fixed message in post-commit trigger

added debugging options to master to behave sporadically, and a 
test case to make sure the slave can deal with it


BitKeeper/triggers/post-commit:
  fixed message
sql/mysqld.cc:
  added debugging options to force the master to occasionally break
  replication
sql/slave.cc:
  cosmetic change to improve readability
sql/sql_repl.cc:
  support for --sporadic-binlog-dump-fail and --max-binlog-dump-events
sql/sql_repl.h:
  support for --sporadic-binlog-dump-fail and --max-binlog-dump-events

10 files changed, 108 insertions, 12 deletions

diff --git a/BitKeeper/triggers/post-commit b/BitKeeper/triggers/post-commit
index 619b6f88cea..dc5f2f2b824 100755
--- a/BitKeeper/triggers/post-commit
+++ b/BitKeeper/triggers/post-commit
@@ -37,9 +37,12 @@ From: $FROM
 To: $INTERNALS
 Subject: bk commit into 3.23 tree
 
-Below is the list of changes that have just been pushed into main
-3.23. repository. For information on how to access the repository
-see http://www.mysql.com/doc/I/n/Installing_source_tree.html
+Below is the list of changes that have just been commited into a local
+3.23. repository of $USER. When $USER does a push, they will be
+propogaged to the main repository and within 24 hours after the push into
+the public repository. For information on how to access
+the public repository see
+http://www.mysql.com/doc/I/n/Installing_source_tree.html
 
 EOF
   bk changes -v -r+
diff --git a/BitKeeper/triggers/post-incoming b/BitKeeper/triggers/post-incoming
new file mode 100644
index 00000000000..f1ea2255de9
--- /dev/null
+++ b/BitKeeper/triggers/post-incoming
@@ -0,0 +1,3 @@
+#! /bin/sh
+
+echo "Test: post-incoming works"
diff --git a/BitKeeper/triggers/post-outgoing b/BitKeeper/triggers/post-outgoing
new file mode 100755
index 00000000000..3fc2cdbad67
--- /dev/null
+++ b/BitKeeper/triggers/post-outgoing
@@ -0,0 +1,3 @@
+#! /bin/sh
+
+echo "Test: post-outgoing works"
diff --git a/mysql-test/r/rpl_sporadic_master.result b/mysql-test/r/rpl_sporadic_master.result
new file mode 100644
index 00000000000..414468f0998
--- /dev/null
+++ b/mysql-test/r/rpl_sporadic_master.result
@@ -0,0 +1,7 @@
+n
+1
+2
+3
+4
+5
+6
diff --git a/mysql-test/t/rpl_sporadic_master-master.opt b/mysql-test/t/rpl_sporadic_master-master.opt
new file mode 100644
index 00000000000..6d5b66bed61
--- /dev/null
+++ b/mysql-test/t/rpl_sporadic_master-master.opt
@@ -0,0 +1 @@
+--sporadic-binlog-dump-fail --max-binlog-dump-events=2
diff --git a/mysql-test/t/rpl_sporadic_master.test b/mysql-test/t/rpl_sporadic_master.test
new file mode 100644
index 00000000000..ce6568f659e
--- /dev/null
+++ b/mysql-test/t/rpl_sporadic_master.test
@@ -0,0 +1,24 @@
+#test to see if replication can continue when master sporadically fails on
+# COM_BINLOG_DUMP and additionally limits the number of events per dump
+source include/master-slave.inc;
+connection master;
+drop table if exists t1;
+create table t1(n int not null auto_increment primary key);
+insert into t1 values (NULL),(NULL);
+delete from t1;
+insert into t1 values (NULL),(NULL);
+insert into t1 values (NULL),(NULL);
+flush logs;
+delete from t1;
+insert into t1 values (NULL),(NULL);
+insert into t1 values (NULL),(NULL);
+insert into t1 values (NULL),(NULL);
+save_master_pos;
+connection slave;
+sync_with_master;
+select * from t1;
+connection master;
+drop table t1;
+save_master_pos;
+connection slave;
+sync_with_master;
diff --git a/sql/mysqld.cc b/sql/mysqld.cc
index 991b0e73c51..361479b4365 100644
--- a/sql/mysqld.cc
+++ b/sql/mysqld.cc
@@ -20,6 +20,7 @@
 #include <my_dir.h>
 #include "sql_acl.h"
 #include "slave.h"
+#include "sql_repl.h"
 #include "stacktrace.h"
 #ifdef HAVE_BERKELEY_DB
 #include "ha_berkeley.h"
@@ -2461,7 +2462,8 @@ enum options {
                OPT_TEMP_POOL, OPT_TX_ISOLATION,
 	       OPT_GEMINI_FLUSH_LOG, OPT_GEMINI_RECOVER,
                OPT_GEMINI_UNBUFFERED_IO, OPT_SKIP_SAFEMALLOC,
-	       OPT_SKIP_STACK_TRACE, OPT_SKIP_SYMLINKS
+	       OPT_SKIP_STACK_TRACE, OPT_SKIP_SYMLINKS,
+	       OPT_MAX_BINLOG_DUMP_EVENTS, OPT_SPORADIC_BINLOG_DUMP_FAIL
 };
 
