Backport BUG#47298 Semisync: always wait until timeout if no semi-sync slave available

Add an option to control whether the master should keep waiting
until timeout when it detected that there is no semi-sync slave
available.

The bool option 'rpl_semi_sync_master_wait_no_slave' is 1 by
defalt, and will keep waiting until timeout. When set to 0, the
master will switch to asynchronous replication immediately when
no semi-sync slave is available.

mysql-test/suite/rpl/r/rpl_semi_sync.result

@@ -29,13 +29,13 @@ set global rpl_semi_sync_master_enabled = 1;
 show variables like 'rpl_semi_sync_master_enabled';
 Variable_name	Value
 rpl_semi_sync_master_enabled	ON
-[ status of semi-sync on master should be OFF without any semi-sync slaves ]
+[ status of semi-sync on master should be ON even without any semi-sync slaves ]
 show status like 'Rpl_semi_sync_master_clients';
 Variable_name	Value
 Rpl_semi_sync_master_clients	0
 show status like 'Rpl_semi_sync_master_status';
 Variable_name	Value
-Rpl_semi_sync_master_status	OFF
+Rpl_semi_sync_master_status	ON
 show status like 'Rpl_semi_sync_master_yes_tx';
 Variable_name	Value
 Rpl_semi_sync_master_yes_tx	0
@@ -210,7 +210,7 @@ include/stop_slave.inc
 [ Semi-sync master status variables before FLUSH STATUS ]
 SHOW STATUS LIKE 'Rpl_semi_sync_master_no_tx';
 Variable_name	Value
-Rpl_semi_sync_master_no_tx	3
+Rpl_semi_sync_master_no_tx	302
 SHOW STATUS LIKE 'Rpl_semi_sync_master_yes_tx';
 Variable_name	Value
 Rpl_semi_sync_master_yes_tx	302
@@ -355,7 +355,7 @@ Variable_name	Value
 Rpl_semi_sync_master_clients	0
 show status like 'Rpl_semi_sync_master_status';
 Variable_name	Value
-Rpl_semi_sync_master_status	OFF
+Rpl_semi_sync_master_status	ON
 set global rpl_semi_sync_master_enabled= 0;
 [ on slave ]
 SHOW VARIABLES LIKE 'rpl_semi_sync_slave_enabled';
@@ -364,17 +364,17 @@ rpl_semi_sync_slave_enabled	ON
 include/start_slave.inc
 [ on master ]
 insert into t1 values (8);
-[ master semi-sync clients should be 0, status should be OFF ]
+[ master semi-sync clients should be 1, status should be OFF ]
 show status like 'Rpl_semi_sync_master_clients';
 Variable_name	Value
-Rpl_semi_sync_master_clients	0
+Rpl_semi_sync_master_clients	1
 show status like 'Rpl_semi_sync_master_status';
 Variable_name	Value
 Rpl_semi_sync_master_status	OFF
 [ on slave ]
 show status like 'Rpl_semi_sync_slave_status';
 Variable_name	Value
-Rpl_semi_sync_slave_status	OFF
+Rpl_semi_sync_slave_status	ON
 include/stop_slave.inc
 [ on master ]
 set sql_log_bin=0;

mysql-test/suite/rpl/t/rpl_semi_sync.test

@@ -74,7 +74,7 @@ echo [ enable semi-sync on master ];
 set global rpl_semi_sync_master_enabled = 1;
 show variables like 'rpl_semi_sync_master_enabled';
 
-echo [ status of semi-sync on master should be OFF without any semi-sync slaves ];
+echo [ status of semi-sync on master should be ON even without any semi-sync slaves ];
 show status like 'Rpl_semi_sync_master_clients';
 show status like 'Rpl_semi_sync_master_status';
 show status like 'Rpl_semi_sync_master_yes_tx';
@@ -497,7 +497,10 @@ source include/start_slave.inc;
 connection master;
 echo [ on master ];
 insert into t1 values (8);
-echo [ master semi-sync clients should be 0, status should be OFF ];
+let $status_var= Rpl_semi_sync_master_clients;
+let $status_var_value= 1;
+source include/wait_for_status_var.inc;
+echo [ master semi-sync clients should be 1, status should be OFF ];
 show status like 'Rpl_semi_sync_master_clients';
 show status like 'Rpl_semi_sync_master_status';
 sync_slave_with_master;

plugin/semisync/semisync_master.cc

@@ -40,6 +40,7 @@ unsigned long long rpl_semi_sync_master_net_wait_num = 0;
 unsigned long rpl_semi_sync_master_clients          = 0;
 unsigned long long rpl_semi_sync_master_net_wait_time = 0;
 unsigned long long rpl_semi_sync_master_trx_wait_time = 0;
+char rpl_semi_sync_master_wait_no_slave = 1;
 
 
 static int getWaitTime(const struct timeval& start_tv);
@@ -525,6 +526,14 @@ void ReplSemiSyncMaster::remove_slave()
 {
   lock();
   rpl_semi_sync_master_clients--;
+
+  /* If user has chosen not to wait if no semi-sync slave available
+     and the last semi-sync slave exits, turn off semi-sync on master
+     immediately.
+   */
+  if (!rpl_semi_sync_master_wait_no_slave &&
+      rpl_semi_sync_master_clients == 0)
+    switch_off();
   unlock();
 }
 
