MySQL Bugs: #48388: additional logging for shutting down threads

Bug #48388	additional logging for shutting down threads
Submitted:	28 Oct 2009 18:43	Modified:	25 Apr 2011 19:06
Reporter:	Lig Isler-Turmelle	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Server: General	Severity:	S4 (Feature request)
Version:		OS:	Any
Assigned to:	Tatiana Azundris Nuernberg	CPU Architecture:	Any

Description:
Request that during the shutdown procedure add logging indicate that threads are being shutdown and when this is complete.  These extra lines will make it clear what is happening if the server shuts down unexpectedly and also how long these particular stages take.

How to repeat:
na

Suggested fix:
Maybe something like:
YYMMDD HH:MM:SS InnoDB: Starting shutdown...
YYMMDD HH:MM:SS Shutting down any running threads <<<<===== Add this line
YYMMDD HH:MM:SS All threads shut down.            <<<<===== Add this line

mysqld_main:
  ...
  mysql_mutex_lock(&LOCK_thread_count);
  while (!ready_to_exit)
    mysql_cond_wait(&COND_thread_count, &LOCK_thread_count);
  mysql_mutex_unlock(&LOCK_thread_count);
  ...
  clean_up(1);
  mysqld_exit(0);

kill_server_thread:
  // created in kill_mysql(). maybe.

handle_shutdown:
  // Win only

kill_server:
  // called from kill_server_thread() / handle_shutdown() / kill_broken_server()
  if MYSQL_KILL_SIGNAL or !signal
    print ER_NORMAL_SHUTDOWN
  else
    print ER_GOT_SIGNAL
  close_connections();
  if (sig != MYSQL_KILL_SIGNAL &&
      sig != 0)
    unireg_abort(1);  // funky signal -> fail -> abort; terminal
  else
    unireg_end();     // regular shutdown; terminal
  // terminal

unireg_abort:
  // called from ... pretty much anywhere, generic fatal handler
  clean_up(...)
  mysqld_exit()   // terminal

unireg_end:
  // called from kill_server(), regular shutdown
  clean_up(1);      // may be terminal ...
  pthread_exit(0);  // if SIGNALS_DONT_BREAK_READ, this is main thr, exit() instead
  // terminal

clean_up:
  // called from mysqld_main(), unireg_abort(), unireg_end()
  ...
  print_inf("shutdown complete")
  ...
  MYSQL_CALLBACK(thread_scheduler, end, ());
  mysql_client_plugin_deinit();
  ...
  mysql_mutex_lock(&LOCK_thread_count);
  DBUG_PRINT("quit", ("got thread count lock"));
  ready_to_exit=1;
  /* do the broadcast inside the lock to ensure that my_end() is not called */
  mysql_cond_broadcast(&COND_thread_count);
  mysql_mutex_unlock(&LOCK_thread_count);
  rts -- if main thread hasn't killed us ...

mysqld_exit:
  // called from unireg_abort(), or mysqld_main() if !EMBEDDED
  wait_for_signal_thread_to_end();
  ...
  clean_up_mutexes();
  clean_up_error_log_mutex();
  ...
  my_end(opt_endinfo ? MY_CHECK_ERROR | MY_GIVE_INFO : 0);
  exit(exit_code); /* purecov: inspected */
  // terminal

my_end:
  // myISAM debugging mostly
  // called from mysqld_exit(), if !EMBEDDED
  ...
  my_thread_end();
  my_thread_global_end();
  rts

my_thread_global_end:
  // wait until all threads have exited, then destroy mutexes
  // called from my_end()
  rts

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/130961

3475 Tatiana Azundris Nurnberg	2011-02-10 [merge]
      merge & fix Bug#48388

correcting, pushed for 5.6.2-m5

Noted in 5.6.2 changelog.

The server now writes thread shutdown messages to the error log
during the shutdown procedure. 

CHANGESET - http://lists.mysql.com/commits/130961