Re: [monit] How to Achieve "if x restarts within y cycles then exec ..."

[Martin Pala]

Art Age Software wrote:
> > Yes, this could work. Optionally it could be possible to use the pidfile
> > using similar approach:
> >
> > --8<--
> > check process mysql with pidfile /var/run/mysqld/mysqld.pid
> >   mode manual
> >   every 2 cycles
> >   if failed unix "/var/lib/mysql/mysql.sock" then restart
> >   if failed host localhost port 3306 protocol mysql then restart
> >   if 5 restarts within 10 cycles then timeout
> >
> > check file mysql_pidfile with path /var/run/mysqld/mysqld.pid
> >   mode manual
> >   if changed timestamp for 5 cycles then exec
> > "/usr/lib/heartbeat/hb_standby"
> > --8<--
> >
> > On restart the script will change the timestamp of the pidfile => the second
> > service check can trigger failover/standby based on it.
>
> I like your approach better because it seems more reliable. With my
> approach, I think it is possible for the mysql protocol checks to
> succeed if the timing is just right (i.e. the check happens between
> restart and crash).
>
> With your approach, should the condition match exactly for both
> checks? For example:
>
> if 5 restarts within 10 cycles then timeout
> if changed timestamp 5 times in 10 cycles then exec "hb_standby"
>
> Will it then be guaranteed that if the restart/timeout condition
> occurs, then the changed timestamp condition/action will also occur?
> Or is there any potential for funny timing anomalies?


I have not tested it, i recommend using -v monit option to tune the 
solution ;) If you'll need help, let me know.

Thanks,
Martin