@@ -812,7 +821,7 @@ int ReplSemiSyncMaster::commitTrx(const char* trx_wait_binlog_name,
                                              trx_wait_binlog_pos));
     
     /* Update the status counter. */
-    if (is_on() && rpl_semi_sync_master_clients)
+    if (is_on())
       rpl_semi_sync_master_yes_transactions++;
     else
       rpl_semi_sync_master_no_transactions++;
@@ -1078,7 +1087,7 @@ int ReplSemiSyncMaster::writeTranxInBinlog(const char* log_file_name,
     commit_file_name_inited_ = true;
   }
 
-  if (is_on() && rpl_semi_sync_master_clients)
+  if (is_on())
   {
     assert(active_tranxs_ != NULL);
     if(active_tranxs_->insert_tranx_node(log_file_name, log_file_pos))
@@ -1153,7 +1162,7 @@ int ReplSemiSyncMaster::readSlaveReply(NET *net, uint32 server_id,
     {
       sql_print_error("Semi-sync master wait for reply "
                       "gettimeofday fail to get start time");
-      timefunc_fails_++;
+      rpl_semi_sync_master_timefunc_fails++;
     }
     else
     {
@@ -1164,12 +1173,12 @@ int ReplSemiSyncMaster::readSlaveReply(NET *net, uint32 server_id,
       {
         sql_print_error("Semi-sync master wait for reply "
                         "gettimeofday fail to get wait time.");
-        timefunc_fails_++;
+        rpl_semi_sync_master_timefunc_fails++;
       }
       else
       {
-        total_net_wait_num_++;
-        total_net_wait_time_ += wait_time;
+        rpl_semi_sync_master_net_wait_num++;
+        rpl_semi_sync_master_net_wait_time += wait_time;
       }
     }
   }
@@ -1242,7 +1251,7 @@ void ReplSemiSyncMaster::setExportStats()
 {
   lock();
 
-  rpl_semi_sync_master_status           = state_ && rpl_semi_sync_master_clients;
+  rpl_semi_sync_master_status           = state_;
   rpl_semi_sync_master_avg_trx_wait_time=
     ((rpl_semi_sync_master_trx_wait_num) ?
      (unsigned long)((double)rpl_semi_sync_master_trx_wait_time /

plugin/semisync/semisync_master.h

@@ -363,4 +363,12 @@ extern unsigned long long rpl_semi_sync_master_trx_wait_num;
 extern unsigned long long rpl_semi_sync_master_net_wait_time;
 extern unsigned long long rpl_semi_sync_master_trx_wait_time;
 
+/*
+  This indicates whether we should keep waiting if no semi-sync slave
+  is available.
+     0           : stop waiting if detected no avaialable semi-sync slave.
+     1 (default) : keep waiting until timeout even no available semi-sync slave.
+*/
+extern char rpl_semi_sync_master_wait_no_slave;
+
 #endif /* SEMISYNC_MASTER_H */

plugin/semisync/semisync_master_plugin.cc

@@ -76,7 +76,7 @@ int repl_semi_binlog_dump_start(Binlog_transmit_param *param,
     /*
       Let's assume this semi-sync slave has already received all
       binlog events before the filename and position it requests.
-  */
+    */
     repl_semisync.reportReplyBinlog(param->server_id, log_file, log_pos);
   }
   sql_print_information("Start %s binlog_dump to slave (server_id: %d), pos(%s, %lu)",
@@ -176,6 +176,13 @@ static MYSQL_SYSVAR_ULONG(timeout, rpl_semi_sync_master_timeout,
   fix_rpl_semi_sync_master_timeout,	// update
   10000, 0, ~0L, 1);
 
+static MYSQL_SYSVAR_BOOL(wait_no_slave, rpl_semi_sync_master_wait_no_slave,
+  PLUGIN_VAR_OPCMDARG,
+ "Wait until timeout when no semi-synchronous replication slave available (enabled by default). ",
+  NULL, 			// check
+  NULL,                         // update
+  1);
+
 static MYSQL_SYSVAR_ULONG(trace_level, rpl_semi_sync_master_trace_level,
   PLUGIN_VAR_OPCMDARG,
  "The tracing level for semi-sync replication.",
@@ -186,6 +193,7 @@ static MYSQL_SYSVAR_ULONG(trace_level, rpl_semi_sync_master_trace_level,
 static SYS_VAR* semi_sync_master_system_vars[]= {
   MYSQL_SYSVAR(enabled),
   MYSQL_SYSVAR(timeout),
+  MYSQL_SYSVAR(wait_no_slave),
   MYSQL_SYSVAR(trace_level),
   NULL,
 };

plugin/semisync/semisync_slave_plugin.cc

@@ -56,10 +56,11 @@ int repl_semi_slave_request_dump(Binlog_relay_IO_param *param,
   }
 
   row= mysql_fetch_row(res);
-  if (!row || strcmp(row[1], "ON"))
+  if (!row)
   {
-    /* Master does not support or not configured semi-sync */
-    sql_print_warning("Master server does not support or not configured semi-sync replication, fallback to asynchronous");
+    /* Master does not support semi-sync */
+    sql_print_warning("Master server does not support semi-sync, "
+                      "fallback to asynchronous replication");
     rpl_semi_sync_slave_status= 0;
     return 0;
   }