mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2026-05-05 06:35:33 +02:00
Code Issues Projects Releases Packages Wiki Activity

BUG#31024: STOP SLAVE does not stop attempted connect()s

Browse source 
Problem: if the IO slave thread is attempting to connect,
STOP SLAVE waits for the attempt to finish. 
It may take a long time.
Fix: don't wait, stop the slave immediately.


sql/slave.cc:
  Send a SIGALRM signal to the slave thread when stopping it (using
  pthread_kill()). This breaks current socket(), connect(), poll() etc.
  calls, and makes the subsequent thd->awake() call effective.
  
  Also, move the definition of KICK_SLAVE to slave.cc.
sql/sql_repl.h:
  Removed KICK_SLAVE and inlined it in slave.cc because:
   - it was only called once, so better to make it local to where it is used
   - it needed to include a preprocessor conditional in the middle
This commit is contained in:
unknown 2008-03-11 14:42:54 +01:00
commit 1836625fb4
2 changed files with 14 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

15
sql/slave.cc
View file

@ -699,7 +699,20 @@ int terminate_slave_thread(THD* thd, pthread_mutex_t* term_lock,
while (*slave_running) // Should always be true
{
DBUG_PRINT("loop", ("killing slave thread"));
KICK_SLAVE(thd);
pthread_mutex_lock(&thd->LOCK_delete);
#ifndef DONT_USE_THR_ALARM
/*
Error codes from pthread_kill are:
EINVAL: invalid signal number (can't happen)
ESRCH: thread already killed (can happen, should be ignored)
*/
IF_DBUG(int err= ) pthread_kill(thd->real_id, thr_client_alarm);
DBUG_ASSERT(err != EINVAL);
#endif
thd->awake(THD::NOT_KILLED);
pthread_mutex_unlock(&thd->LOCK_delete);
/*
There is a small chance that slave thread might miss the first
alarm. To protect againts it, resend the signal until it reacts