 static struct option long_options[] = {
@@ -2545,6 +2547,10 @@ static struct option long_options[] = {
      (int) OPT_DISCONNECT_SLAVE_EVENT_COUNT},
   {"abort-slave-event-count",      required_argument, 0,
      (int) OPT_ABORT_SLAVE_EVENT_COUNT},
+  {"max-binlog-dump-events",      required_argument, 0,
+     (int) OPT_MAX_BINLOG_DUMP_EVENTS},
+  {"sporadic-binlog-dump-fail", no_argument, 0,
+     (int) OPT_SPORADIC_BINLOG_DUMP_FAIL},
   {"safemalloc-mem-limit",  required_argument, 0, (int)
      OPT_SAFEMALLOC_MEM_LIMIT},
   {"new",                   no_argument,       0, 'n'},
@@ -3301,6 +3307,17 @@ static void get_options(int argc,char **argv)
       abort_slave_event_count = atoi(optarg);
 #endif      
       break;
+    case (int)OPT_SPORADIC_BINLOG_DUMP_FAIL:
+#ifndef DBUG_OFF      
+      opt_sporadic_binlog_dump_fail = 1;
+#endif      
+      break;
+     case (int)OPT_MAX_BINLOG_DUMP_EVENTS:
+#ifndef DBUG_OFF      
+      max_binlog_dump_events = atoi(optarg);
+#endif      
+      break;
+
     case (int) OPT_LOG_SLAVE_UPDATES:
       opt_log_slave_updates = 1;
       break;
diff --git a/sql/slave.cc b/sql/slave.cc
index a768e8494a0..9b9a4041616 100644
--- a/sql/slave.cc
+++ b/sql/slave.cc
@@ -1236,6 +1236,8 @@ pthread_handler_decl(handle_slave,arg __attribute__((unused)))
     goto err;
   }
   
+connected:
+  
   while (!slave_killed(thd))
   {
       thd->proc_info = "Requesting binlog dump";
@@ -1278,10 +1280,9 @@ try again, log '%s' at postion %s", RPL_LOG_NAME,
 	      goto err;
 	    }
 
