Description:
When the MySQL server was killed without the pid file being removed then trying to stop
the server with 'mysql.server stop' will wait 900 seconds before giving up.
This regularly bites users using a DRBD/Heartbeat cluster. Usually they test the cluster
by manually killing mysqld_safe und mysqld with SIGKILL. When Heartbeat tries to shutdown
the 'mysql' resource it hangs for a long time and eventually kills the mysql resource
script after a timeout.
How to repeat:
Start MySQL using the mysql.server script:
~ $/usr/local/mysql/current/share/mysql/mysql.server start
Starting MySQL done
Now manually kill mysqld_safe and mysqld with kill -9 and then try to stop the MySQL
server:
~ $/usr/local/mysql/current/share/mysql/mysql.server status
MySQL running (12709) done
~ $kill -9 12680
~ $kill -9 12709
~ $/usr/local/mysql/current/share/mysql/mysql.server stop
Shutting down MySQL/usr/local/mysql/current/share/mysql/mysql.server: line 336: kill:
(12709) - No such process
..........
(interrupted with CTRL+C)
BTW, the 'status' method of mysql.server diagnoses right:
~ $/usr/local/mysql/current/share/mysql/mysql.server status
MySQL is not running, but PID file exists failed
Suggested fix:
The mysql.server script should first check if a process with the pid from the pid file
exists before it tries to stop it:
--- mysql.orig 2007-10-23 12:30:25.000000000 +0200
+++ mysql 2007-10-23 12:43:58.000000000 +0200
@@ -332,10 +332,17 @@
if test -s "$pid_file"
then
mysqlmanager_pid=`cat $pid_file`
- echo $echo_n "Shutting down MySQL"
- kill $mysqlmanager_pid
- # mysqlmanager should remove the pid_file when it exits, so wait for it.
- wait_for_pid removed; return_value=$?
+
+ if (kill -0 $mysqlmanager_pid 2>/dev/null)
+ then
+ echo $echo_n "Shutting down MySQL"
+ kill $mysqlmanager_pid
+ # mysqlmanager should remove the pid_file when it exits, so wait for it.
+ wait_for_pid removed; return_value=$?
+ else
+ log_failure_msg "MySQL manager or server process #$mysqlmanager_pid is not
running!"
+ rm $pid_file
+ fi
# delete lock for RedHat / SuSE
if test -f $lock_dir
after applying this patch:
~ $/usr/local/mysql/current/share/mysql/mysql.server stop
MySQL manager or server process #12709 is not running! failed
Description: When the MySQL server was killed without the pid file being removed then trying to stop the server with 'mysql.server stop' will wait 900 seconds before giving up. This regularly bites users using a DRBD/Heartbeat cluster. Usually they test the cluster by manually killing mysqld_safe und mysqld with SIGKILL. When Heartbeat tries to shutdown the 'mysql' resource it hangs for a long time and eventually kills the mysql resource script after a timeout. How to repeat: Start MySQL using the mysql.server script: ~ $/usr/local/mysql/current/share/mysql/mysql.server start Starting MySQL done Now manually kill mysqld_safe and mysqld with kill -9 and then try to stop the MySQL server: ~ $/usr/local/mysql/current/share/mysql/mysql.server status MySQL running (12709) done ~ $kill -9 12680 ~ $kill -9 12709 ~ $/usr/local/mysql/current/share/mysql/mysql.server stop Shutting down MySQL/usr/local/mysql/current/share/mysql/mysql.server: line 336: kill: (12709) - No such process .......... (interrupted with CTRL+C) BTW, the 'status' method of mysql.server diagnoses right: ~ $/usr/local/mysql/current/share/mysql/mysql.server status MySQL is not running, but PID file exists failed Suggested fix: The mysql.server script should first check if a process with the pid from the pid file exists before it tries to stop it: --- mysql.orig 2007-10-23 12:30:25.000000000 +0200 +++ mysql 2007-10-23 12:43:58.000000000 +0200 @@ -332,10 +332,17 @@ if test -s "$pid_file" then mysqlmanager_pid=`cat $pid_file` - echo $echo_n "Shutting down MySQL" - kill $mysqlmanager_pid - # mysqlmanager should remove the pid_file when it exits, so wait for it. - wait_for_pid removed; return_value=$? + + if (kill -0 $mysqlmanager_pid 2>/dev/null) + then + echo $echo_n "Shutting down MySQL" + kill $mysqlmanager_pid + # mysqlmanager should remove the pid_file when it exits, so wait for it. + wait_for_pid removed; return_value=$? + else + log_failure_msg "MySQL manager or server process #$mysqlmanager_pid is not running!" + rm $pid_file + fi # delete lock for RedHat / SuSE if test -f $lock_dir after applying this patch: ~ $/usr/local/mysql/current/share/mysql/mysql.server stop MySQL manager or server process #12709 is not running! failed