strmake(name, relay_log_basename_channel, FN_REFLEN + 1);
-  *str++ = '.';
-  sprintf(str, "%06u", number);
-}
-
-uint Relay_log_info::relay_log_name_to_number(const char *name) {
-  return static_cast<uint>(atoi(fn_ext(name) + 1));
-}
-
 bool is_mts_db_partitioned(Relay_log_info *rli) {
   return (rli->current_mts_submode->get_type() == MTS_PARALLEL_TYPE_DB_NAME);
 }

sql/rpl_rli.h

@@ -81,8 +81,8 @@ typedef Prealloced_array<Slave_worker *, 4> Slave_worker_array;
 
 typedef struct slave_job_item {
   Log_event *data;
-  uint relay_number;
   my_off_t relay_pos;
+  char event_relay_log_name[FN_REFLEN + 1];
 } Slave_job_item;
 
 /**
@@ -699,8 +699,6 @@ class Relay_log_info : public Rpl_info {
   char group_relay_log_name[FN_REFLEN];
   ulonglong group_relay_log_pos;
   char event_relay_log_name[FN_REFLEN];
-  /* The suffix number of relay log name */
-  uint event_relay_log_number;
   ulonglong event_relay_log_pos;
   ulonglong future_event_relay_log_pos;
 
@@ -1647,26 +1645,9 @@ class Relay_log_info : public Rpl_info {
   inline void set_event_relay_log_name(const char *log_file_name) {
     strmake(event_relay_log_name, log_file_name,
             sizeof(event_relay_log_name) - 1);
-    set_event_relay_log_number(relay_log_name_to_number(log_file_name));
     notify_relay_log_change();
   }
 
-  uint get_event_relay_log_number() { return event_relay_log_number; }
-  void set_event_relay_log_number(uint number) {
-    event_relay_log_number = number;
-  }
-
-  /**
-    Given the extension number of the relay log, gets the full
-    relay log path. Currently used in Slave_worker::retry_transaction()
-
-    @param [in]   number      extension number of relay log
-    @param[in, out] name      The full path of the relay log (per-channel)
-                              to be read by the slave worker.
-  */
-  void relay_log_number_to_name(uint number, char name[FN_REFLEN + 1]);
-  uint relay_log_name_to_number(const char *name);
-
   void set_event_start_pos(my_off_t pos) { event_start_pos = pos; }
   my_off_t get_event_start_pos() { return event_start_pos; }