-	  continue;
+	  goto connected;
 	}
 
-
       while(!slave_killed(thd))
 	{
 	  thd->proc_info = "Reading master update";
@@ -1291,7 +1292,6 @@ try again, log '%s' at postion %s", RPL_LOG_NAME,
 	      sql_print_error("Slave thread killed while reading event");
 	      goto err;
 	    }
-
 	  	  
 	  if (event_len == packet_error)
 	  {
@@ -1329,8 +1329,9 @@ reconnecting to retry, log '%s' position %s", RPL_LOG_NAME,
 reconnect done to recover from failed read");
 	        goto err;
 	      }
-	    break;
-	  }
+	    
+	    goto connected;
+	  } // if(event_len == packet_error)
 	  
 	  thd->proc_info = "Processing master log event"; 
 	  if(exec_event(thd, &mysql->net, &glob_mi, event_len))
@@ -1373,9 +1374,8 @@ the slave thread with \"mysqladmin start-slave\". We stopped at log \
 	        events_till_disconnect++;
 	    }
 #endif	  
-
-	}
-    }
+	} // while(!slave_killed(thd)) - read/exec loop
+  } // while(!slave_killed(thd)) - slave loop
 
   // error = 0;
  err:
diff --git a/sql/sql_repl.cc b/sql/sql_repl.cc
index 7ddf04f1b9a..a5bbe8551cb 100644
--- a/sql/sql_repl.cc
+++ b/sql/sql_repl.cc
@@ -27,6 +27,11 @@
 extern const char* any_db;
 extern pthread_handler_decl(handle_slave,arg);
 
+#ifndef DBUG_OFF
+int max_binlog_dump_events = 0; // unlimited
+bool opt_sporadic_binlog_dump_fail = 0;
+static int binlog_dump_count = 0;
+#endif
 
 static int fake_rotate_event(NET* net, String* packet, char* log_file_name,
 			     const char**errmsg)
@@ -265,8 +270,19 @@ void mysql_binlog_send(THD* thd, char* log_ident, ulong pos, ushort flags)
   int error;
   const char *errmsg = "Unknown error";
   NET* net = &thd->net;
+#ifndef DBUG_OFF
+  int left_events = max_binlog_dump_events;
+#endif  
   DBUG_ENTER("mysql_binlog_send");
 
+#ifndef DBUF_OFF
+  if (opt_sporadic_binlog_dump_fail && (binlog_dump_count++ % 2))
+  {
+    errmsg = "Master failed COM_BINLOG_DUMP to test if slave can recover";
+    goto err;
+  }
+#endif  
+
   bzero((char*) &log,sizeof(log));
 
   if(!mysql_bin_log.is_open())
@@ -325,6 +341,14 @@ void mysql_binlog_send(THD* thd, char* log_ident, ulong pos, ushort flags)
       
     while (!(error = Log_event::read_log_event(&log, packet, log_lock)))
     {
+#ifndef DBUG_OFF
+      if(max_binlog_dump_events && !left_events--)
+      {
+	net_flush(net);
+	errmsg = "Debugging binlog dump abort";
+	goto err;
+      }
+#endif      
       if (my_net_write(net, (char*)packet->ptr(), packet->length()) )
       {
 	errmsg = "Failed on my_net_write()";
@@ -399,6 +423,15 @@ void mysql_binlog_send(THD* thd, char* log_ident, ulong pos, ushort flags)
 
 	bool read_packet = 0, fatal_error = 0;
 
+#ifndef DBUG_OFF
+	if(max_binlog_dump_events && !left_events--)
+	{
+	  net_flush(net);
+	  errmsg = "Debugging binlog dump abort";
+	  goto err;
+	}
+#endif      
+
 	// no one will update the log while we are reading
 	// now, but we'll be quick and just read one record
 	pthread_mutex_lock(log_lock);
diff --git a/sql/sql_repl.h b/sql/sql_repl.h
index 68f2b4ba6c4..c6a79ec4650 100644
--- a/sql/sql_repl.h
+++ b/sql/sql_repl.h
@@ -9,6 +9,11 @@ extern uint32 server_id;
 extern bool server_id_supplied;
 extern I_List<i_string> binlog_do_db, binlog_ignore_db;
 
+#ifndef DBUG_OFF
+extern int max_binlog_dump_events;
+extern bool opt_sporadic_binlog_dump_fail;
+#endif
+
 File open_binlog(IO_CACHE *log, const char *log_file_name,
 	      const char **errmsg);