Re: test

[Martin Pala]

Jan-Henrik Haukeland wrote:

> Martin Pala <address@hidden> writes:
>
>  
>
> >  check process apache with pidfile /var/run/httpd.pid
> >    start program = "/etc/init.d/apache start"      stop program =
> >    "/etc/init.d/apache stop"
> >    if failed port 80 protocol http then restart
> >    alert address@hidden
> >    depends on mysql
> >
> >  check host mysql on address 10.1.1.1
> >    if failed port 3306 then stop
> >    alert address@hidden
> >
> >  check process mmp with pidfile /var/run/mmp.pid
> >    start program = "/etc/init.d/mmp start"      stop program =
> >    "/etc/init.d/mmp stop"
> >    if failed port 110 protocol pop then restart
> >    alert address@hidden
> >    depends on pop-backend1
> >    depends on pop-backend2
> >
> >  check host pop-backend1 on address 10.1.1.1
> >    if failed port 110 then stop
> >    alert address@hidden
> >
> >  check host pop-backend2 on address 10.1.1.2
> >    if failed port 110 then stop
> >    alert address@hidden
> >
> >
> > As you can see, host 10.1.1.1 provides pop3 as well as mysql, first
> > syntax is more general and allows such setups.
> >    
>
> Good point, with my proposal you can only have *one* check-host entry
> in the control file per hostname/address. And the above example will
> not work with this proposal. But on the other hand, I could argue that
> you can roll-up the above entries into the process entry like so :-)
>
>  
>
> >  check process apache with pidfile /var/run/httpd.pid
> >    start program = "/etc/init.d/apache start" 
> >    stop program =  "/etc/init.d/apache stop"
> >    if failed port 80 protocol http then restart
> >    
>      if failed host 10.1.1.1 port 3306 then stop 
>  
>
> >    alert address@hidden
> >    depends on mysql
> >
> >  check process mmp with pidfile /var/run/mmp.pid
> >    start program = "/etc/init.d/mmp start"      stop program =
> >    "/etc/init.d/mmp stop"
> >    if failed port 110 protocol pop then restart
> >    alert address@hidden
> >    
>      if failed host 10.1.1.1 port 110 then stop
>      if failed host 10.1.1.2 port 110 then stop
Nice shortcut :) On the other hand if you specify remote connection test 
in its own monitoring service instance, you can use remote monitoring 
specific tests (such as mentioned icmp), which need not to be supported 
in process monitoring class.

> But, yes, you will miss out on the possibilities a depend tree can
> provide. Hmm, I still like my proposal, but I can see that it's not so
> flexible. How about a compromise? I think that it is a possibility
> that host in your proposal could be mistaken for a real DNS host
> record and not as a descriptive name, which in fact it is, so maybe we
> should rewrite the statement to something like this?
>
>   check connection pop-backend2 on host 10.1.1.2 # or host xzzy.org
>     if failed port 110 then alert
>     alert address@hidden
No problem - both notation (hostname or IP address) are usefull - in the 
case that the hostname is in hosts table (which is usual for critical 
hosts) there will be no problem. If the monit user wants not to depend 
on hosts table, he can choose raw form (IP address).

> PS. I know we are in a freeze, and I know that I'm shooting myself in
> the foot when I suggest that we add this remote host test in the
> upcomming 4.0 release. But the reason is that it's very useful, and it
> is easy to realize. It will take me (or anyone of you) only about 1-2
> hours to add this check.
You are rigth, the modification will be very simple and it is probably 
better to add it to 4.0. You choose :)

> PPS. But the ICMP test you suggested in the other mail should be
> postponed to a later release. It could be fun and I even have code
> available for building a handmade ICMP ip-packet using a raw socket.
>
>  
Super :)

Martin