[ClusterLabs] [Question] Question about mysql RA.

[Ken Gaillot]

On 11/04/2015 04:36 AM, renayama19661014 at ybb.ne.jp wrote:
> Hi All,
> 
> I contributed a patch several times about mysql, too.
> 
> I did not mind it very much before, but mysql RA makes next move.
> 
> Step1) Constitute a cluster using mysql in Pacemaker.
> Step2) The mysql process kill by signal SIGKILL.
> Step3) Stop Pacemaker before monitor error occurs and stop mysql
> 
> As a result, mysql RA causes trouble in stop.
> By this trouble, Pacemaker does not stop until escalation.
> 
> 
> The cause is processing unlike pgsql RA.
> When a process of pid does not exist, in the case of a stop not to go by way of monitor trouble, mysql RA produces ERR_GENERIC.
> When a process of pid does not exist, pgsql becomes the success of the stop.
> 
> ---------------------------------------------------------
> * mysql
> (snip)
> mysql_monitor() {
>     local rc
>     local status_loglevel="err"
> 
>     # Set loglevel to info during probe
>     if ocf_is_probe; then
>         status_loglevel="info"
>     fi
>  
>     mysql_common_status $status_loglevel
> 
>     rc=$?
> 
>     # TODO: check max connections error
> 
>     # If status returned an error, return that immediately
>     if [ $rc -ne $OCF_SUCCESS ]; then
>         return $rc
>     fi
> (snip)
> mysql_stop() {
>     if ocf_is_ms; then
>         # clear preference for becoming master
>         $CRM_MASTER -D
> 
>         # Remove VIP capability
>         set_reader_attr 0
>     fi
> 
>     mysql_common_stop
> }
> (snip)
> 
> * mysql-common.sh
> (snip)
> mysql_common_status() {
>     local loglevel=$1
>     local pid=$2
>     if [ -z "$pid" ]; then
>         if [ ! -e $OCF_RESKEY_pid ]; then
>             ocf_log $loglevel "MySQL is not running"
>             return $OCF_NOT_RUNNING;
>         fi
> 
>         pid=`cat $OCF_RESKEY_pid`;
>     fi
>     if [ -d /proc -a -d /proc/1 ]; then
>         [ "u$pid" != "u" -a -d /proc/$pid ]
>     else
>         kill -s 0 $pid >/dev/null 2>&1
>     fi
> 
>     if [ $? -eq 0 ]; then
>         return $OCF_SUCCESS;
>     else
>         ocf_log $loglevel "MySQL not running: removing old PID file"
>         rm -f $OCF_RESKEY_pid
>         return $OCF_NOT_RUNNING;
>     fi
> }
> (snip)
> mysql_common_stop()
> {
>     local pid
>     local rc
> 
>     if [ ! -f $OCF_RESKEY_pid ]; then
>         ocf_log info "MySQL is not running"
>         return $OCF_SUCCESS
>     fi
> 
>     pid=`cat $OCF_RESKEY_pid 2> /dev/null `
>     /bin/kill $pid > /dev/null

I think before this line, the RA should do a "kill -0" to check whether
the PID is running, and return $OCF_SUCCESS if not. That way, we can
still return an error if the real kill fails.

>     rc=$?
>     if [ $rc != 0 ]; then
>         ocf_exit_reason "MySQL couldn't be stopped"
>         return $OCF_ERR_GENERIC
>     fi
> (snip)
> ---------------------------------------------------------
> 
> The mysql RA does such a code from old days.
>  * http://hg.linux-ha.org/agents/file/67234f982ab7/heartbeat/mysql
> 
> Does mysql RA know the reason becoming this made?
> Possibly is it a factor to be conscious of mysql cluster?
> 
> I think about a patch of this movement of mysql RA.
> I want to know the detailed reason.
> 
> Best Regards,
> Hideo Yamauchi.